Systems and methods for near infrared spectroscopy
A flexible NIRS device with multiple light sources and detectors addresses the limitations of current cerebral oximeters by detecting multiple clinical measurements and conforming to the forehead curvature, enhancing signal integrity and data acquisition for diverse age groups.
Patent Information
- Application Number
- PCT/US2026/027174
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-07
- Filing Date
- 2026-05-07
- Publication Date
- 2026-11-12
AI Technical Summary
Current cerebral oximeters provide limited clinical measurements, such as oxygen saturation, and have rigid substrates that fail to accommodate the natural curvature of the forehead, leading to diminished signal integrity, especially in subjects with faster heart rates like neonates and children.
A flexible NIRS device with multiple light sources and detectors, coupled to a computing device, that emits light to detect various tissue chromophores, including oxyhemoglobin, deoxyhemoglobin, and oxidized cytochrome c oxidase, and determines intra-arterial blood pressure and intracranial pressure, using a flexible substrate to conform to the forehead curvature.
The device enables non-invasive detection of multiple clinical measurements, including oxygen saturation, total hemoglobin, and intracranial pressure, with improved signal integrity and data acquisition rates suitable for various age groups, including neonates and children.
Smart Images

Figure US2026027174_12112026_PF_FP_ABST
Abstract
Description
Atty. Dtk. No. 125141.05019. MGH2O24-394SYSTEMS AND METHODS FOR NEAR INFRARED SPECTROSCOPY CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application claims priority to U. S. Patent Application No. 63 / 801,697 filed May 7, 2025, and entitled, “FlexNIRS,” U. S. Patent Application No. 63 / 801,700 filed May 7, 2025, and entitled, “System for and Method of Evaluating Glymphatic Health and Function Using EEG And NIRS,” U. S. Patent Application No. 63 / 801,703 filed May 7, 2025, and entitled, “System for and Method Of Determining Tissue Chromophore Content Given a Specific Hardware Configuration,” U. S. Patent Application No. 63 / 801,706 filed May 7, 2025, and entitled, “Flexible PCB With Frontal EEG And NIRS For Cerebral Health Monitoring,” and U. S. Patent Application No. 63 / 801,709 filed May 7, 2025, and entitled, “Flexible PCB Frontal EEG Based Sleep Staging and Feature Identification,” each which is hereby incorporated by reference in its entirety.STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH
[0002] This invention was made with government support under 5R01GM116177-04, 1R21AG072481-01, 5R21NS137436-02, and 1R21NS137436-01 awarded by the National Institutes of Health (“NIH”), and government support under AY2AX000007 (agreement 2024A014963), awarded by the Advanced Research Projects Agency for Health (“ARPA-H”). The government has certain rights in the invention.BACKGROUND
[0003] Cerebral oximetry measure issue oxygen saturation and can be helpful when monitoring patients. For example, cerebral oxygen saturation (SO2) can provide insight into the current health of the patient. However, cerebral oxygen saturation is only one clinical measurement. Therefore, it would be desirable to have improved systems and methods for near infrared spectroscopy.SUMMARY OF THE DISCLOSURE
[0004] In some examples, the present disclosure provides a near infrared spectroscopy (NIRS) device including a substrate having a body and an arm extending from a side of theAtty. Dtk. No. 125141.05019. MGH2O24-394body, the arm being bendable relative to the body, a light source coupled to the arm, and a light detector coupled to the arm.
[0005] In some examples, the arm can include a base, a first finger extending from the base, and a second finger extending from the base, and the device can further include a first group of light sources, including the light source, and a second group of light sources, where the first group of light sources is substantially similar to the second group of light sources.
[0006] In some examples, the first group of light sources can have the same number and type of light sources as the second group of light sources.
[0007] In some examples, the first group of light sources can include two different colored light sources, and the second group of light sources can include two different colored light sources.
[0008] In some examples, the first group of light sources can include four different colored light sources, eight different colored light sources, or nine different colored light sources.
[0009] In some examples, the first group of light sources can include nine different colored light sources and the second group of light sources can include nine different colored light sources.
[0010] In some examples, at least one light source of the first group of light sources can be a light source having a wavelength of about 735 nm, about 765 nm, about 785 nm, about 805 nm, about 830 nm, about 855 nm, about 870 nm, or about 910 nm.
[0011] In some examples, the arm can include a third finger and a fourth finger extending from the base, the light detector can be a first light detector, the device can further include a second light detector, the first light detector can be coupled to the third finger, the second light detector can be coupled to the fourth finger, the first finger can be substantially parallel to the third finger, and the second finger can be substantially parallel to the fourth finger.
[0012] In some examples, the device can further include a group of light sources coupled to a distal end of the arm, where the group of light sources includes the light source.
[0013] In some examples, the light detector can be a first light detector, and the device can further include a second light detector coupled to the arm, where the first light detector, the second light detector, and the group of light sources can be aligned with each other.
[0014] In some examples, the device can further include a computing device coupled to the substrate at the base.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0015] In some examples, the substrate can be a printed circuit board (PCB), and the PCB can be flexible.
[0016] In some examples, the present disclosure provides a NIRS device including one or more light sources, one or more light detectors, and a computing device in communication with the one or more light sources and the one or more light detectors, where the computing device can be configured to cause the one or more light sources to emit light toward tissue of a subject, receive optical data from the one or more light detectors in response to the emitted light, extract a photoplethysmogram (PPG) waveform from the optical data, receive a derivative PPG waveform as a time derivative of the PPG waveform, and determine an intraarterial blood pressure measurement based on the derivative PPG waveform.
[0017] In some examples, the computing device can be configured to present the intraarterial blood pressure measurement on a display or notify a user of the intraarterial blood pressure measurement of the subject.
[0018] In some examples, the computing device can be configured to provide the derivative PPG waveform to a machine learning model trained to determine an intraarterial blood pressure measurement and receive the intraarterial blood pressure measurement from the machine learning model.
[0019] In some examples, the present disclosure provides a NIRS device including one or more light sources, one or more light detectors, and a computing device in communication with the one or more light sources and the one or more light detectors, where the computing device is configured to cause the one or more light sources to emit light towards tissue of a subject, receive, from the one or more light detectors, optical data in response to the emitted light, extract a photoplethysmogram (PPG) waveform from the optical data, receive a derivative PPG waveform as the time derivative of the PPG waveform, and determine an intraarterial blood pressure measurement, based on the derivative PPG waveform.
[0020] In some examples, the computing device can be configured to determine a first time to a systolic peak and a second time to a diastolic peak for the PPG waveform, and determine that the subject has the cognitive impairment when the first time is above a first time threshold or when the second time is above a second time threshold.
[0021] In some examples, the present disclosure provides a NIRS device including one or more light sources, one or more light detectors, and a computing device in communicationAtty. Dtk. No. 125141.05019. MGH2O24-394with the one or more light sources and the one or more light detectors, where the computing device can be configured to cause the one or more light sources to emit light towards tissue of a subject, receive, from the one or more light detectors, optical data in response to the emitted light, extract a photoplethysmogram (PPG) waveform from the optical data, and determine that the user has a cognitive impairment, based on the PPG waveform.
[0022] In some examples, the computing device can be configured to determine that the subject has ischemia or an occluded blood vessel when a second amplitude of the PPG waveform at a second time occurring after a first time is less than a first amplitude of the PPG waveform at the first time.
[0023] In some examples, the present disclosure provides a system including a pulsed light source, an image sensor, one or more NIRS light sources, one or more NIRS light detectors, and a computing device in communication with the pulsed light source, the image sensor, the one or more NIRS light sources, and the one or more NIRS light detectors, where the computing device can be configured to cause the one or more light sources to emit light towards tissue of a subject, receive, from the one or more light detectors, optical data in response to the emitted light, extract a photoplethysmogram (PPG) waveform from the optical data, and determine that the user has ischemia or an occluded blood vessel, based on the PPG waveform.
[0024] In some examples, the blood flow waveform can be a cerebral blood flow waveform and the internal pressure can be an intracranial pressure.
[0025] In some examples, the computing device can be configured to determine a presence of a brain injury based on the intracranial pressure.
[0026] The foregoing and other aspects and advantages of the present disclosure will appear from the following description. In the description, reference is made to the accompanying drawings that form a part hereof, and in which there is shown by way of illustration one or more exemplary versions. These versions do not necessarily represent the full scope of the disclosure.BRIEF DESCRIPTION OF THE DRAWINGS
[0027] The following drawings are provided to help illustrate various features of nonlimiting examples of the disclosure, and are not intended to limit the scope of the disclosure or exclude alternative implementations.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0028] FIG. 1 shows a block diagram of a NIRS system.
[0029] FIG. 2 shows a top view of another NIRS device.
[0030] FIG. 3 shows an example of a data acquisition scheme for the NIRS device of FIG. 2.
[0031] FIG. 4 A and 4B show an example of another NIRS device.
[0032] FIG. 5 shows a schematic illustration of a bock diagram of a system for detecting internal pressure within the body of a subject (e.g., intracranial pressure).
[0033] FIG. 6 shows an example of a data acquisition scheme for the system of FIG. 5.
[0034] FIG. 7 shows a schematic of the system of FIG. 5.
[0035] FIG. 8 shows a flowchart of a process for detecting blood pressure.
[0036] FIG. 9 shows a flowchart of a process for detecting an internal pressure (e.g., intracranial pressure).
[0037] FIG. 10A-B show Pictures of Gen 1 (a) and Gen 2 (b) FlexNIRS modules.
[0038] FIG. 11A-B show a dimensional drawing of the symmetric multi-distance optical probe, b) Pulse oximetry configuration with only one source and one detector.
[0039] FIG. 12A-B show a complete setup circuit of (a) Gen 1 FlexNIRS. with a second probe expansion b) Gen 2 FlexNIRS circuit with a second probe expansion.
[0040] FIG. 13 shows pictures of FlexNIRS packaging.
[0041] FIG. 14 shows an example of fMRI cortical BOLD (green-volume) and CSF (purple-flow) signals during wakefulness (left) and sleep (right), as reported in FIG. 16A- B of Bottom: Example of FlexNIRS optical density signals at 850 nm (volume) recorded from a subject during REM (left) and deep sleep (right), as indicated by a sleep tracker wristband. The LOD oscillations observed during deep sleep closely resemble the low-frequency BOLD oscillations associated with glymphatic activity.
[0042] FIG. 15A-C show photographs of the latest 2-color FlexNIRS prototype, b) Specs of the proposed LEDs and PD components for the GS-FlexNIRS. c) Simplified block diagram of the proposed GS-FlexNIRS device.
[0043] FIG. 16A-B shows estimated recovery of chromophores changes using Monte Carlo simulation of a realistic head geometry. The noise properties for the current (+970 nm) and the planned GS-FlexNIRS devices were used. When changing concentration of brain HbO (top), HbR (middle), H2O (bottom) with the selected wavelengths, chromophoresAtty. Dtk. No. 125141.05019. MGH2O24-394concentration changes are recovered with minimal crosstalk. Moreover, the improved SNR of the GS-FlexNIRS significantly reduces the variance of the recovered concentrations, b) Data collected at 100Hz and 3.3 cm separation on the forehead of a volunteer with a modified I hex X IRS using the proposed PDs and 2 LEDS at 980 nm (SMT980D). The SNR is comparable to the one achieved at shorter wavelengths and both cardiac and lower frequency signals are present at this water sensitive wavelength. Further improvement of SNR is expected with the new high-power LEDs and higher current AFE.
[0044] FIG. 17A-C show a multi -taper spectrogram of the long-separation FlexNIRS optical density (OD at 850 nm, after short-separation regression) during sleep. Gray area corresponds to motion artefacts. The cardiac, respiratory, and glymphatic frequencies can be seen to vary in both amplitude and frequency throughout the night, b) Time domain representation of the glymphatic frequency band, with c) extracted segments showing examples of lower amplitude fluctuations, higher amplitude fluctuations and a transition from low to high.
[0045] FIG. 18A-D show?a photo of the (a) 8-color FlexNIRS prototype, b) Photographs of the FlexNIRS prototype embedded in an infant cap. c) Marubeni LED package and measured spectra, d) Recovered Ink spectra in an intralipid / water solution. The high SNR of the device allows to recover small changes in absorption at 8 wavelengths.
[0046] FIG. 19 shows a simplified block diagram of the proposed FlexNIRS version 2 of the oxCCO-NIRS device molded case and embedded in a fabric.
[0047] FIG. 20A-B show (a) Mean arterial blood pressure (MAP) estimates from FlexNIRS, based on pulsation waveform feature analysis (P2 / P1) during a carotid endarterectomy (CEA) in an adult patient, showing a strong correlation with arterial-line readings (R2=.87) and (b) Accumulative results in 6 CEA patients showing (bl) the prediction vs. the true target blood pressures (R2=0.83) and (b2) corresponding Bland- Altman plot (LoA: 95% Limit of agreement; CI: 95% confidence interval).
[0048] FIG. 21 shows that consistent with adult data, a higher P2 / P1 ratio found by FlexNIRS in preterm infants indicated elevated blood pressure, as in arm sphygmomanometer readings (purple dots).
[0049] FIG. 22A-C shows a currently developed FlexNIRS and (b) modification plan, (c) Small FlexNIRS can be embedded into an infant cap and (d) operated wirelessly via a tablet.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0050] FIG. 23 A-C shows a schematic of the fiber-based pulsed source SCOS system with the detection fiber bundle rectangle distributed at the camera end. MMF=multimode fiber, FB= fiber bundle b) Light intensity, K and BFi measured with a sCMOS camera at 2.5 cm SDS using CW illumination (top) and 10% duty cycle pulsed illumination (bottom), both with an average power of ~33mW. Camera T 2 ms and 46 Hz fps. The 10 times higher power of the pulsed source increase SNR. lOx. c) Pictorial representation of the data analysis processing stream: The raw intensity image from the detection fiber bundle at time t is defocused to reduce the non-uniform intensity distribution over the fiber bundle ROI. An average of 500 dark images is subtracted from the raw intensity, and a region of interest (ROI) is selected. The raw speckle contrast square is calculated for each sliding window of 7x7 pixels within the ROI, to reduce the noise arising from calculated within each window. Then, the bias introduced by shot noise K2, read noise K2, spatial non-uniformity noise K2, and digitization error K2is removed for each 7x7 pixel window. Finally, a weighted average by I2is performed for all the windows within the ROI to obtain a single value of K2(t) for the camera frame.
[0051] FIG. 24A-B shows a comparison of pCBFi (4 heartbeats) acquired on the forehead of different subjects with the fiber-based light-pulsed Basler SCOS prototype (top), a dual illumination 4 detectors SNSPD-DCS at 1064 nm (middle), and a conventional 3 channel DCS system (bottom). In all cases light power is within ANSI standard limits, signals are processed at ~50Hz, no smoothing averages applied. Despite being at a 1cm shorter source-detector separation the pCBFi resolved with the conventional DCS is much noisier pCBFi with SCOS is comparable to the one acquired with the most expensive and best performing DCS system currently available in the field, b) Simulated cerebral perfusion measurement performances of the Basler fiber-based light-pulsed SCOS (T 1ms, fpslOOHz, peak power 380 mW, duty cycle 10%, s / p:::0.7), vs. SNSPD-DCS at 1064 nm and conventional DCS at 850 nm, both with 1 source and 1 detector and illumination at ANSI limits. In the panels we report cerebral sensitivity (top), CoV (standard deviation of the fit results / average value of the fit results) (middle) and CNR=sensitivity / CoV (bottom), for detecting a change in CBF at 10 Hz using a 3D head model (extracerebral thickness of 1.5 cm) with typical tissue properties. Pulsed fiber-based SCOS achieves orders of magnitude lower CVR and higher CNR than DCS across all SDS. SCOS CNR benefits are achieved despite the lower brain sensitivity of SCOS.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0052] FIG. 25A-C shows a fiber-based SCOS FlexNTRS implementation a) Simplified block diagram (shown unilateral); b) Acquisition scheme for bilateral acquisition at 100Hz sampling rate; c) Schematic drawing of the set up. The SCOS fibers have pictorially rendered into an actual FlexNIRS probe.
[0053] FIG. 26A-C shows the implementation of the contact SCOS FlexNIRS a) Simplified block diagram; b) photograph of a FlexNIRS system integrated with a Basler camera (Sony IMX392 sensor, 6.6mm x 4.2 mm, larger than what we plan to use) to help visualize final device. The SCOS source and short separation detector fibers have pictorially added, c) Preliminary results of contact SCOS, SDS 3.5 cm, fr 100Hz, T 1ms.
[0054] FIG. 27 shows an example of DCS data collected on the forehead of a TBI patient at a SDS of 2.5 cm, together with arterial blood pressure and intracranial pressure. A 6 sec average and 0.05 Hz low pass filter is applied to CBFi, ICP and ABP data. CrCP is calculated gate averaging 40 s of data (100Hz g2) as in. CBFi in this patient strongly correlate with ABP (R=0.88), indicating loss of cerebral autoregulation. The correlation between ICP and DCS CrCP is statistically significant (R=0.18), but the strong CrCP averaging makes it difficult to identify the intracranial pressure spikes. In the bottom inserts we show the zoomed-in raw ICP data (250 Hz) during a spike (left) and during normal ICP (right). The 3 cardiac peaks in the pressure wave are very distinct, with P2 always higher than Pl, indicating poor vascular compliance, which gets considerably worse during the pressure spikes. The DCS pCBFi shown in green is not able to resolve the 3 cardiac peaks because the large averaging needed dampen and corrupt the cardiac peaks.
[0055] FIG. 28A-D shows a) NIRS-PPG, d / dt( NIRS-PPG), b) Dt pCBFi and pCBFi-fit, and c) blood inflow and blood outflow waveforms obtained averaging 4 heartbeats (subject 10 of the lower left image). Dashed lines in panels a and b are the standard deviation over 1 minute rest, e) Coefficient of variation of pCBFi and pCBFi-fit across 10 subjects. Pl = systolic peak, P2 = tidal peak, DN = dicrotic notch, DP = diastolic peak.
[0056] FIG. 29A-B shows BFi comparison of fiber based SCOS at 3 cm and conventional DCS at 2.5 cm (average of 4 detectors) on a intralipid solution during temperature titration, a) 5 s averaging vs. time, b) 1 min averaging vs temp. Exposure time 5 ms, pa=0.1 cm'1, ps’=9.4 cm"1. DCS is much noisier than SCOS despite the shorter SDS, SCOS BFi changes agree with DCS.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0057] FIG. 30 shows a flow of a patient study.
[0058] FIG. 31 shows a schematic illustration of PPG to access vascular age.
[0059] FIG. 32 shows a schematic illustration of PPG to access cerebrovascular health.
[0060] FIG. 33 shows a schematic illustration of the pulsatile signal as an index to perfusion.
[0061] FIG. 34 shows a schematic illustration of PPG for non-invasive arterial pressure monitoring.
[0062] FIG. 35 shows a schematic illustration quantifying critical closing pressure (“CrCP”) with diffuse correlation spectroscopy (“DCS”).
[0063] FIG. 36 shows a schematic illustration illustrating the relationship between NIRS-PPGand pCBFi.
[0064] FIG. 37 shows a schematic illustration detaining CrCP changes at four second temporal resolution.
[0065] FIG. 38 shows a schematic illustration detaining measuring cytochrome c oxidase using NIRS and specifically the FlexNIRS.
[0066] FIG. 39 shows graphs detailing the glymphatic system activity during sleep using NIRS and specifically the FlexNIRS.
[0067] FIG. 40 shows graphs detailing glymphatic system oscillations during sleep using NIRS and specifically the FlexNIRS.
[0068] FIG. 41 shows a schematic illustration of a system for FlexNIRS validation.
[0069] FIG. 42 shows a schematic illustration of a combined SNSPD-DCS and FlexNIRS system.
[0070] FIG. 43 shows a graph detailing the relationship between NIRS PPG and pCBFi.
[0071] FIG. 44 shows graphs detailing the enhancement of pCBFi with NIRS-PPG (e.g., from FlexNIRS).
[0072] FIG. 45 shows graphs detailing FlexNIRS data during sleep.
[0073] FIG. 46 shows a schematic illustration of the NIRS multi -di stance self-calibrating method.
[0074] FIG. 47 shows another schematic illustration of the NIRS multi-distance selfcalibrating method.
[0075] FIG. 48 shows a block diagram for the MW -Flex NIRS device.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0076] FIG. 49A and 49B show a spectroscopy device, while FIG. 49C shows the wearable spectroscopy device of FIG. 49A and 49B secured to a subject's forehead using a headband and skin-safe tape during in vivo monitoring in a clinical environment.
[0077] FIG. 50 shows a three-dimensional printed phantom container with optical windows on multiple walls, configured to allow simultaneous measurements by the wearable spectroscopy device, a reference frequency-domain NIRS system, and a broadband imaging spectrometer for cross-validation.
[0078] FIG. 51 shows a measured emission spectra for each source in the 8-wavelength LED package and the photodetector responsivity profile.
[0079] FIG. 52A-D shows system stability over an approximately 10-hour operating period, illustrating that optical intensity, recovered effective attenuation coefficients, battery voltage, and LED temperature remained stable throughout the measurement.
[0080] FIG. 53 shows noise equivalent power measurements for each wavelength. FIG. 18 shows recovered effective attenuation coefficients on a solid phantom comparing a per-source multi-distance approach.
[0081] FIG. 54 shows recovered effective attenuation coefficients on a solid phantom comparing a per-source multi-distance approach.
[0082] FIG. 55A-B shows correlation and Bland- Altman agreement analyses across three solid phantoms using the spectrally weighted multi-distance method.
[0083] FIG. 56A-B shows measured changes in optical density and recovered cytochrome concentration changes in a bloodless yeast phantom, illustrating the spectral characteristics of mitochondrial chromophores in the near-infrared range.
[0084] FIG. 57 shows chromophore concentration time traces during the same experiment, demonstrating close agreement between the spectrally weighted method and the reference systems.
[0085] FIG. 58A-D shows in vivo cerebral monitoring data acquired from a subject with transient respiratory distress, including total hemoglobin, cerebral oxygen saturation, changes in oxidized cytochrome c oxidase, heart rate, and peripheral oxygen saturation, illustrating the feasibility of the system 100 for non-invasive cerebral oxygenation and metabolic monitoring using the spectrally weighted method described herein.Atty. Dtk. No. 125141.05019. MGH2O24-394DETAILED DESCRIPTION OF THE PRESENT DISCLOSURE
[0086] As described above, current cerebral oximeters measure blood oxygen saturation, one clinical measurement, but lack the ability to determine other relevant and helpful clinical measurements. For example, such cerebral oximeters only have a pair of light sources (or pair of photodetectors) to detect deoxyhemoglobin and oxyhemoglobin at different wavelengths. Therefore, these approaches fail to detect other tissue chromophores, which could serve as useful clinical measurements, markers, indicators, etc. Current cerebral oximeters also have a fairly low data sampling or acquisition rate, which can be unsuitable for neonatal (or other infants, children, etc.) subjects that have faster heart rates. Further, some cerebral oximeters can have substrates that are too rigid and fail to accommodate the natural curvature of the forehead, thereby diminishing signal integrity.
[0087] Some examples of the disclosure provide advantages to these issues, and others, by providing improved systems and methods for near infrared spectroscopy. For example, some examples of the disclosure provide a near infrared spectroscopy (“NIRS”) device that can include one or more light sources, one or more light detectors, and one or more computing devices in communication with the one or more light sources and the one or more light detectors. The computing devices can cause the light sources to emit light towards tissue of a subject and can receive NIRS data from the one or more light detectors in response to the emitted light. The NIRS data can advantageously be used to derived a variety of clinical measurements (e.g., cerebral clinical measurements) for the subject including oxygen saturation, total hemoglobin, oxidized cytochrome c oxidase (“CCO”), tissue water content, intra-arterial blood pressure, intracranial pressure, etc. Accordingly, a single, lightweight device, coupled to the patient’s tissue (e.g., forehead) can non-invasively (e.g., using probing light and received light) to detect a variety of different clinical measurements for the subject.
[0088] FIG. 1 shows a block diagram of a NIRS system 100, which can include a computing device 102, a NIRS device 104, and sensor(s) 106. The computing device 102 can be implemented in different ways. For example, the computing device can be a tablet, a smartphone, a computer, a personal computer, a monitor, etc.
[0089] The NIRS device 104 can include one or more light sources 108, one or more light detectors 110, a computing device 112, and headwear 114. In some cases, the one or more light sources 108 can include one, two, three, four, five, six, seven, eight, nine, ten, etc., lightAtty. Dtk. No. 125141.05019. MGH2O24-394sources, each of which has a different light emission spectra (e g., wavelength range) for emitted light (e.g., each light source having a different color). In some cases, the combined light emission spectra for the light sources 108 can span a wavelength range. In some cases, the wavelength range can be the near infrared light range (e.g., about 650 nm to about 1000 nm, about 700 nm to about 1000 nm, etc.). In some cases, the light emission spectra of each light source can be evenly distributed throughout the wavelength range. For example, adjacent wavelength peaks for each adjacent light emission spectra can be separated by substantially the same wavelength (e.g., about 30 nm). In some cases, each light source 108 can be a light emitting diode (“LED”).
[0090] Although the NIRS device 104 has been described as including headwear 114 to couple the NIRS device 104 to the tissue 118, in other configurations, the NIRS device 104 can include other coupling devices, mechanisms, etc. For example, the NIRS device 104 can include a clip (e.g., for an ear (an earlobe), an appendage, such as, for example, a finger, a toe, an arm, a leg, etc.) to couple the NIRS device 104 to the tissue 118.
[0091] The one or more light detectors 110 can include one, two, three, four, five, etc., detectors. In some cases, each light detector 110 can be separated from the light sources 108 at different distances. In some cases, including when the NIRS device 104 includes multiple groups of light sources and the NIRS device 104 includes a plurality of light detectors 110, a first group of light sources and a first light detector can be substantially symmetrical to a second group of light sources and second light detector. In this way, the symmetric configuration can avoid the need to perform calibrations. In some configurations, each light detector 108 can be a photodetector (e.g., a photodiode).
[0092] The computing device 112 of the NIRS device 104 can be implemented in different ways. For example, the computing device 112 can be a processor, a controller, a microcontroller, etc. The computing device 112 can be in communication with some or all of the components of the NIRS device 104, to, for example, control the components (e.g., transmit instructions thereto) of the NIRS device 104, receive data from the components, etc. For example, the computing device 112 can cause the light sources 108 to emit light 116 towards tissue 118 (e.g., cerebral tissue) including a tissue chromophore 120 therein (e.g., oxyhemoglobin, deoxyhemoglobin, CCO, water, etc.). The tissue chromophore 120 absorbs the emitted light 116, which is dependent on the wavelength components of the emitted lightAtty. Dtk. No. 125141.05019. MGH2O24-394116. The emitted light 116 is modified by various optical properties including the absorbance by the tissue 118, the tissue chromophore 120, etc., and the residual emitted light 122 is reflected off the tissue 118 (or sometimes transmitted through the tissue 118, when, for example, the light detectors 110 are positioned at a different end of the tissue 118 as the side the emitted light 116 is directed onto) and received by the light detectors 110 as NIRS data. The NIRS data can be used to determine a concentration of the tissue chromophore 120, a photoplethysmography (“PPG”) signal, etc. These concentration and PPG signals can be used to determine other clinical measurements including, for example, intra-arterial blood pressure, intracranial pressure, etc.
[0093] In some cases, the NIRS device 104 can include headwear 114, which can be coupled to each of the components of the NIRS device 104, via for, example, a substrate. For example, the light sources 108, light detectors 110, computing device 112, etc., can be coupled to a substrate, which can then be coupled to the headwear 114. In some cases, the substrate can be flexible (e.g., can be a flexible printed circuit board) to accommodate and conform to the curvature of the head (e.g., forehead) of the subject for better data acquisition. The headwear 114 can couple the NIRS device 104 to the tissue 118 (e.g., a head of a subject). The headwear 114 can be implemented in different ways. For example, the headwear 114 can be a hat, a cap, a headband, a strap, a band, a piece of fabric (e.g., having ends that couple together, such as, for example, via hook and loop fasteners). In some cases, the headwear 114 can be sized for a subject that is a neonatal infant, infant, child, adult, etc.
[0094] As shown in FIG. 1, the sensor(s) 106, which can also be in communication with the computing device 112 and can be part of the NIRS device 104 (e.g., coupled to the substrate of the NIRS device 104). In other cases, the sensor(s) 106 can be separate from the NIRS device 104, but otherwise in communication with the computing device 112. In some cases, the sensor(s) 106 can include a motion sensor (e.g., an accelerometer, a gyroscope, an inertial measurement unit (“IMU”), etc.). The motion sensor can advantageously identify patient movement, such that optical data acquired by the light detectors(s) 110 during the identified patient movement can be deleted, removed, or otherwise, not used to determine a clinical measurement.
[0095] FIG. 2 shows a top view of a NIRS device 150. The NIRS device 150 can be a specific implementation of the NIRS device 104. Accordingly, the description of the NIRSAtty. Dtk. No. 125141.05019. MGH2O24-394device 104 pertains to the NIRS device 150 and vice versa. The NIRS device 150 can include a substrate 152 having a body 154 and arms 156, 158 coupled to opposing ends of the body 154. The substrate 152 can be formed out of different materials. For example, advantageously, the substrate 152 can be flexible (e.g., formed out of a flexible material) so that the features of the substrate 152 can accommodate the curvature of the tissue in which the NIRS device 150 is placed on (e.g., the curvature of the forehead). In some cases, the substrate 152 can be a flexible printed circuit board (“PCB”), in which traces that electrically connect the electrical components of the NIRS device 150 together can be positioned on top of the substrate 152 (along with the electrical components). In some cases, the substrate 152 can include multiple layers. For example, wires that connect the electrical components of the NIRS device 150 can be sandwiched between a top layer of the substrate 152 and a bottom layer 154 of the substrate 152. In some cases, the top layer and the bottom layer can each be an insulation layer (e.g., formed out of an insulating material, such as a polymer). In some cases, a top of the NIRS device 150 (e.g., the side positioned away from the tissue when placed on a subject) can be coated with a flexible material, such as a polymer (e.g., silicone). In some cases, a bottom of the NIRS device 150 (e.g., the side positioned towards the tissue when placed on a subject) can be covered with a flexible material in a similar way as the top side. In some configurations, the flexible material at the bottom can include a plurality of holes, with each hole aligned with respective light source or detector (e g., to ensure that light can pass into the tissue and can be received out of the tissue). In some cases, the flexible material can encapsulate the NIRS device 150. In some cases, the substrate (e.g., the PCB, the flexible PCB, etc.) of the NIRS device 150 can have a thickness that is less than or equal to about 400pm. In other cases, the NIRS device 150 (e.g., the entire NIRS device 150, such as the substrate coated in a material, such as a polymer) can have a thickness that is less than about 5 millimeters, greater than about 1 millimeter, in a range between about 1 millimeter to about 5 millimeters, etc.
[0096] The NIRS device 150 can also include optical probes 160, 162 coupled to opposing ends of the body 154. For example, the optical probe 160 is coupled to a first end of the body 154, while the optical probe 162 is coupled to a second end of the body 154. Each optical probe 160, 162 can include one or more groups of light sources and one or more light detectors. In some cases, each optical probe 160, 162 can receive optical data (e.g., NIRS data) from a respective hemisphere of the brain (e.g., cerebral hemispheres). For example, the optical probeAtty. Dtk. No. 125141.05019. MGH2O24-394160 can acquire optical data from a right hemisphere (e.g., a right cerebral hemisphere) of the subject, while the optical probe 162 can acquire optical data from a left hemisphere (e.g., a left cerebral hemisphere) of the subject. In this way, the NIRS device 150, with the optical probes 160, 162 can acquire bilateral measurements.
[0097] As shown in FIG. 2, the substrate 152 can include arms 156, 158 coupled to opposing ends of the body 154. Each arm 156, 158 can bend away from the body 154, when, for example, the NIRS device 150 is placed on the tissue of the subject (e.g., the forehead). In some cases, each arm 156, 158 can include a respective bridge 164, 166 that couples the respective arm 156, 158 to the body 154. In some cases, each bridge 164, 166 can have a smaller width than the width of the body 154. Each arm 156, 158 can include a plurality of fingers extending therefrom. For example, the arm 156 can include fingers 168, 170, 172, 174, while the arm 158 can include fingers 176, 178, 180, 182. Each finger 168, 170, 172, 174 can extend away from a base 184 of the arm 156 (e.g., in which the base 184 can include the bridge 164), while each finger 176, 178, 180, 182 can extend away from a base 186 of the arm 158 (e.g., in which the base 186 can include the bridge 166). As shown in FIG. 2, the fingers 168, 172 can extend upwardly from the base 184, while the fingers 170, 174 can extend downwardly from the base 184. Similarly, the fingers 180, 176 can extend upwardly from the base 186, while the fingers 178, 182 can extend downwardly from the base 186. In some cases, each finger can advantageously bend away from the respective base, which can conform better to the curvature of the tissue (e.g., forehead). In this way, the probing light from a light source can interact with a greater amount of tissue, while at the same time, the light detector can similarly collect a greater amount of the probing light. In some cases, each finger of each optical probe can be substantially parallel to each other. As shown in FIG. 2, the periphery of each finger can be curved to, for example, avoid corners that could be uncomfortable for the subject.
[0098] Each finger (e.g., at a free end thereof) can include a group of light sources (e.g., or one light source, a single light source, etc.) or a light detector. In this way, the group of light sources or the light detector can bend away from the base of the arm to probe greater amounts of tissue with light or receive greater amounts of the probing light (e.g., each of which can improve the signal to noise ratio of the underlying signals). For example, each finger 168, 170 can each include a respective group of light sources 188, 190 (e.g., positioned at a free endAtty. Dtk. No. 125141.05019. MGH2O24-394thereof). In some cases, each group of light sources 188, 190 can include one or more light sources, and each group of light sources 188, 190 can be substantially identical or identical. That is, each group of light sources 188, 190 can include the same number of individual light sources and the same type of light sources (e.g., the emission spectra of the light sources being substantially the same). In some cases, each group of light sources 188, 190 can include an about 735 nm light source and an about 850 nm light source. In some configurations, each group of light sources 188, 190 can include an about 735 nm light source, about 765 nm light source, about 785 nm light source, about 805 nm light source, about 830 nm light source, about 855 nm light source, about 870 nm light source, about 910 nm light source. In some cases, each group of light sources 188, 190 can include an about 740 nm light source, about 760 nm light source, about 780 nm light source, about 810 nm light source, about 850 nm light source, about 890 nm light source, about 910 nm light source, about 940 nm light source, and an about 1050 nm light source. In some cases, an about a particular wavelength light source means that the light emission spectra emitted from that light source has a peak wavelength (or dominant wavelength) at the particular wavelength. In some configurations, each group of light sources 188, 190 can be a light package, such as, for example, SMT735D / 850D manufactured by Marubeni, or COB-8040-MGH-X manufactured by Marubeni.
[0099] As another example, each finger 172, 174 (e.g., at a free end thereof) can include a light detector 192, 194. In some cases, a light detector 196 can be positioned between the groups of light sources 168, 170 at the base 184. In some cases, each light detector 192, 194, 196 can have a larger surface area than the surface area of each light source of the group of light sources 168, 170, and the collective surface area of the groups of light sources 168, 170. In some cases, therefore, the spatial footprint of each light detector 192, 194, 196 can be larger than the combined spatial footprint of each group of light sources 168, 170. In some cases, the group of light sources 188 and the light detector 192 can be substantially symmetrical to the groups of light sources 188 and the light detector 194. That is, the group of light sources 188 can be positioned substantially the same distance from the base 184 as the group of light sources 190 is from the base 184. Similarly, the light detector 192 can be positioned substantially the same distance from the base 184 as light detector 194 is from the base 184. In a similar way, the light detector 196 is positioned the substantially the same distance away from the group of light sources 188 as the group of light sources 190 is away from the lightAtty. Dtk. No. 125141.05019. MGH2O24-394detector 196. In some cases, the distance between the group of light sources 188 and the light detector 192 can be substantially similar (e.g., about 2.4 cm) can be substantially the same as the distance between the group of light sources 190 and the light detector 194. In some configurations, the distance between the group of light sources 188 and the light detector 194 can be substantially similar (e.g., about 2.8 cm) can be substantially the same as the distance between the group of light sources 190 and the light detector 192. This symmetrical configuration of the light sources and light detectors (e.g., the source detector distances being the same) can prevent the need for calibration. That is, this configuration can be self-calibrating (e.g., errors from one side can be canceled out from errors on the opposing side). In some cases, the optical probe 162 can be implemented in a similar manner as the optical probe 160, and therefore, the optical probe 162 will not be described in detail for brevity (e.g., the optical probe can include two groups of light sources, three light detectors, etc.). In some cases, light sources that are positioned farther away from the middle of the substrate (e.g., the body 154 of the substrate 152) can reduce the crosstalk between each of the optical probes 160, 162. In some configurations, an optical probe (e.g., the optical probe 156) can include a light detector, which can be positioned between the light detectors 192, 194 and can be positioned closer to the base 154 of the substrate 152 than the light detector 196. In some cases, this fourth light detector can allow for an additional distance for the multi-distance, self-calibrating process, which can improve SNR in optical property recovery.
[0100] In some cases, the lack of material between adjacent fingers, and between a finger and the body 154 of the substrate 152 (e.g., each defining a recess), can be advantageous to ensure that probing light from the light sources are received at the light detectors and not blocked by the substrate. That is, if material of the substrate extended laterally between the fingers, the probing light from the light sources may not be adequately received by the light detectors by being blocked, reflected, absorbed, attenuated, etc., through the substrate.
[0101] Although the NIRS device 150 can capture bilateral cerebral data, in other cases, the NIRS device 150 can capture unilateral cerebral data (e.g., from one hemisphere). In this way, for example, the NIRS device 150 can then include one optical probe (e.g., as opposed to two optical probes).
[0102] In some cases, the NIRS device 150 can include a computing device 198 and a communication system 200. The computing device 198 can be implemented in variousAtty. Dtk. No. 125141.05019. MGH2O24-394ways, such as, for example, the computing device 102. For example, the computing device 198 can be a controller, microcontroller, processor, etc. The computing device 198 can be in electrical communication with some or all of the components of the NIRS device 150. For example, the computing device 198 can be electrically coupled (e.g., via wires, electrical traces, etc.) to each component of the optical probes 160, 162 (e.g., each light source, each light detector, etc.). In some cases, the computing device 198 can selectively cause light to be emitted from each light source of a group of light sources, and similarly, can cause optical data to be received from each light detector during activation of the selected light source. In this way, optical crosstalk can be minimized. In some cases, parallel acquisition of optical data from each light detector can be facilitated by a dedicated amplifier (e.g., transimpedance amplifier, a current to voltage converter, which can amplify the current signals indicative of received light from a photodiode as a light detector). That is, each light detector can include a transimpedance amplifier to amplify the captured current signals, which can be digitized by an analog to digital converter (“ADC”) (e.g., in which the digitized data can be the optical data).
[0103] FIG. 3 shows an example of a data acquisition scheme for the NIRS device 150. As shown in FIG. 3, ambient light is captured from each light detector in the form of ambient optical data (e.g., which can be averaged together). Then, while the 735 nm light source emits light, each light detector captures optical data (e.g., which can be averaged together). During emission of light from the 735 nm light source, no other light sources are activated (e.g., emit light) including light from the same group of light sources or the other group of light sources. Then, while the 760 nm light source emits light, each light detector captures optical data (e.g., which can be averaged together). Again, during emission of light from the 760 nm light source, no other light sources are activated (e.g., emitting light). Then, the next light source having the next lowest peak wavelength (e.g., the 780 nm light source) can be activated to emit light, with each light detector receiving optical data. This process proceeds until all the light sources within the first group of light sources have been activated and corresponding optical data has been captured therefrom. Then, a similar process can acquire optical data using each light sources from a second group of light sources. In some cases, optical data corresponding to each light source can be combined (e.g., averaged) with the ambient optical data subtracted therefrom (e.g., to compensate or otherwise remove the ambient light). For example, opticalAtty. Dtk. No. 125141.05019. MGH2O24-394data corresponding to light received during the emission of light by the 735 nm light source of the first group of light sources can be combined with optical data corresponding to light received during emission of light by the 735 nm light source of the second group of light sources.
[0104] In some cases, a computing device can implement a boosted approach to improve the SNR of the underlying NIRS signals. In a boosted approach, for example, multiple samples (e.g., iterations, measurements, etc.) can be completed to, for example, boost the signal of light sources that have a particularly low sensitivity at that particular wavelength. For example, as shown in FIG. 3, when one optical probe is iterating through individual light sources (and receiving optical data therefrom) the other optical probe, during each iteration, can acquire optical data from each photodetector while the 910 nm light source emits 910 nm light. In this way, rather than one optical data sample for the 910 nm light source, nine different optical data samples can be acquired, which can greatly improve the SNR (e.g., by 3 times). This can be particular helpful for wavelengths (e.g.,. higher wavelengths) where the light detector is less sensitive to that particular wavelength of light. Therefore, in some cases, this boosted approach can be implemented for each light source having a wavelength 870 nm or above, including for example, an about 1050 nm light source. In other cases, this boosted approach can be implemented for each light source that has a peak wavelength that results in less than or equal to about 0.8 in the normalized emission profile of the light detector responsivity (e.g., the light at that wavelength resulting in about 0.8). Further, although nine optical data samples have been described, it is appreciated that other number of samples for a NIRS cycle can be implemented, such as, for example, a plurality of optical data samples (e.g., two, three, four, five, six, seven, eight, ten, eleven, etc.).
[0105] Referring back to FIG. 2, the NIRS device 150 can include a communication system 200, which can be coupled to the substrate 152 (e.g., in addition to the computing device 198). The communication system 200 can transmit data to a computing device 202 or others (e.g., optical data including raw optical data, processed optical data, clinical measurements, etc.) and can receive data (e.g., instructions) from the computing device 202 (or others). In some cases, therefore, the communication system 200 can include a transmitter, a receiver, a transceiver, etc. In some cases, the communication system 200 can be a wirelessAtty. Dtk. No. 125141.05019. MGH2O24-394module, such as a Bluetooth® module for transmitting and receiving a Bluetooth® wireless signals (e.g., to the computing device 202 and from the computing device 202).
[0106] In some cases, the NIRS device 150 can include analog front ends (“AFEs”) 204, 206, each of which can be coupled to the substrate 152. The AFE 204 can be in electrical communication with the optical probe 156 to facilitate control of the light sources and conditioning of the optical signals received from the light detectors, while the AFE 206 can be in electrical communication with the optical probe 158 to facilitate control of the light sources and conditioning of the optical signals received from the light detectors. For example, each AFE 204, 206 can include a multiplexer, an analog to digital converter (“ADC”), one or more transimpedance amplifiers, a communication interface (e.g., a serial peripheral interface (“SPI”).
[0107] In some cases, the NIRS device 150 can include a power source 208, which can provide power to some or all of the components of the NIRS device 150. In some cases, the power source 208 can be a battery (e.g., having one or more cells therein), a rechargeable battery, a wired power source (e.g., an AC wire), etc. In some cases, an electrical connector can be coupled to the substrate 152 to removably receive the power source 208. For example, when the power source 208 is a rechargeable battery, the rechargeable battery can be removed from the NIRS device 150 at the electrical connector and can be recoupled to the NIRS device 150 at the electrical connector when the rechargeable battery is sufficiently charged.
[0108] FIG. 4A and 4B show an example of a NIRS device 250, and specifically, a different configuration of an optical probe of the NIRS device 250. Accordingly, the NIRS device 250 relates to the NIRS devices 100, 150, and therefore, the description of the NIRS device 250 pertains to the NIRS devices 100, 150 and vice versa.
[0109] FIG. 4A shows an example of an optical probe 252 of the NIRS device 250. The optical probe 252 can include a substrate 254 that can couple each of the components of the NIRS device 250 together. In some examples, the substrate 254 can be rigid (e.g., a PCB). In some examples, the substrate 254 can be flexible (e.g., a flexible PCB).
[0110] The substrate 254 of the optical probe 252 can include a base 256 and an arm 258 that extends from the base 256. The arm 258 extends from a proximal end coupled to the base 256 to a distal end spaced apart from the base 256. In some examples, the group of light sources 260 are disposed on the distal end of the arm 258 (e.g., at a free end thereof). In some examples,Atty. Dtk. No. 125141.05019. MGH2O24-394the light detectors 262, 264, 266, 268 are coupled to and disposed on the arm 258 (e.g., between the distal end and the proximal end of the arm 258). That is, the light detectors 262, 264, 266, 268 can be disposed between the group of light sources 260 and the base 256. In some examples, the base 256 includes an accelerometer. The accelerometer can send acceleration or orientation data to a computing device to remove motion artifacts from the NIRS data, thereby reducing noise from the NIRS data.
[0111] Referring to FIG. 4B, the group of light sources 260 can include nine light sources. Each light source of the group of light sources 260 can be configured to emit light at a peak wavelength. In some examples, each light source of the group of light sources 260 can also be configured to emit light at one or more additional wavelengths (e.g., at wavelengths less than the peak wavelength). For example, the group of light sources 260 can include a first light source 260a configured to emit light at a first peak wavelength (e.g., about 890 nm), a second light source 260b configured to emit light at a second peak wavelength (e.g., about 850 nm), a third light source 260c configured to emit light at a third peak wavelength (e.g., about 810 nm), a fourth light source 260d configured to emit light at a fourth peak wavelength (e.g., about 780 nm), a fifth light source 260e configured to emit light at a fifth peak wavelength (e.g., about 760 nm), a sixth light source 260f configured to emit light at a sixth peak wavelength (e.g., about 740 nm), a seventh light source 260g configured to emit light at a seventh peak wavelength (e.g., about 910 nm), an eighth light source 260h configured to emit light at an eighth peak wavelength (e.g., about 1050 nm), and a ninth light source 260i configured to emit light at a ninth peak wavelength (e.g., about 940 nm). In some examples, the group of light sources 260 can be arranged in a grid (e.g., a 3x3 grid), array, etc.
[0112] Continuing to refer to FIG. 4A-B, the light detectors 262, 264, 266, 268 can include four light detectors. For example, the light detectors 262, 264, 266, 268 can include a first light detector 262, a second light detector 264, a third light detector 266, and a fourth light detector 268. In some examples, the light detectors 262, 264, 266, 268 can be arranged in a linear array (e g., aligned along a longitudinal axis of the arm 258). In some examples, the first light detector 262, the second light detector 264, the third light detector 266, and the fourth light detector 268 can be aligned with one another along the longitudinal axis of the arm 258. In some examples, the light detectors 262, 264, 266, 268 can be spaced apart from the group of light sources 260 by different source-to-detector distances. For example, the first light detectorAtty. Dtk. No. 125141.05019. MGH2O24-394146a can be spaced apart from the group of light sources 260 (e.g., at a centroid, center, etc., of the group of light sources 260, for example, the center light source, such as for example, the light source 260e) by a first distance, the second light detector 264 can be spaced apart from the group of light sources 260 by a second distance that is greater than the first distance, the third light detector 266 can be spaced apart from the group of light sources 260 by a third distance that is greater than the second distance, and the fourth light detector 268 can be spaced apart from the group of light sources 260 by a fourth distance that is greater than the third source-to-detector distance. In this way, the light detectors 262, 264, 266, 268 can detect light from different light penetration distances. In other words, as the source to detector distance increases, the light from the light source can probe deeper tissues, while the light detector can correspondingly receive this probing light from the deeper tissues. In some cases, the first distance can be about 7.7 mm, the second distance can be about 25 mm, the third distance can be about 30 mm, and the fourth distance can be about 36 mm. In some cases, the second distance can be about two times, three times, four time, etc., the first distance. In some cases, the distance between the light detectors 264, 266, 268 can be substantially or exactly the same. In some configurations, the linear based NIRS device 250 can be advantageous as compared to a self-calibrating NIRS device (e.g., the NIRS device 150). For example, an optical probe having one group of light sources as opposed to two groups of light sources can save a considerable amount of time with signal acquisition speed. This can be particularly important for neonatal, infants, children, etc., that have relatively higher heart rates than adults.
[0113] FIG. 5 shows a schematic illustration of a bock diagram of a system 400 for detecting internal pressure within the body of a subject (e.g., intracranial pressure). The system 400 can receive both NIRS data (e.g., optical data) and can receive speckle contrast optical (“SCOS”) data (e.g., in the form of images) data to determine an internal pressure. The system 400 can include a NIRS device 402, which can be similar to the other NIRS devices described herein (e.g., the NIRS devices 100, 150, 250, etc.). For example, the NIRS device 402 can include a substrate 406, groups of light sources, light detectors, etc. The substrate 406 can include an optical input 408 and an optical output 410. The optical input 408 can receive a pulsed light source (e.g., a laser) to direct the light from the pulsed light source into the tissue (e.g., into the forehead). Similarly, the optical output 410 can receive light (e.g., from the pulsed light source) after the pulsed light has interacted with the underlying tissue.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0114] The system 400 can include a pulsed light source 412 optically coupled to the optical input 408 and an image sensor 412 optically coupled to the optical output 410. In some cases, the pulsed light source 412 can be optically coupled to the optical input 408 via one or more optical fibers (e.g., one optical fiber, a single optical fiber, which can have a width of less than or equal to about 400 pm, 300 pm, 200 pm, etc.), while the image sensor 412 can be optically coupled to the optical output 410 via one or more optical fibers (e.g., an optical fiber bundle). In some cases, the optical fiber bundle 416 optically coupled to the optical output and the image sensor 414 can include a plurality of optical fibers (e.g., greater than about 50, 100, 150, 200 optical fibers), with each optical fiber being a rectangular optical fiber (e.g., a rectangular waveguide). In some cases, each rectangular optical fiber can have a rectangular cross-section. In some cases, each rectangular optical fiber (e.g., a rectangular fiber) can be a square optical fiber (e.g., a square fiber). Each optical fiber of the optical fiber bundle 416 can be less than or equal to about 4 mm, 3 mm, 2.5 mm, etc. In some cases, the rectangular optical fibers of the optical fiber bundle 416 can improve light coupling to the image sensor 414. For example, the rectangular fiber bundles better match the shape of the image sensor (e.g., the CMOS camera sensor) and therefore allow for more effective filling of the full aperture of the image sensor (e.g., camera). Conversely, for a similar surface area circular fiber bundle that fills the full face of the camera, much more light would be lost, leading to a larger SNR degradation. In some cases, one or more optical elements can be optically coupled to the one or more optical fibers optically coupled to the image sensor. In some cases, the one or more optical elements can be a lens (e.g., a diverging lens).
[0115] In some cases, the pulsed light source 412 can emit light have a particular peak wavelength (e.g., about 1064 nm), which can be within the near-infrared light spectrum. That is, the peak wavelength can be greater than the wavelengths within the visible light spectrum (e.g., between 380 nm and 750 nm). In some cases, the peak wavelength of the pulsed light source 412 can be different than the peak wavelength of each light source of the NIRS device 404 (e.g., those that capture NIRS optical data). In some cases, the pulsed light source 412 can be a laser (e.g., a laser diode, a Nd: YAG laser, etc.). In some cases, the pulsed light source 412 can steer the emitted light (e.g., laser beam) to different positions (e.g., a first position, a second position, etc.), using for example, a galvanometer, chopper wheel, etc. In some cases, the first position can be a first X-Y position and the second position can be a second X-Y position. InAtty. Dtk. No. 125141.05019. MGH2O24-394some cases, the image sensor 414 can be a CMOS image sensor, a progressive scan CMOS image sensor, etc. In some cases, the image sensor 414 can be a camera (e.g., the image sensor including one or more other optical components, such as, for example, a lens). In some cases, the pulsed light source 412 can be optically coupled to the image sensor 414 (e.g., in addition to the optical output 410). In some cases, a portion of the pulsed light can be directed to the image sensor 414 and can interact with the pulsed light received after interacting with the tissue. In some configurations, this can be implemented using a beam splitter, where a fraction (e.g., about 1 / 99) of the pulsed light is directed to the image senor 414, and the remaining faction (e.g., about 98 / 99) of the pulsed light received after interacting with the tissue is received by the image sensor 414. In this way, the SNR of the flow signal can be increased (e.g., from the interference between the light).
[0116] In some examples, although optical fibers and optical fiber bundles have been described, in other cases, the pulsed laser source 412 can be coupled to the substrate 406 (e.g., at the optical input 408) and the image sensor 414 can be coupled to the substrate 406 (e.g., at the optical output 410). In other words, the pulsed laser source 412 and the image sensor 414 can be positioned directly on the substrate 406. Further, in some cases, although the pulsed light source 412 and the image sensor 414 has been described with respect to an NIRS device, in other cases, the pulsed light source 412 and the image sensor 414 can be part of a speckle contrast optical spectroscopy (“SCOS”) system, which can function independently of an NIRS device, to, for example, determine internal pressure (e g., an intracranial pressure).
[0117] FIG. 6 shows an example of a data acquisition scheme for the system 400, which can improve the signal to noise ratio (“SNR”) by reducing optical cross-talk between components. The data acquisition scheme shown in FIG. 6 can be implemented using a computing device (e.g., as described herein, such as to drive the light sources, adjust positions of the emitted light, acquire optical data, etc.). As shown in FIG. 6, the pulsed light source emits light at a first position (e.g., a first galvanometer position) during a first time period towards the tissue, while the image sensor receives optical data corresponding to the light from the pulsed light source interacting with the tissue for a second time period, which can be less than the first time period. In some cases, the second time period is an exposure time of the image sensor. Then, the pulsed light source emits light at a second position (e.g., a second galvanometer position) during a third time period towards the tissue, while the image sensorAtty. Dtk. No. 125141.05019. MGH2O24-394receives optical data corresponding to the light from the pulsed light source interacting with the tissue for a fourth time period, which can be less than the first time period. Again, the fourth time period can be an exposure time of the image sensor. During emission of light from the pulsed light source, the NIRS light sources do not emit light (e.g., and the light detectors do not receive light). Then, the NIRS light sources can emit light and receive NIRS optical data (e.g., according to the data acquisition scheme of FIG. 3) for a fifth time period, which can be longer than the first, second, third, and fourth time periods. In some cases, this fifth time period corresponds to a first group of light sources. Similarly, during the fifth time period, the NIRS light sources (e.g., a second group of light sources) can emit light and receive NIRS optical data (e.g., according to the data acquisition scheme of FIG. 3) for a sixth time period, in which this sixth time period can correspond to a second group of light sources. During emission of the NIRS light sources, the pulsed light source does not emit light (e.g., and the image sensor does not acquire image data, images, etc., the shutter is closed, the image sensor is not exposed to light), the first group of light sources do not emit light, etc. In some cases, the first time period (e.g., about 2 ms) can be about double the second time period (e.g., about 1 ms), the third time period (e.g., about 2 ms) can be about double the fourth time period (e.g., about 1 ms), etc. In some cases, the fifth time period (e.g., about 5 ms) can be about double the first time period, about five times the second time period, etc.
[0118] In some cases, the pulsing of the pulsed light source and the exposure time of the image sensor coinciding with the emission time of the pulsed light source can improve SNR of the received images (e.g., speckle images), due to, for example, being able to use a higher optical power during the pulse of light by the pulsed light. Similarly, since the NIRS light sources and light detectors are not active during pulsing of the pulsed light source and the image sensor, optical cross talk between the systems is advantageously reduced (e.g., the NIRS system and the SCOS system).
[0119] FIG. 7 shows a schematic of the system 400, showing an example implementation of the system 400.
[0120] FIG. 8 shows a flowchart of a process 500 for detecting blood pressure (e.g.,. intraarterial blood pressure). The process 500 can be implemented using any of the NIRS devices, systems, etc. Further, the process 500 can be implemented using one or more computing devices. Accordingly, the process 500 can be a computer-implemented method.Atty. Dtk. No. 125141.05019. MGH2O24-394Although the following description of the process 500 will describe a computing devices implementing the process 500, it is appreciated that this can include one or more computing devices to implement the process 500 (e.g., a plurality of computing devices).
[0121] At block 501, the process 500 can include positioning a near infrared spectroscopy (“NIRS”) device on a subject. In some cases, this can include coupling the NIRS device to the subject at a tissue location, such as, for example, the forehead (e.g., when measuring cerebral parameters). In some cases, this can include coupling (e.g., adhering, such as with adhesive) a bottom of the NIRS device to a subject (e.g., a forehead of the subject), such that an emission surface of each light source faces the tissue and a detection surface of each light detector faces the tissue.
[0122] At block 502, the process 500 can include a computing device causing one or more light sources to each emit light towards a tissue (e.g., a forehead). In some cases, the light form each light source can penetrate the tissue and interact with the tissue (e.g., absorb, reflect, etc.). In some cases, this can include sequentially activating each light source of one or more groups of light sources (e.g., two groups of light sources). In some configurations, the sequential activation of each light source can follow the data acquisition scheme according to FIG. 3.
[0123] At block 504, the process 500 can include a computing device receiving optical data from each light detector. In some cases, this can include receiving optical data, from each light detector, during emission by light of each light sources individually (e.g., while only one light source emits light at a given time). In some cases, the optical data is produced from residual light that is received at a light detector, after interacting with the tissue (e.g., being absorbed, reflected, etc.). In some cases, the optical data can be the intensity of the received light (e.g., the intensity of the residual light from a light source after the initially emitted light has interacted with the tissue).
[0124] At block 506, the process 500 can include the computing device combining optical data together. In some cases, this can include combining (e.g., averaging) optical data together from multiple light detectors. This can also include combining optical data acquired during emission of the same type of light sources from each group of light sources. For example, optical data (e.g., combined optical data from multiple light detectors, such as light detectors 192, 194, 196) acquired during emission of light from a first light source of a first group of light source (e.g., the 735 nm light source) can be combined with optical data (e.g., combinedAtty. Dtk. No. 125141.05019. MGH2O24-394optical data from the same light detectors) acquired during emission of light from a first light source of a second group of light source (e.g., the 735 nm light source). In some cases, as described, combining optical data together can cancel out errors in a self-calibrating manner.
[0125] In some cases, the block 506 can include the computing device filtering, conditioning, etc., the optical data. For example, this can include removing the ambient light signal from the optical data. For example, ambient light optical data can be acquired by the one or more light detectors while each light source from each group of light sources is off (e.g., not emitting light towards the tissue). Then, the computing device can subtract each optical data by the ambient light optical data. This can effectively scrub each set of optical data (e.g., derived from probing light) of ambient light noise.
[0126] In some cases, a computing device can provide a warning, notification, etc., for example, based on a quality of the optical data (e.g., if the optical data quality is below a threshold value). In some cases, the computing device can use different criteria to determine a quality of the optical data, or otherwise determine that the NIRS device should be repositioned. For example, the computing device can determine an ambient light intensity (e g., from the optical data corresponding to the ambient light, that is, when no light sources of the NIRS device emit light), a detector intensity value, and a linearity of the relationship between the source-detector separation (“SDS”) and the logarithm of the product of the detector intensity and the square of the source-detector separation (e.g., log(I*SDS2)). In some cases, the computing device can provide the warning, notification, bad quality result, etc., based on one or more of these parameters. For example, the computing device can provide the warning, notification, bad quality result, etc., based on one or more of the ambient light intensity being greater than two percent of the maximum possible light detector value for the photodetector, a detector intensity value being less than two percent of the maximum possible light detector value, the detector intensity value being greater than 95 percent of the maximum possible light detector value, and the linearity of the relationship between the source-detector separation (“SDS”) and the logarithm of the product of the detector intensity and the square of the source-detector separation (e.g., log(I*SDS2)) being less than 0.99. In some cases, the computing device can evaluate these criteria for each light detector, each light detector and light source pair, etc. In some cases, the computing device can reject optical data that fails one or more of these criteria.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0127] At block 508, the process 500 can include the computing device extracting one or more photoplethysmogram (“PPG”) waveforms from the optical data (e.g., combined optical data). In some cases, the optical data can be a light intensity over time. In some cases, the photoplethysmogram waveforms can indicate blood volume changes in a blood vessel (e.g., artery over time). In some cases, each photoplethysmogram waveform can be the blood volume (e.g., in arbitrary units) over time. In some cases, the block 508 can include the computing device combining one or more of the photoplethysmogram waveforms together (e.g., averaging the photoplethysmogram waveforms, such as four waveforms) to produce a combined photoplethysmogram waveform.
[0128] At block 510, the process 500 can include the computing device taking the time derivative (e.g., the first derivative) of each photoplethysmogram waveform to obtain, for each photoplethysmogram waveform, a derivative photoplethysmogram waveform. In some cases, the block 510 can include the computing device determining one or more parameters for each photoplethysmogram waveform. In some cases, the one or more parameters can include a peak, a plurality of peaks, etc. In some configurations, a first peak can be a systolic peak (e.g., Pl) and a second peak can be a diastolic peak (e.g., P2). In some cases, the computing device can determine the subject’s heart rate from the photoplethysmogram waveform (e.g., each photoplethysmogram waveform).
[0129] In some cases, at the block 510, the process 500 can include the computing device determining that a blood pressure of a subject is high, elevated, etc., based on the ratio of the Pl and P2 in the derivative photoplethysmogram waveform of the photoplethysmogram waveform. For example, the computing device can determine that the blood pressure of the subject is high, elevated, etc., based on P2 being greater than Pl in the derivative photoplethysmogram waveform (e.g., where Pl and P2 are determined in the photoplethysmogram waveform used to form the derivative photoplethysmogram waveform). Similarly, the computing device can determine that a blood pressure of a subject is low, based on Pl being greater than the Pl in the derivative photoplethysmogram waveform.
[0130] At block 512, the process 500 can include the computing device providing each time derivative photoplethysmogram waveform to a machine learning model trained to output a blood pressure, an intra-arterial blood pressure, etc. In some cases, providing a time derivative photoplethysmogram to the machine learning model can include the machineAtty. Dtk. No. 125141.05019. MGH2O24-394learning model extracting one or more features, in which the one or more features can be used by the machine learning model to determine the blood pressure measurement (e.g., by comparing each feature to an arbitrary feature threshold of a decision tree of the machine learning model). In some cases, the machine learning model can be trained by providing the subject’s own intra-arterial pressure measurements (e.g., from an arterial pressure catheter, A-line, etc.) along with corresponding PPG waveforms (e.g., PPG waveforms, first time derivative PPG waveforms, etc.). In this way, the machine learning model can advantageously be subject specific, to account for unique subject variabilities. Further, after training, the arterial line can be removed, and the NIRS device can provide accurate intraarterial pressure measurements accordingly.
[0131] At block 514, the process 500 can include the computing device receiving a blood pressure measurement for the subject (e.g., a cerebral blood pressure measurement) from the machine learning model. In some cases, the blood pressure measurement can be an intraarterial blood pressure measurement.
[0132] At block 516, the process 500 can include the computing device presenting the blood pressure measurement on a display. In some cases, this can include notifying a user of the blood pressure measurement, for example, when the blood pressure measurement is below a blood pressure threshold or above a blood pressure threshold. In some cases, this can include the computing device transmitting the blood pressure measurement to another computing device. In some cases, this can include a computing device presenting one or more tissue chromophore concentrations (e.g., total hemoglobin, hemoglobin oxygen saturation, etc.) on the display. In some configurations, this can include notifying a user of each tissue chromophore concentration, transmitting each tissue chromophore concentration to another computing device, etc. In some cases, each tissue chromophore concentration can be determined from the optical data (e.g., at the block 506, 508, etc.). For example, the computing device can compare each optical data to a respective concentration curve of the specific tissue chromophore.
[0133] In some cases, the process 500 can include using the one or more PPG waveforms (or one or more derivative PPG waveforms) to determine other parameters, clinical measurements, clinical conditions, etc. For example, the process 500 can include a computing device determining that the subject has a cognitive impairment (e.g., mild cognitiveAtty. Dtk. No. 125141.05019. MGH2O24-394impairment), based on the time to the systolic peak (e.g., tPl) or the time to the diastolic peak (e.g., tP2). As a more specific example, the tPl can be the time at the Pl peak in the derivative PPG waveform, while the tP2 can be the time at the P2 peak in the derivative PPG waveform. The computing device can determine a presence of a cognitive impairment (e.g., mild cognitive impairment), based on the tPl being greater than a threshold value, based on the tP2 being greater than a threshold value, or both. As yet another example, the process 500 can include the computing device determining that the subject has ischemia (e.g., hypoperfusion, cerebral hypoperfusion, etc.), an occluded blood vessel (e.g., a carotid artery, etc.) based on a second amplitude of a PPG waveform at a second time period being less than a first amplitude of the PPG waveform at a first time period. In some cases, this can include the computing device determining that the subject has an ischemia (e.g., hypoperfusion, cerebral hypoperfusion, etc ), based on an amplitude of the PPG waveform being less than a threshold value. In still other cases, this can include the computing device determining that the subject has an ischemia (e.g., hypoperfusion, cerebral hypoperfusion, etc.), based on a ratio of the first amplitude to the second amplitude is greater than a threshold value.
[0134] FIG. 9 shows a flowchart of a process 550 for detecting an internal pressure (e.g., intracranial pressure). The process 550 can be implemented using any of the NIRS devices, systems, etc., any SCOS devices, etc. Further, the process 550 can be implemented using one or more computing devices. Accordingly, the process 550 can be a computer-implemented method. Although the following description of the process 550 will describe a computing devices implementing the process 550, it is appreciated that this can include one or more computing devices to implement the process 550 (e.g., a plurality of computing devices).
[0135] At block 552, the process 550 can include a computing device receiving one or more PPG waveforms (e.g., similar to the block 508, the block 510, etc ).
[0136] At block 554, the process 550 can include the computing device causing a pulsed light source to emit light towards tissue.
[0137] At block 556, the process 550 can include the computing device receiving a plurality of images in response to the pulsed light. For example, at least one image can be acquired from an image sensor during each pulse of the pulsed light. In some cases, each image of the one or more images can be a speckle image. In some cases, the plurality of images are a time series of images.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0138] At block 558, the process 550 can include the computing device extracting a blood flow waveform (e.g., a cerebral blood flow waveform) from the plurality of images (e.g., speckle images). In some cases, the blood flow waveform can be an index of blood flow (e.g., an index of cerebral blood flow (“CBFi”)).
[0139] At block 560, the process 550 can include the computing device modifying the blood flow waveform to output or provide a modified blood flow waveform, using, for example, one or more PPG waveforms (e.g., the PPG waveforms at the block 552). In some cases, modifying the blood flow waveform can include enhancing the blood flow waveform using the one or more PPG waveforms. For example, a computing device can modify the blood flow waveform using the derivative PPG waveform (e.g., the first time derivative of a PPG waveform), as these are related to each other. For example, the blood flow waveform equals or is proportional to the blood flow in (BFin) + the blood flow out (BFout), while the derivative PPG waveform is proportional or equals the BFin - BFout. In some cases, this modified blood flow waveform is enhanced and improvers the SNR, accuracy, etc., of an internal pressure measurement (e.g., an intracranial pressure measurement). In some cases, in addition to the derivative PPG waveform, the underling PPG waveform can be combined with the derivative PPG waveform and the blood flow waveform to determine the modified blood flow waveform. For example, the PPG waveform can be proportional to the BFout.
[0140] At block 562, the process 550 can include the computing device determining an internal pressure measurement (e.g., an intracranial pressure measurement) from the modified cerebral blood flow waveform. In some cases, the computing device can determine the critical closing pressure from the modified cerebral blood flow waveform. The computing device can then determine the internal pressure (e.g., intracranial pressure) based on the critical closing pressure (e.g., and the vessel wall tension of the blood vessel being measured).
[0141] In some cases, the block 562 can include the computing device presenting the internal pressure measurement (e.g., intracranial pressure measurement) on a display. In some cases, this can include notifying a user of the internal pressure measurement, for example, when the internal pressure measurement is below an internal pressure threshold or above an internal pressure threshold. In some cases, this can include the computing device transmitting the internal pressure measurement to another computing device. In some cases, this can include a computing device presenting one or more tissue chromophore concentrations (e.g., totalAtty. Dtk. No. 125141.05019. MGH2O24-394hemoglobin, hemoglobin oxygen saturation, etc.) on the display. In some configurations, this can include notifying a user of each tissue chromophore concentration, transmitting each tissue chromophore concentration to another computing device, etc. In some cases, each tissue chromophore concentration can be determined from the optical data (e.g., at the block 506, 508, etc.). For example, the computing device can compare each optical data to a respective concentration curve of the specific tissue chromophore.
[0142] At block 564, the process can include determining a presence of a brain injury (e.g., a traumatic brain injury, a stroke, an intracranial hemorrhage, a brain bleed, etc.), based on the intracranial pressure measurement (e.g., when the internal pressure measurement is an intracranial pressure measurement). For example, the computing device can compare the intracranial pressure (“ICP”) measurement to an ICP threshold (e.g., an ICP threshold value). The computing device can determine that the subject has a brain injury, based on the ICP measurement being greater than the ICP threshold. In other cases, the computing device can determine that the subject does not have the brain injury, based on the ICP measurement being less than the ICP threshold.EXAMPLES
[0143] The following examples have been presented in order to further illustrate aspects of the disclosure, and are not meant to limit the scope of the disclosure in any way. The examples below are intended to be examples of the present disclosure and these (and other aspects of the disclosure) are not to be bounded by theory.EXAMPLE 1
[0144] Over the past thirty years, cerebral oximetry and functional Near-Infrared Spectroscopy (“NIRS”) developed as non-invasive, non-ionizing methods for brain imaging and monitoring. NIRS measurements of brain hemodynamics are predominantly performed with the continuous-wave NIRS (“CWNIRS”) modality, which uses the simplest, least expensive devices to measure changes in cerebral oxy-, deoxy- and total hemoglobin concentrations.
[0145] CWNIRS uses red and / or near-infrared light shone on the scalp and detected a few centimeters away, after it propagates diffusively through the scalp, skull and brain. The amountAtty. Dtk. No. 125141.05019. MGH2O24-394of diffuse light that reaches the detector varies as the optical absorption of the brain varies in response to hemoglobin concentration changes. By measuring these changes in light attenuation at two or more wavelengths, CWNIRS allows for the retrieval of changes in cerebral oxy-, deoxy-, and total hemoglobin concentrations resulting from cortical activation.
[0146] FIG. 10A-B show Pictures of Gen 1 (a) and Gen 2 (b) FlexNIRS modules.
[0147] FIG. 11A-B show a dimensional drawing of the symmetric multi-distance optical probe, b) Pulse oximetry configuration with only one source and one detector.
[0148] FIG. 12A-B show a complete setup circuit of (a) Gen 1 FlexNIRS. b) Gen 2 FlexNIRS circuit with a second probe expansion.
[0149] FIG. 13 shows pictures of FlexNIRS packaging.
[0150] The previous NIRS devices are fiber optics based, with a main unit which includes all the hardware and the power supply, a computer for display and data collection and a fiber optics probe to connect to the subject. The new NIRS device, the FlexNIRS, is a wearable and wireless device which allow to monitor subjects while performing any activity.
[0151] As seen in FIG. 12, the prototypes consist of a set of modules connected together with either flexible wiring (a) or embedded in the same PCB (b). the modules include: an optical probe, an analog front end (AFE) and digitization board, a Bluetooth wireless board (BLE) and a rechargeable battery.
[0152] The optical probe hosts dual or 8-color LED light sources (“SRC”) (dual color: either 735 and 850 nm, SMT735D / 850D or 760 and 850 nm, SMT760 / 850; 8 colors; 735, 765, 785, 805, 830, 855, 870, 910 nm COB-8040-MGH-X; Marubeni America Corporation) and pin photodiode detectors (DET) (either VEMD5060X01 or VEMD5160X01, Vishay Semiconductors) and is realized in a flexible PCB technology, allowing it to adapt to the surface of the forehead and guarantee maximum contact with the skin. The typical geometry includes 2 sources and 3 detectors symmetrically arranged to obtain multiple source-detector separations (see FIG. 12). Other source-detector arrangements are also used, as for the ear clip optical probe, which includes only one source and one detector, and operates in transmission mode, as pulse oximeters.
[0153] The analog front-end and digitization board hosts a specialized analog front-end circuit (Texas Instruments AFE4900 or AFE4950, Ultra-Low-Power Analog Front End) designed for wearable optical and electrical sensing. The board has a transmitter section thatAtty. Dtk. No. 125141.05019. MGH2O24-394drives the four / sixteen light sources, a receiver section that supports three detectors and an electrocardiogram (ECG) signal chain. This device has a timing engine that can be used to create customized acquisition sequences to optimize power usage. The current driver of the transmitter section drives the LEDs in common anode configuration with current independently adjustable up to 200 mA. The current from the photodetector is converted to a differential voltage using a gain adjustable transimpedance amplifier (TIA). The output voltage is sampled and converted by a 24-bits ADC. An offset cancellation stage can be programmed to subtract a DC current from the current signal allowing the increase of the signal-to-noise ratio. The ECG acquisition section has a differential couple of inputs plus an output commonmode bias voltage that is intended to close a feedback circuit to maintain the input common mode bias. The front-end of the ECG is a programmable gain INA. The prototype allows to record electrical activity of the brain by using electrodes and special signal amplifiers to convert the ECG channel into an EEG channel. The communication of the board is ensured by a SPI interface and a proper connector.
[0154] The communication board hosts a microcontroller, a Bluetooth module for wireless data transmission, an inertial module to detect movements of the head, and a battery management circuit. The microcontroller programs and retrieves data from the AFE4900 / AFE4950, implements post-processing and send data out using the Bluetooth module (either BGM121 or BGM220S, Blue Gecko Bluetooth® SiP Module, Silicon Labs). These are low power and small size models with an integrated chip antenna. The BLE are characterized by a transmission power up to 8 dBm and receiving sensitivity of -90 dBm that allow it to work up to a range of 10 m. The Inertial module (LSM6DS3H, STMicroelectronics or ADXL344, Analog Devices) is a low power MEMS 3-axis accelerometer and 3-axis gyroscope used to detect the movement of the cerebral oximeter and therefore distinguish motion artifacts.
[0155] A small and lightweight Lithium-ion battery with a capacity of 550 mAh allows the prototype to be a wireless and wearable device, while recording data for more than 8 hours. The battery charge and discharge cycles are monitored by a dedicated circuit.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0156] For both the Gen 1 and Gen 2 FlexNIRS modules, the Bluetooth board hosts a second connector that allows the acquisition from a second channel (e.g., a second probe and analog front end) through a second SPI interface as reported in FIG. 12.
[0157] Acquisition rate and battery endurance. The AFE4900 front end can interrogate each LED source — photodiode pair sequentially and is able to dynamically switch the LED current and photodiode gain. The internal programmable timing engine creates the pattern of signals related to the different phases of the signal acquisition. The acquisition is based on a sequence of phases: activation of the light source, sampling, and conversion, that are repeated for every source. Setting the timing engine is possible to configure the period of the acquisition, the order and the duration of every phase. Testing of 1 source and one detector shows a 1000 Hz data acquisition rate is sustainable (from a Bluetooth Low Energy data rate standpoint). The most common cycle includes both active and null measurements (e.g. LED and / or photodiodes being off) to provide “dark references” allowing the SNR of the acquisition to be maximized. This high acquisition rate enables advanced measurement modes, including pulse oximetry (“SpO2”) and heart rate variability (“HRV”), and permits approximately 8-12 hours of battery powered operation. The firmware allows setting lower rates / duty cycles as needed when longer operational times are desired and only StO2 is needed. LED current and photodiode gain are set automatically when data acquisition is initiated to obtain optimal signal quality for each patient. In some cases, since optical signals vary from person to person (e.g., due to tissue hemoglobin concentration, tissue water content, skin color, etc.), in some cases, light source luminosity (e.g., light source driving currents, such as LED driving currents) and light detector gains (e.g., the programable gain of a transimpedance amplifier) can be tailored or adjusted for each specific subject. This can provide a more accurate estimation of StO2 and SpO2, among other measurements.
[0158] Software. The FlexNIRS can be controlled using a Lab VIEW software on a laptop. Alternatively, to maximize the portability, a mobile application has been developed that implements a graphical user interface. The app is built on the Android 10 operating system and works with BLE protocols. It communicates with the device, visualizes the light intensity, cerebral oxygenation, and motion on the screen in real time, and saves the data on the device. Upon connecting to the FlexNIRS, the app opens four different tabs (“fragments” in Android), each with various features including real-time feedback that help the user optimize data qualityAtty. Dtk. No. 125141.05019. MGH2O24-394and monitor results. A banner is located on the top of the screen of all the tabs, allowing for easy access to essential functions: connect / disconnect to the FlexNIRS Bluetooth, start / stop data recording, log events, and a pop-up settings window to change assumed tissue scattering, water concentration, sources-detectors separations, and data quality thresholds. At the bottom of all the tabs, the values of the data quality parameters are displayed and followed by green / red dots to warn the user about detector saturation and aid correct probe positioning on the forehead.
[0159] Packaging The FlexNIRS modules are integrated into a stretchable, breathable wrapping material that can be strapped around the head, arm, ear, leg, and held in place using either Velcro, tape or clips. Two layers of the material are used, embedding the circuitry and connecting cables in between the layers. The sensor area has a cutout on the inner side for contact with the skin. Hydrogel can be used as interface between the probe and the patient skin. The analog front end and Bluetooth board are protected within the wrap. Finally, a clip in socket is mounted on the outer surface for the battery. The analog front end and Bluetooth board are contained within a low profde, 3D printed flexible case, with openings for flat cables to the optical probe, and to the battery holder. The Li-ion battery is also encapsulated in a 3D printed low-profde flexible case with a retaining clip and electrical contacts added to it. The weight of the entire assembly is less than 3 ounces, somewhat more than a modern smart watch but less than half the weight of a typical modern smartphone. Example of an alternate packaging with longer cable and remote control electronics is shown in FIG. 13.
[0160] Light power. The Marubeni LEDs have a max radiated power of- 300 mW at 500 mA. In the FlexNIRS, the max current to the LED is 200 mA, and assuming a linear dependence of light power with current, the max LED power is <200 mW, if the LEDs are always on. In the FlexNIRS, each LED is turned on and off in sequence at a low duty cycle. It has been verified that the maximum light power combining all colors, do not go over 20-35 mW, which is still very low and safe for skin illumination. In fact, while these are LEDs and not Lasers, this light power is below the ANSI standard limits for Laser illumination.
[0161] The FlexNIRS devices are not disposable at this time and can be sanitized before and after each use in a subject. All the wrapping materials are single use. The FlexNIRS can be easily detached from the subject in case of an emergency. The lithium-ion battery has noAtty. Dtk. No. 125141.05019. MGH2O24-394more risk associated with it than a small smartphone battery. The LEDs light power below ANSI standard and is safe for human skin illumination for any period.EXAMPLE 2
[0162] The glymphatic system (GS), a recently discovered neural pathway, plays a crucial role in the removal of cerebral waste and toxins from the central nervous system. Comprising a network of perivascular channels alongside cerebral blood vessels, the glymphatic system demonstrates higher activity during sleep. Specifically, the increased flow of cerebrospinal and interstitial fluid (CSF and ISF) during sleep significantly amplifies the clearance of extracellular metabolites, amyloid-P and other proteins, surpassing the efficacy observed in wakeful states. The improved clearance of waste by the GS during sleep is multi-factorial, it involves an increase in the volume of the interstitial space, and is associated with slow-delta electrophysiologic oscillations, low frequency arterial fluctuations, and autonomic nervous system fluctuations. The glymphatic system's efficiency appears to decline with age, which may contribute to the increased susceptibility to neurodegenerative diseases in older individuals, and GS dysfunction has been implicated in various neurological disorders, including Alzheimer's disease.
[0163] Understanding and promoting a healthy glymphatic system may impact brain health and cognitive function. To date, investigations of GS function have largely consisted of invasive imaging studies in animal models. In humans, the assessment of GS function is done using MRI; however, the costs and impracticalities associated with MRI hinder its widespread use in the general population. What is needed is a non-invasive, cost-effective, portable modality capable of monitoring GS function in humans during naturalistic sleep at home.
[0164] In this study, we propose utilizing near-infrared spectroscopy (NIRS) to monitor GS function. NIRS is a non-invasive diffuse optical technique that employs light at various wavelengths to discern absorption changes of tissue chromophores (hemoglobin, lipid, water, etc.). The preliminary data indicate that NIRS can detect an increase in low-frequency hemoglobin concentration (Hb) oscillations during deep sleep, consistent with the BOLD MRI oscillations observed during GS activity (FIG. 14) In this project, we plan to leverage a low-cost, wearable, wireless NIRS device called FlexNIRS that we have recentlyAtty. Dtk. No. 125141.05019. MGH2O24-394developed, and add wavelengths around the water absorption peak to also resolve cerebral water fraction changes. The hypothesis is that with the GS-FlexNIRS device we will be able to monitor glymphatic CSF and Hb fluctuations during naturalistic sleep.
[0165] The specific aims of the project are:
[0166] Aim 1. Hardware modification: Modify the FlexNIRS device to acquire overnight data at 760, 810, 850 and 970 nm, to best separate deoxy-hemoglobin, oxyhemoglobin, and water concentrations changes.
[0167] Aim 2. Laboratory testing: Test the GS-FlexNIRS device in 30 young (18-39) and 10 older subjects (60-75) during a full sleep cycle in a controlled lab environment. Sleep stages will be assessed with co-acquired polysomnography (PSG). We will refine data processing flows, demonstrate increases in amplitude of glymphatic system hemoglobin and water fluctuations during deep sleep, and investigate age-related reductions in GS activity.
[0168] Aim 3. At home feasibility: Demonstrate feasibility of conducting overnight GS-FlexNIRS measurements at home in 10 healthy young subjects.
[0169] FIG. 14 shows an example of fMRI cortical BOLD (green-volume) and CSF (purple-flow) signals during wakefulness (left) and sleep (right), as reported in FIG. I6A- B of Bottom: Example of FlexNIRS optical density signals at 850 nm (volume) recorded from a subject during REM (left) and deep sleep (right), as indicated by a sleep tracker wristband. The LOD oscillations observed during deep sleep closely resemble the low-frequency BOLD oscillations associated with glymphatic activity.
[0170] These pilot studies play a crucial role in guiding the design of the ultimate NIRS wearable device and refining data analysis processes before incorporating the method into a more extensive clinical study. The successful development of this wearable GS monitor will facilitate the evaluation of glymphatic function at the consumer level, longitudinally within the home environment during sleep. The cutting-edge monitoring capabilities are poised to redefine sleep quality assessment, providing valuable insights into the efficacy of prospective therapeutic interventions aimed at enhancing GS function. Furthermore, this technology harbors the capability to identify early deviations in GS functions before the onset of irreversible damage, thereby contributing significantly to proactive health management.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0171] The glymphatic system (GS) is an integral component of the brain’s waste clearance mechanism, playing a crucial role in the removal of metabolic waste and toxins, including amyloid-P and tau proteins implicated in the pathogenesis of Alzheimer's disease and other neurodegenerative disorders. Active primarily during the slow-wave sleep stage, the GS operates as a specialized network of perivascular channels that facilitate the exchange of cerebrospinal fluid (CSF) and interstitial fluid (ISF). This unique system, known for its ability to efficiently eliminate neurotoxic byproducts, underscores its significance in maintaining brain health. The notion that failure in the GS may precede or predispose individuals to protein accumulations in neurodegenerative diseases emphasizes the importance of gaining a deeper understanding of the dynamics of the GS and its connection to cognitive decline. Methods are being explored to enhance glymphatic function, including approaches like vagal nerve stimulation, acupuncture, maintaining adequate sleep, regular physical exercise, and a balanced diet, showing considerable potential for therapeutic strategies aimed at addressing neurodegenerative diseases and other neurological disorders even before their manifestation.
[0172] These findings highlight the need for innovative tools capable of visualizing glymphatic system activity in real-time, providing a crucial avenue for feedback on strategies aimed at preserving brain health, GS activity is manifested as an expansion of the interstitial space as demonstrated in a study by Xie et al.. Utilizing in vivo two-photon imaging in mice, the study first visually captured the glymphatic pathway, revealing a significant increase in CSF influx from the cerebral vasculature into the brain interstitial space during sleep. GS activity is further characterized by convective exchange of CSF and ISF driven by cerebrovascular oscillations that span cardiac, respiratory, and lower frequencies. These oscillations drive fluid into the perivascular space, enabling efficient clearance and exchange of ISF and their disruption or irregularities have been associated with imbalances in ISF dynamics. Finally, electroencephalography (EEG) studies in humans have provided further insights into neuronal influences of glymphatic activity, revealing that that brain wide oscillations evident in delta frequency power during deep sleep are associated with an enhanced glymphatic clearance process. Dr. Lewis and collaborators first demonstrated the correlation between CSF oscillations, slow hemoglobin oscillations and EEG delta activity, highlighting the intricate interplay between vascular dynamics and neural activity in theAtty. Dtk. No. 125141.05019. MGH2O24-394context of glymphatic function. These findings demonstrate the significance of comprehending how cerebral water fraction volume and oscillatory dynamics changes, cerebral hemodynamic fluctuations, and neuronal activity collectively influence waste clearance, thereby presenting opportunities for precise GS monitoring in the context of cognitive health and neurodegenerative disorders.
[0173] As indicated in the literature mentioned above, glymphatic system (GS) function studies primarily rely on invasive animal imaging methods, including ex -vivo microscopy, two-photon imaging and contrast-enhanced magnetic resonance imaging (MRI) following intra-cranial / lumbar injections. Approaches for evaluating GS function in humans almost exclusively utilize MRI, which is expensive, time-consuming, and impractical for routine monitoring. Advancements in technology have opened avenues for noninvasive and cost-effective monitoring solutions, and one such promising approach we propose is the utilization of near-infrared spectroscopy (NIRS). NIRS, a safe and portable technology, enables the measurement of changes in main tissue chromophores by detecting attenuation of near-infrared light. The preliminary findings, illustrated in FIG. 14, underscore the capability of NIRS to detect low-frequency cerebral hemoglobin oscillations during deep sleep. Remarkably, these observations align with the glymphatic-associated, cortical hemoglobin fluctuations previously measured by co-investigators Lewis and Bonmassar using BOLD functional MRI. The NIRS data presented in FIG. 14 were acquired utilizing the recently developed wearable wireless app-controlled device, FlexNIRS This innovative device allows for the measurement of changes in oxy- and deoxy-hemoglobin concentrations (HbO and HbR) at high sampling rate and signal-to-noise ratio (SNR). As a result, the FlexNIRS enables the resolution of cardiac, respiratory, and lower frequency hemoglobin oscillations in ecologically valid conditions during overnight sleep (FIG. 17A-C).
[0174] Building on these important and innovative findings, the project aims to enhance the monitoring of GS function with FlexNIRS by incorporating an additional wavelength at 970 nm, centered around a water absorption peak (Aim 1). This addition will allow us to detect low-frequency CSF oscillations alongside hemodynamic oscillations and potentially resolve level changes in cerebral water fraction (H2O), corresponding to increases in interstitial and perivascular volume during sleep, thereby offering a more comprehensive understanding ofAtty. Dtk. No. 125141.05019. MGH2O24-394the dynamics of the glymphatic system in humans. NIRS has been used in the past to detect changes in tissue water fraction, however, a novel approach is needed to enable high SNR measurements with a device suitable for ecologically valid sleep. To address these, the approach involves the use of 4 wavelengths in a self-calibrated multi -di stance geometry', and specialized electronics to enhance SNR at the longer wavelengths. Preliminary data and simulations, illustrated in FIG. 17A-C, indicate the feasibility of the proposed method. Once the hardware is ready, we will evaluate the GS-FlexNIRS device in a controlled laboratory environment using young, healthy subjects (Aim 2). This assessment will be conducted alongside wearable polysomnography (PSG), which provides EEG, ECG, respiration rate, EMG, and pulse oximetry, to accurately characterize sleep stages and validate the correlation between GS activity, deep sleep stages, and slow variation in delta EEG power. Furthermore, we will demonstrate the device's capability to detect glymphatic-related changes in Hb and H2O in individuals with different skin tones. In the subsequent phase of Aim 2, we plan to include 10 older subjects to begin investigating reduction in GS functionality associated with aging. Finally, in Aim 3 'e intend to evaluate the GS-FlexNIRS in a real-world scenario by conducting an at-home sleep study to demonstrate the feasibility and practicality of utilizing the GS-FlexNIRS device outside a laboratory setting.
[0175] Rigor of the Prior Research. The glymphatic system plays a crucial role in preserving cognitive performances. Currently, the measurement of GS function in humans relies solely on MRI, which limits access to a potentially significant biomarker. Recognizing the significance of monitoring the GS during naturalistic sleep, we propose to use a state-of-the-art, noninvasive, cost-effective near-infrared wearable device. Previous investigations by the group have demonstrated that the sleeping brain exhibits slow CSF oscillations which are interlinked with hemodynamic and neural rhythms. The BOLD oscillations, reflective of cortical blood volume changes, are readily detectable using NIRS methods, and, for the first time, the preliminary data shows the feasibility of capturing these hemoglobin oscillations with NIRS (FIGS 14 and 17). Beyond Hb oscillations, NIRS at longer wavelengths holds promise in resolving changes in water fraction. The plan to add 970nm wavelength to the GS-FlexNIRS aims to provide sufficient SNR and sensitivity to resolve water fraction changes of <1%, as shown in the preliminary results (FIG. 17A--C). The team, led by Dr. Franceschini, an expert in NIRS technology development and translation, coupled with Dr.Atty. Dtk. No. 125141.05019. MGH2O24-394Bonmassar's expertise in EEG data collection and analysis, and Dr. Lewis's knowledge in multimodal GS monitoring and data analysis, benefits from a multidisciplinary collaboration crucial for the successful execution of this project. The realization of this consumer GS NIRS monitor not only addresses current limitations in GS function assessment but also opens new avenues for early detection, intervention, and personalized management of neurological disorders, contributing to improved outcomes and quality of life for individuals at risk of cognitive decline.
[0176] INNOVATION. The proposed GS-FlexNIRS device presents a highly innovative opportunity to monitor glymphatic system function. By simultaneously measuring cerebral water fraction and hemodynamic fluctuations during sleep, and allowing for smooth integration with EEG, it will provide a comprehensive understanding of cerebral nervous system waste clearance processes and insights into macroscopic features of GS dysfunction, providing possible biomarkers for early developing features of cognitive decline. Designed with accessibility and user-friendliness in mind, the device will be readily translated into clinical practice, promoting early detection of GS dysfunction, and providing valuable feedback into the efficacy of prospective therapeutic interventions aimed at enhancing GS function. This technology aligns well with the current trend of personalized and preventive medicine, emphasizing its broader impact on public health and neurological well-being. Beyond clinical applications, the GS-FlexNIRS device's high performance, cost-effectiveness, and wearability position itself as an innovative solution for seamless conversion into a consumer device. This transition goes beyond conventional sleep monitoring devices, offering a unique perspective into sleep quality by monitoring glymphatic function and suggesting strategies to enhance GS function, promoting lifestyle modifications.
[0177] FIG. 15A-C show photographs of the latest 2-color FlexNIRS prototype, b) Specs of the proposed LEDs and PD components for the GS-FlexNIRS. c) Simplified block diagram of the proposed GS-FlexNIRS device.
[0178] Approach. The objectives are: i) to develop a wearable, wireless multiwavelength NIRS device designed for home use during overnight sleep, and ii) to assess its capability to monitor glymphatic system function by measuring changes in hemoglobin and water concentrations across different sleep stages. We have tested the current 2-color FlexNIRS prototype in humans during sleep and we have confirmed its abilityAtty. Dtk. No. 125141.05019. MGH2O24-394to detect increased low-frequency Hb fluctuations during deep sleep (FIG. 14 and 17). The plan is to extend the FlexNIRS capability to also resolve cerebrospinal fluid changes and perform pilot measurements in healthy subjects to show feasibility and initial utility of the method.
[0179] Aim 1. Hardware modification: Enhance the FlexNIRS hardware to enable estimate of water fraction changes. The glymphatic system NIRS device will be based on the latest FlexNIRS prototype (Fig. 2a), and new optoelectronic components will be selected to optimize the device for this application. In particular, the dual-color LED package (SMT760 / 850, Marubeni America Corporation) will be replaced by a new high-power 4-color LED package (EDCC family, Marubeni). The 4-color LEDs are fabricated within independent ceramic packages measuring 1.5 mm x 1.85 mm (Fig. 2b), and, for the same driving current, provide almost 2 times higher optical power than the SMD dual LEDs. Selecting from the available colors, the customized 4-LED package will comprise LEDs emitting at 760, 810, 850, and 970 nm. These wavelengths are strategically chosen to ensure accurate assessment of HbO, HbR and H2O. To discern changes in water concentration, we will utilize a 970 nm wavelength, which is close to the 980 nm water peak. Further, including a fourth wavelength around the isosbestic point of hemoglobin (810 nm) enables a more distinct differentiation between Hb and H2O. The hardware and sampling parameters will be optimized to compensate for the high tissue absorption at 970 nm. Consequently, we will replace the current photodetectors (PD) (VEMD5060X01, Vishay Semiconductors) with PDs featuring enhanced sensitivity and a peak detection shifted to 950 nm (VEMD5080X01) (FIG. 15A-C).
[0180] Preliminary multilayer Monte Carlo simulations of a realistic head geometry which incorporate the noise properties of selected components, indicate that with this wavelengths choice we can recover small brain changes in oxy- deoxy-hemoglobin and water concentration changes with minimal crosstalk, and that with the new hardware we achieve a ~3x contrast-to-noise ratio improvement with respect to the SMT LEDs series and current PDs (FIG. 16A--B). The simulations suggest this approach will improve SNR by -7dB at 970 nm and increase the sensitivity to cerebral water fraction signal by threefold. Preliminary measurements conducted with a modified FlexNIRS operating at 980nm (SMT980D),Atty. Dtk. No. 125141.05019. MGH2O24-394demonstrate the capability to detect cerebral pulsatile and respiratory signals at 3.3 cm sourcedetector separation at this longer wavelength with excellent SNR (FIG. 16A-B).
[0181] FIG. 16A-B show estimated recovery of chromophores changes using Monte Carlo simulation of a realistic head geometry'. We used the noise properties for the current (+970 nm) and the planned GS-FlexNIRS devices. When changing concentration of brain HbO (top), HbR (middle), H2O (bottom) with the selected wavelengths we recover chromophores concentration changes with minimal crosstalk. Moreover, the improved SNR of the GS-FlexNIRS significantly reduces the variance of the recovered concentrations b) Data collected at 100Hz and 3.3 cm separation on the forehead of a volunteer with a modified FlexNIRS using the proposed PDs and 2 LEDS at 980 nm (SMT980D). The SNR is comparable to the one achieved at shorter wavelengths and both cardiac and lower frequency signals are present at this water sensitive wavelength. Further improvement of SNR is expected with the new' high-power LEDs and higher current AFE.
[0182] As implemented in the original FlexNIRS, we will use the self-calibrating geometry based on 2 sources and 2 detectors placed symmetrically to provide two sets of identical long source-detector separations (LS) The symmetric geometry compensates for component variability without need for external calibration and provides quality control metrics The multi-distance scheme utilizes the slopes of intensity vs distance to quantify chromophores, reducing sensitivity to skin color, and increasing depth sensitivity compared to a single intensity measure. A third PD placed at a short-separation (SS, <1 cm) enables the acquisition of superficial scalp signals for subtraction from long source-detector channels, further enhancing depth sensitivity.
[0183] To drive 8 LEDs (4 per source location) the Analog Front End (AFE) integrated circuit (IC) AFE4900 (Texas Instruments) will be replaced by a newly-developed AFE4460. The AFE4460 IC allows for driving up to 32 independent LEDs with up to 350mA driving current (current system is limited to 200mA) and sampling 4 PD (2 in parallel). The shortseparation and background ambient light will also be sampled, leading to an overall number of 15 phases driven by the AFE at 50 Hz sampling rate Finally, to sustain the higher LED driving current, the current battery? will be replaced with another medical-grade Li-Po with increased charge capacity of 1900mAh (LP103450JH, Jauch Quartz America, Inc ) and comparable size-factor, maintaining the estimated battery life of 8 hours. As currentlyAtty. Dtk. No. 125141.05019. MGH2O24-394implemented (FIG. 15A-C), the GS-FlexNIRS will be manufactured on a flex printed circuit board (PCB), allowing the hosting of narrow-pitch ICs while offering maximum flexibility to accommodate for the head curvature of the patient and maximize the optical coupling. Specifically, the source-detector separation will be 08, 2.8, and 3.3 cm and the overall board dimension will be 8 cm x 2.3 cm (FIG. 15A --C). A 3-axis accelerometer and gyroscope IC (ADXL344, Analog Devices) will be embedded on the PCB to allow tracking subjects movements and reject motion artifacts in the optical data. Finally, a microcontroller (BGM220S, Blue Gecko Bluetooth® SiP Module, Silicon Labs) will run the measurement routine: programming the AFE, acquiring the sampled light and motion signals and transmitting data to the external tablet through a Bluetooth low energy 5.2 communication. Despite the higher number of sources, thanks to the parallel acquisition of the new AFE, the higher power of the LEDs, and the higher PD sensitivity in the infrared, we expect the 4-color FlexNIRS when operating at 50 Hz will achieve a NEP 70 fW / √Hz and a dynamic range > 80 dB. The device will be encapsulated in a low-profile, medical-grade silicone elastomer (Elkem RTV-4420) vacuum-cast case and embedded in a fabric headband. The sensor area will have a cutout on the inner side for contact with the skin. Finally, a clip-in socket for the battery will be either attached to the headband or routed to the chest of the subject and attached to the PSG system. Magnetic pogo-pin connectors will be used to allow for easy battery replacement. The weight of the entire assembly is expected to be around 2-3 ounces (-60-80 grams). The plan is to finalize the circuit design and manufacturing by month 3 of year 1, and to complete firmware programming, packaging, and characterization by month 6 of year 1. Meanwhile the current Android app will be modified to control and display the 4-wavelength data, estimate water fraction changes in addition to currently available hemoglobin oxygenation and concentration changes, and provide additional data quality feedback at the additional wavelengths.
[0184] Aim 2. Laboratory testing: Test the GS-FlexNIRS system during overnight sleep within a controlled laboratory environment. Collected NIRS and PSG data will be used to refine data processing processes and demonstrate increases in amplitude of low frequency Hb and H2O fluctuations during deep sleep, and age-related reductions in GS activity. These data will also be used to evaluate the ability to detect changes in water level across sleep stages and the effect of skin color on measured parameters.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0185] Subject recruitment: Forty healthy subjects will be recruited and enrolled for a sleep study aimed at validating the GS-FlexNIRS device. We plan to start, recruitment six months into the grant period. Inclusion and exclusion criteria, along with protocol details, are outlined in the human subject section.
[0186] We aim to enroll 30 young subjects, aged between 18 and 39 years, representing diverse skin pigmentation. This will include 10 individuals each with light, medium, and dark skin, classified according to the Fitzpatrick classification and colorimeter measurements (DSM-4, Cortex Technology). We will ensure an equal distribution of male and female participants. Additionally, we will enroll 10 older subjects, aged between 60 and 75 years. All subjects will undergo overnight sleep monitoring in the lab using both the GS-FlexNIRS and a wireless polysomnography system that allows for the acquisition of 12-el ectrode EEG, ECG, EMG, pulse oximetry, and respiration rate (Nox Als, Nox Medical). The PSG data will be used to assess sleep stages and correlate EEG and systemic physiology with the NIRS data.
[0187] Data preprocessing: Raw data will be preprocessed by removing periods with excessive motion artifacts as detected by the built-in accelerometer. Changes in head position will be flagged in the data. For the fluctuation analysis ODs will be bandpass filtered (0.01-10 Hz) to extract relevant oscillations frequencies. For slow water level changes, after ambient light subtraction, we will compute an average OD for each cardiac pulse to remove high frequency signals. In both cases, the multi -wavelength multi -di stance self-calibrated approach described in will be used to estimate HbO, HbR and H2O changes. Chromophore concentration estimates will be computed in a spectrally aware manner, considering the linewidth of the LEDs and the responsivity of the PDs,
[0188] Increase in GS fluctuation's amplitude during deep sleep: To assess the contributions at different frequencies we will perform a sleep stage specific analysis, following established fNIRS approaches for analyzing resting state network data. Both OD and concentration data will be considered. Data will be separated into 5-minute windows, and a generalized linear model (GLM) framework will be used to remove correlated changes in large and short separation signals data. SS signals are first bandpass filtered into three distinct bands representing the cardiac pulsation (0.6 Hz - 10 Hz), the respiratory fluctuationAtty. Dtk. No. 125141.05019. MGH2O24-394(0.15 Hz - 0.6 Hz), and low frequency oscillation (001 Hz - 0 15 Hz) representative of glyraphatic flow
[0189] An example of a full night sleep analyzed with this method can be seen in FIG.17A-C. Data were acquired with the current dual-color FlexNIRS on a young subject. The multi-taper spectrogram computed from the 850 nm A OD regressed signal contains cardiac, respiratory, and low-frequency bands that can be seen to vary' in both amplitude and frequency throughout the night. In the spectrogram (FIG. 17A--C), the low frequency component (<0.1Hz) is the most prominent earlier in the night with increases in the respiratory band (~-0.2Hz) becoming apparent later in the night. The amplitude of the low frequency oscillations, analogous to those reported in FIG. 14 for the same dataset, fluctuate throughout the night. In these preliminary data we used a wristband sleep tracker to assess sleep stages. In the proposed project, we will use EEG signals from the PSG to demonstrate that the increases in the low-frequency band occur during deep sleep stages.
[0190] To demonstrate increases in amplitude of GS hemoglobin and water low-frequency fluctuations during deep sleep we will add the PSG time courses information into the spectrograms and compute sleep stage specific frequency power spectra. One-sided paired t-tests will be utilized to demonstrate a higher signal power in the low frequency bands during deep sleep compared to wakefulness or REM. Assuming a standard deviation of 3, to detect a difference of 1.5 (effect size 0.5, > 0.05) with a sample size of 30, the power will be 83%.
[0191] Relationship with neuronal activity: To examine the relationship between cerebrovascular driven oscillations and neuronal activity we will attempt a sleep stimulus driven analysis following standard INIRS evoked response data analysis methods. EEG derived relevant sleep features (infraslow oscillation in EEG delta power, sleep spindles, etc.) will be used as stimulus triggers, and a block average GLM approach will be adapted for analysis. For each subject and stimulus type, a response function will be simultaneously fit using GLM SS regression. Aggregated responses will be derived from the individual stimulus events, and the extracted cerebral response will be compared between stimuli.
[0192] Water content changes during sleep: Using the filtered data, we will attempt to estimate the slow cerebral water fraction levels across sleep stages. As shown in small animals, interstitial cerebral water content increases by -10% between sleep and wake, which will increase the optical absorption coefficient of the brain at 970 nm byAtty. Dtk. No. 125141.05019. MGH2O24-394approximately the same factor. To investigate the influence of sleep state on cerebral water content, relative changes in water fraction will be compared between sleep and wake and between different sleep stages.
[0193] GS function decrease with age: Comparisons of the length of low-frequency oscillation periods, normalized by the hours of sleep, will be conducted between younger and older subjects. We will use one-sided t-tests to examine whether age influences the amount of low-frequency oscillations during sleep. While this is an exploratory' aim, if the effect size is large (0.8), with a sample size of 10 old and 30 young subjects the power to detect a difference will be 70%.
[0194] FIG. 17A-C show a multi-taper spectrogram of the long-separation FlexNIRS optical density (OD at 850 nm, after short-separation regression) during sleep. Gray area corresponds to motion artefacts The cardiac, respiratory, and glyniphatic frequencies can be seen to vary in both amplitude and frequency throughout the night, b) Time domain representation of the glymphatic frequency band, with c) extracted segments showing examples of lower amplitude fluctuations, higher amplitude fluctuations and a transition from low to high.
[0195] Skin tone effect: To explore the potential influence of skin pigmentation on the detection of low-frequency GS oscillations, we will conduct grouped analyses based on skin color. These analyses will compare the relative changes of low-frequency oscillations during stable wake, stable sleep, and the difference between sleep and wake. While we anticipate lower SNR in the darker skin group, we do not expect signal degradation at 970 nm, as melanin absorption is significantly reduced at longer wavelengths. We anticipate similar relative changes in GS low-frequency oscillations in the young healthy population regardless of skin tone, as the FlexNIRS employs a multi-distance approach to quantify chromophores that is insensitive to an absorbing thin superficial layer
[0196] Because of intensity changes due to motion artifacts and head position changes in some individuals is possible during a night of sleep, quantification of overnight changes in water fraction levels based purely on intensity may not be feasible. Concentration change estimates using the multi-distance self-calibrating geometry method are less sensitive than those computed with a single distance and without the self-calibrating geometry. Nonetheless, efforts will be made to minimize these artifacts. To prevent probe movement, medical-gradeAtty. Dtk. No. 125141.05019. MGH2O24-394adhesive will be used to secure the GS-FlexNIRS to the skin as needed. Additionally, to minimize head turning during sleep, in addition to ensuring a comfortable sleep environment, subjects will be encouraged to bring their own favorite firm pillow and to sleep on their backs, ii) Measurement sessions impacted by restless sleep or an inability to sleep for at least 4 hours will be excluded from the analysis. Additional subjects will be enrolled as needed to maintain an adequate sample size, iii) If the 12 EEG electrodes montage of the PSG is insufficient for the sleep stimulus driven analysis, we will use the 256-channel geodesic nets and the NA410 amplifier (Electrical Geodesics, Inc., Eugene, OR USA) as in.
[0197] Aim 3 At home feasibility: Evaluate the ability of conducting at-home overnight GS-FlexNIRS measurements, with a pilot study on 10 young subjects.
[0198] In the second year, we will recruit 10 young subjects who are sufficiently techsavvy to test the GS-FlexNIRS at home during one night of sleep As detailed in the human subject section, before the at-home measurements, subjects will receive training on the GS-FlexNIRS and PSG devices in the lab. Additionally, a study staff member will conduct a presleep Zoom session at their homes to guide them in correctly attaching the devices and starting data acquisition. Feedback regarding wearability issues, potential app crashes, or any other issues will be addressed promptly. Data quality will be evaluated by comparing measurements collected at home with those obtained in the lab using the same devices. We will quantify the incidence of motion corrupted data segments, signals and noise levels, deterioration of FlexNIRS symmetry' quality criteria, and EEG lead detachments and compare to the in- lab study.
[0199] Consideration of Sex and Other Biological Variables: Groups will be matched for sex. All analyses will be performed across and within sex to determine any potential differences. This is needed because females, in general, have thinner scalp and skull, thus NIRS measurements in females may have higher brain sensitivity. With this analysis we want to confirm the method works equally well in male and female. Groups will be balanced for race and ethnicity by targeting enrollment of minorities, as well as balanced for skin tone types.
[0200] Scientific Rigor: We have designed the study based on previous literature reports and supported by the preliminary' results. The foundation of the study is grounded in the highly innovative development of a cutting-edge, non-invasive, and cost-effective near-infrared wearable device designed to monitor glymphatic system function duringAtty. Dtk. No. 125141.05019. MGH2O24-394naturalistic sleep. A rigorous device development plan has been proposed. Comprehensive validation studies involving healthy subjects have been carefully designed, considering statistical factors and the anticipated enrollment capacity within the grant period. As the project progresses through each Aim, the limitations of the device will be rigorously evaluated. Co-registration with PSG signals and precise sleep staging analysis provide a well- accepted comparative benchmark and a standard for evaluating the method's ability to measure GS function in humans. As the FlexNIRS is not MRI-compatible, MRI validation falls outside the scope of this initial proof-of-principle study. However, upon successful completion of this study, a future larger project can address MRI validation by developing a fiber-based 970 nm NIRS device and providing support for overnight MRI scan costs. By project's end, we anticipate establishing the feasibility of GS function monitoring with the FlexNIRS in both controlled and home settings. Initial findings highlighting age-related differences will inform larger studies targeting aging populations at risk for cognitive decline. The cost-effectiveness, user-friendliness, and portability of the GS-FlexNIRS make it a valuable resource for interested researchers, and we will explore ways to support both open- source distribution and commercialization. Data collected in this study will be shared with the research community as outlined in the data management and sharing plan.
[0201] I. Specific Aim - oxCCO Measurement Hypoxic ischemic encephalopathy (HIE) remains one of the leading causes of neonatal morbidity and mortality and is a substantial public health problem. Randomized controlled trials have demonstrated a reduction in death or disability in newborns with moderate or severe HIE treated with therapeutic hypothermia (TH). TH is used to protect the vulnerable brain after a period of hypoxia-ischemia (HI) by several mechanisms, including reducing cerebral metabolism to limit secondary' energy failure. To date, therapeutic hypothermia is the only treatment for HIE in standard practice. The gold standard for assessment of brain injury after HI is brain magnetic resonance imaging with spectroscopy (MRI-MRS). However, this study is not typically obtained until the 72-hour TH treatment is complete Therefore, there is a need for informative and real-time bedside tools and biomarkers to continuously monitor cerebral hemodynamics and metabolism and guide intervention, to prevent further brain injury'- and potentially improve neurodevelopmental outcome.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0202] Cerebral oximetry with near-infrared spectroscopy (NIRS) is an established in-hospital modality to monitor brain oxygenation status in infants at risk for ischemia and consequent brain damage. Most recently, robust measurements of changes in the concentration of oxidized cytochrome c oxidase (CCO) in neonates with HIE have been demonstrated by using novel broadband NIRS systems. CCO is a critical component of the electron transport chain, and its activity affects all aspects of mitochondrial function from oxidative phosphorylation to oxidative stress and cell death. The ability to measure changes in the concentration of oxidized CCO (oxCCO) allows for monitoring of cerebral energy metabolism at a cellular level. Hence, simultaneous NIRS measurement of hemoglobin concentration (Hb), hemoglobin oxygenation (SO2), and oxCCO provides complimentary information on brain hemodynamics, oxygenation, and metabolism critical for the assessment of brain health. More specifically, in neonates with HIE, oxCCO / SO2 may help identify (1) neonates with more severe HI injury (prior to detection on MRI-MRS) and (2) response to therapy so real-time, targeted interventions can be optimized and early prognostication can be provided to families
[0203] We have recently built the first 8 wavelength, low-cost, wearable wireless oxCCO- NIRS device. This system is an extension of the LED-based wearable device, called FlexNIRS, which by using custom 8-color LED packages uniformly covers the hemoglobin and CCO spectra between 730 and 910 nm and is able to quantify oxCCO concentration changes in addition to hemoglobin concentration and oxygenation at high sampling rate. The device is operated via Bluetooth by a custom app for an Android tablet. We plan to test this novel oxCCO-NIRS device in neonates with HIE to demonstrate feasibility of the method, validate the oxCCO measure against MRI-MRS, and show initial clinical utility of the oxCCO-NIRS device to predict outcome. The use of a low-cost, wearable, and wireless NIRS device will allow continuous monitoring of neonates with HIE throughout the TH treatment, during rewarming, and after rewarming, so that real-time management decisions can be made for more immediate clinical impact. The specific aims of the project are:
[0204] Aim 1: Optimize the oxCCO-NIRS prototype to acquire oxCCO changes in neonates with HIE. Expand the prototype to bilateral cerebral measurement (dual sensors) and implement a more ergonomically advanced design suited for newborns. Validate discrimination of oxCCO and hemoglobin in phantoms.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0205] Aim 2: Validate the oxCCO-NIRS device in 6 term neonates with transient respiratory distress. Test and characterize oxCCO-NIRS in 6 term newborns with transient respiratory distress admitted to the MGH Neonatal Intensive Care Unit. Demonstrate correlation between changes in oxCCO and periods of oxygen desaturation.
[0206] Aim 3: Collect oxCCO-NIRS data for 20 neonates with moderate / severe HIE and correlate changes in oxCCO with presence and type of brain injury'- on post-TH brain MRI. Use the oxCCO-NIRS device to measure cerebral hemodynamics and metabolism in 20 neonates with HIE (10 with moderate and 10 with severe HIE) during hypothermia treatment (first 72 hours), during rewarming (—6-12 hours), and after rewarming (1-2 days), and compare SO2, Hb, and oxCCO during these time periods. Classify neonates with HIE by presence and type of brain injury' on post-TH brain MRI with spectroscopy and correlate with CCO oxidation state. While this is an early "proof of principle" pilot study, we hypothesize that an increased frequency of dynamic changes in cerebral metabolism during each epoch will correlate with presence and severity of brain injury and may provide an individualized opportunity for future treatment to improve outcome.
[0207] II Research Strategy. A. Significance. HIE is estimated to affect. 1-3 per 1000 live births in the developed world and even more in the developing world, contributing to significant disability worldwide. Therapeutic hypothermia (TH) is the standard therapy to treat newborns with moderate or severe encephalopathy. Studies in the mid 2000s demonstrated that among this population. TH reduced death or disability by 20%. Despite treatment with TH, data from 46 Level III NICUs in the United States highlight that among term neonates, HIE remains the most significant contributor to overall NICU mortality (estimated at 23% in neonates >36 weeks gestation at birth). Studi es of neonates with encephalopathy treated with TH suggest the incidence of brain injury is 38-61%. While consensus guidelines exist for management of brain health during therapeutic hypothermia, there is minimal opportunity to make direct impact on cerebral hemodynamics and metabolism at the bedside with the goal of preventing brain injury. Therefore, non-invasive, safe, and real-time bedside tools to monitor brain health during TH are needed to improve clinical practice and management during the critical periods of TH, rewarming, and even after rewarming. Numerous real-time biomarkers have been evaluated as predictors of brain injury'; however, their measurement in serum, urine, and cerebrospinal fluid often requires an invasive procedure, making them less attractive toAtty. Dtk. No. 125141.05019. MGH2O24-394clinicians caring for these neonates at the bedside The evaluation of mitochondrial dysfunction using non-invasive brain monitoring approaches has shown promise in prediction of brain injury' and may allow for other successful treatment approaches. The group has extensive experience with non-invasive brain monitoring, including near infrared spectroscopy (NIRS) and diffuse correlation spectroscopy (DCS) in neonatal populations. We have piloted the use of NIRS and DCS in extremely low gestational age neonates to better understand the predisposition to intraventricular hemorrhage. We have acquired time point measurements in infants during cardiac surgery and during TH in neonates with HIE and have shown how cerebral oxygen metabolism changes with body temperature.
[0208] This proposal aims to expand the previous work, to use a new custom-made, wearable, multicolor NIRS device with the ability to measure the oxidation state of CCO (oxCCO-NIRS) This device is an extension of the NIRS cerebral oximeter device, called FlexNIRS. FlexNIRS differs from commercially available cerebral oximeters by being miniaturized, wearable, wireless, and low cost (<$100 for all components). The original FlexNIRS includes two dual -LED (735 and 850 nm) and 3 photodiode detectors arranged in a symmetric geometry' (source-detector separations 0.8, 2.8 and 3,3 cm), which allows for a selfcalibrated multi-distance method to recover hemoglobin oxygenation (SO2) and hemoglobin concentration (Hb) changes at a rate of 266 Hz. The device is powered by a rechargeable battery (8 hours of continuous operation) and uses Bluetooth Low Energy (BLE) for wireless communication with a custom mobile application for Android tablets. Characterization tests in phantoms and human participants show very low noise (NEP of 70 fW / VHz) and robustness of SO2 quantification in vivo (±6%). The new multi -wavelength FlexNIRS prototype, includes custom 8 wavelength LED (light emitting diode) packages and updated driving electronics to acquire 8-colors in the spectral range of 700-900 nm, needed to separate the redox state of cytochrome c oxidase from hemoglobin concentration changes. In fact, based on the oxCCO literature, the use of 8-wavelengths is sufficient to recover oxCCO with a 2% error. Some have reported encouraging results even when using only 4 LEDs. The oxCCO-NIRS device is the first wearable, wireless multicolor system, aiming to monitor oxCCO changes in newborns
[0209] The utility of measuring oxCCO in neonates with HIE has been investigated by Dr. Bale and Prof. Tachtsidis in the UK, in a series of studies using fiber-based NIRS devices. In a study with 23 neonates they showed the oxCCO metabolic responses to physiologicalAtty. Dtk. No. 125141.05019. MGH2O24-394changes were significantly different between mild and severe brain injury In another study, with 24 neonates experiencing periods of peripheral oxygen desaturation, they demonstrated that a strong relationship between oxCCO and SO2 was associated with unfavorable outcome, and was able to predict brain injury severity with 64% sensitivity and 79% specificity.
[0210] Rigor of the Prior Research: Previous investigations by the group and others have demonstrated the clinical utility of non-invasive monitoring with diffuse optical methods neonates with HIE. These results motivated us to develop a wearable low-cost oxCCO-NIRS system tailored for this neonatal population and application. We aim to extend the findings described above, by demonstrating feasibility of continuous monitoring throughout the entirely of TH, rewarming, and after rewarming (which has not previously been described), as we believe measurement of SO2 and oxCCO can provide invaluable insight into brain health during these distinct periods of time. We will be able to do this using the technology we have developed, an innovative device suitable for prolonged monitoring periods in low' resource settings, including typically underserved communities, and enable new community and telehealth applications. In Aim 1, we propose to expand the design of the multi-color FlexNIRS device with the introduction of bilateral neural recording for expanded cerebral coverage allowing for more robust metric quantification. We plan to reduce the size of the device to achieve optimal brain sensitivity in newborns, and also implement design optimizations for improved wearability for comfortable continuous monitoring and robust signal production. We will test the validity of the sensor in a tissue-like phantom and compared to gold standard high performance NIRS devices. In Aim 2, we will test the reliability and validity of the acquired parameters through repeated measurement sessions in neonates admitted to the NICU with mild respiratory distress. This will allow us to test the device and measure oxCCO during periods of desaturation in relatively healthy neonates before moving to measurements in the population of interest. In Aim 3, we will monitor neonates with moderate and severe HIE and correlate changes in oxCCO with brain injury in hopes of creating a prediction model and identifying periods of time during TH, rewarming, and after rewarming where targeted interventions may improve outcomes. The work proposed fits uniquely under the primary efforts of the investigators. Dr. Franceschini’s group is highly experienced in the development and clinical translation of near-infrared spectroscopy devices. Dr. Herzberg’s group has the necessary knowledge to guide the development of the deviceAtty. Dtk. No. 125141.05019. MGH2O24-394from the application perspective and to validate it in patients. The external collaborator, Dr Gemma Bale, has extensive expertise in oxCCO measurements with broadband and multiwavelength devices, and will guide us on the data analysis to optimize discrimination between hemoglobin and oxCCO changes.
[0211] If successful, this technology has the potential to have direct and immediate impact on bedside interventions to optimize brain health and could be used as clinical standard of care.
[0212] B. Innovation. The device we have built and plan to optimize for use in neonates will be the first non-invasive, bedside, wearable, multi-color continuous monitoring system to measure SO2, Hb, and oxCCO concentration changes using NIRS technology. This project is innovative in several ways:
[0213] Continuous monitoring of 502, Hb concentration, and oxCCO in neonates with HIE during TH, rewarming, and after TH / rewarming is innovative. As described above, while other investigators, including Dr. Bale's group, have demonstrated utility of NIRS technology for monitoring oxCCO, this monitoring has been limited to several hours. This project is innovative in its goal to continuously monitor neonates with HIE throughout distinct periods of changing physiology (during TH, during rewarming, and after TH / rewarming) to better understand the relationship between cerebral metabolism and brain injury.
[0214] The proposed adaptation to miniaturize the wearable device is vital for successful implementation of the oxCCO-NIRS device in the NICU. A wearable, wireless device provides several advantages in the NICU setting. Neonates admitted to the NICU require multiple forms of cardiorespiratory and neurologic monitoring. Wired devices are a challenge when neonates move and traditional neuromonitoring probes (NIRS, DCS) can potentially lose contact or signal when the baby is repositioned or held. A wearable device has ease of application and the potential to be better tolerated by the neonate with better signal with movement and holding and no concerns regarding dislodgement of wires. Additionally, the low-cost nature, easy wearability, and long battery life make this innovative device suitable for continuous monitoring even in low resource settings, which enables new community and telehealth applications.
[0215] We aim to identify points of intervention to change clinical practice and neonatal outcomes by continuously monitoring 502, Hb concentration, and oxCCO in neonates. While the clinical use of NIRS to measure oxygenated hemoglobin has been adopted by some NICUsAtty. Dtk. No. 125141.05019. MGH2O24-394throughout the United States, SO2 is only a proxy for cerebral perfusion. Therefore, clinical NIRS helps to establish a trend, but doesn't allow the medical team to respond with a specific intervention in the majority of cases, nor does it allow for any monitoring of oxCCO or other markers of cerebral metabolism. If this project is successful, it will provide a bedside biomarker that will allow the medical team to understand the state of cerebral perfusion and metabolism in real time during TH, rewarming, and after rewarming and to intervene to prevent ongoing or further brain injury, potentially impacting long term neurodevelopmental impairment.
[0216] C. Approach In this proposal, we aim to miniaturize wearable, wireless oxCCO-NIRS device for use in neonates. We have tested and validated the FlexNIRS prototype device in phantoms and shown its ability to spectrally discriminate small changes in chromophores absorption in liquid phantoms (FIG. 18A--D).
[0217] Aim 1: Optimize the oxCCO-NIRS prototype to acquire oxCCO changes in neonates with HIE. Dual sensor multi-wavelengths FlexNIRS device implementation: The oxCCO-NIRS system will be based on the latest FlexNIRS prototype (FIG. 18A-D). The system relies on custom 8-color LEDs (COB-8040-MGH-X; Marubeni America Corporation), extremely sensitive fast PIN photodiodes (PD) (VEMD5160X01, Vishay Semiconductors) and a dedicated ultra-low-power Analog Front End (AFE) integrated circuit (IC) (AFE4950, Texas Instruments). The 8-color LEDs are manufactured in a custom ceramic package of 2.2 mm x 2.9 mm dimension and hosts 8 independent LEDs, emitting at 735, 765, 785, 805, 830, 855, 870, 910 nm (FIG. 18A-D). The wavelengths have been selected to uniformly cover the near-infrared hemoglobin and oxCCO spectra. While broadband NIRS devices work best at eliminating Hb crosstalk in the recovery of oxCCO, it has been shown that a combination of discrete wavelengths can resolve oxCCO changes with sufficient accuracy. From a practical viewpoint, the discrete wavelength approach is the only viable method to avoid the need of an optical spectrometer per detection channel. As in the original FlexNIRS, we use the selfcalibrating geometry' based on 2 LEDs and 2 photodetectors (PD) symmetrically placed to provide two source-detector separations. The symmetric geometry allows for compensating for inter-device variability without need for external calibration, provides quality control metrics, and the two distances increase the accuracy in estimating cerebral SO2 and changes in Hb and oxCCO. A third PD placed at a short separation (<1 cm) enables the acquisition ofAtty. Dtk. No. 125141.05019. MGH2O24-394superficial scalp signals, for subtraction from the long source-detector channels to enhance depth sensitivity. The AFE4950, originally built for wearable optical heart-rate monitoring and pulse oximetry, allows for driving 8 independent LEDs and sampling 4 photodetectors (2 in parallel). In order to drive two 8-color LED modules (16 LEDs total), an external switch, controlled by the microcontroller, is used to select the active 8-color LED module. The short SD separation and background ambient light are also sampled, leading to an overall number of 36 phases driven by the AFE IC at 66 Hz sampling rate. A 3-axis accelerometer IC (ADXL344, Analog Devices) is embedded on the printed circuit board (PCB) to allow tracking patient movements and detecting and rejecting motion artifacts in the optical data. Finally, a microcontroller runs the measurement routine: programming the AFE, acquiring the sampled light and motion signals, and transmitting data to the external tablet through a Bluetooth low energy 5.2 communication (BGM220S, Blue Gecko Bluetooth® Si P Module, Silicon Labs). The 8-color FlexNIRS when operating at 66 Hz achieves a NEP of about 100 fW / VHz and a dynamic range of 80 dB.
[0218] FIG. 18A--D show a photo of the (a) 8-color FlexNIRS prototype, b) Photographs of the FlexNIRS prototype embedded in an infant cap. c) Marubeni LED package and measured spectra, d) Recovered Ink spectra in an intralipid / water solution. The high SNR of the device allows to recover small changes in absorption at 8 wavelengths.
[0219] We will modify the current 8-color FlexNIRS device to acquire bilateral oxCCO measurements in neonates. This will be done by adding another AFE to control a second optical sensor. A simplified block diagram of the dual system is shown in FIG. 19. The oxCCO-NIRS device will be manufactured on a flex PCB, allowing the hosting of narrow¬ pitch ICs while offering maximum flexibility to accommodate for the head curvature of the patient and maximize the optical coupling. The source-detector separations and overall dimension of the sensors wall be reduced to optimize measurements in new-borns. Specifically, the source-detector separation will be 0.7, 2..4, and 2.8 cm and the overall oxCCO-NIRS dimension will be 8 cm x 1 cm. To account for different head circumferences, we will build additional units with overall lengths of 9, 10 and 11 cm (more sizes can be done if necessary). This is to ensure the 2 optical sensors will be positioned symmetrically in the middle left and right forehead. The device will be encapsulated in a low-profile injection headband or baby¬ cap as shown in FIG. 18A--D. The sensor area will have a cutout on the inner side for contactAtty. Dtk. No. 125141.05019. MGH2O24-394with the skin. Finally, a clip-in socket will be either attached on the baby clothes or positioned on the bed above the head, for the battery. Magnetic pogo-pin connectors will be used to allow for easy battery replacement; estimated battery life will be 5 hours when using the same medical-grade Li-Po battery' as the current FlexNIRS, longer if we will employ larger batteries The weight of the entire assembly is expected to be around 2-3 ounces (-60-80 grams), somewhat more than a modern smart watch but less than half the weight of a typical modern smartphone.
[0220] FIG. 19 shows a simplified block diagram of the proposed oxCCO-NIRS device molded case and embedded in a fabric.
[0221] The plan is to finalize the circuit design and manufacturing by month 3 of year 1, thus completing packaging and characterization by month 6 of year 1. Meanwhile the current Android app will be modified to control and display the dual sensor data, estimate oxCCO changes in addition to currently available SO2 and Hb concentration changes, and provide additional data quality feedback.
[0222] Phantom validation tests: To validate the quantification of the biologically relevant hemoglobin species and CCO redox state we will perform intralipid-blood-yeast phantom experiments as previously described by Lange et al. and Sakata et al. To this end the phantom base will consist of phosphate-buffered saline solution to maintain a physiological pH of 7.4, to which 20% Intralipid will be added for scattering, and continuous stirring on a hot plate will be done to maintain a physiological temperature of 37°C. This will be followed by: injection of increasing concentration of bovine blood, yeast to add CCO to the solution and to help modulate oxygenation, achieved by sealing the container and flowing oxygen through silicone tubing. The deoxygenation / oxygenation cycle will be repeated several times at different concentrations of blood, and the chromophores concentration changes including oxCCO will be measured throughout. Experiments will be done with the oxCCO-NIRS device waterproofed with epoxy resin and if necessary sealed with transparent plastic. A simultaneous measurement using a frequency-domain (FD) NIRS system (ISS MetaOx) will be performed as well to monitor scattering (not expected to change) and to validate hemoglobin changes. Measurements using an imaging spectrometer composed of a Pl-Acton 2150i spectrograph and a Princeton Instruments PhotonMax cooled back-illuminated EMCCD camera as an alternative broadband NIRS device will also be done to further validate the oxCCO-NIRSAtty. Dtk. No. 125141.05019. MGH2O24-394oxCCO quantification. Additionally, we can further validate the oxCCO-NIRS oxCCO quantification using measurements from a broadband NIRS provided by Dr, Bale's group.
[0223] Aim 2: Validate the oxCCO-NIRS device in 6 term neonates with transient respiratory distress. Subject recruitment and methods: A goal of 6 term neonates with transient respiratory distress, but otherwise healthy, will be recruited from the MGH NICU to help validate the oxCCO-NIRS device in the neonatal population. We will aim to start recruiting the validation cohort approximately 6 months into the grant period, as we anticipate it will take approximately 6 months to optimize the device for neonatal application and validate it in phantoms. Parents will consent to participate using an informed consent document approved by the MGH Human Research Committee and MGH's Institutional Review Board. Inclusion criteria for validation cohort: about 36 weeks gestation, <28 days old, admitted to the NICU for respiratory distress, clinically stable (as deemed by medical team). Exclusion criteria for validation cohort: contraindications to wearable oxCCO-NIRS placement, congenital / genetic abnormalities. In this validation cohort, we aim to enroll 3 neonates each with light and dark skin to validate changes in oxCCO in neonates with different skin melanin production. Because melanin is only present on the thin skin layer, by using the multi -di stance method we do not expect differences in the oxCCO concentration changes estimates across skin types, as demonstrated by Dr. Franceschini's group. Nevertheless, with darker skin we expect a lower si nal-to-noise ratio (SNR) and we will evaluate the effect of noise on the oxCCO estimates. More neonates will be recruited if needed to further validate this point.
[0224] We will start monitoring these neonates upon admission to the NICU once parental consent is obtained and will aim to monitor continuously for several days. By monitoring for 2-3 days, we will be able to assess changes due to sensor replacement and repositioning, sensitivity to motion, and infant position, by comparing periods when the baby is on a warmer versus when being held by parents or being cared for by nurses. We selected term neonates with transient respiratory distress as the validation cohort to be able to demonstrate changes in oxCCO during periods of mild oxygen desaturation, defined as SpO2 <90%. By monitoring continuously for 2-3 days, we expect to detect 5-10 relative desaturation events per baby. With 6 subjects and at least 40 desaturation events, we will be able to detect a Cohen's d effect size of 0.6 (coefficient of variation=1.4) with p below 0.05 and a statistical power of 0.9. In addition, in this group, we will perform FD-NIRS point measurements once or twice a day, toAtty. Dtk. No. 125141.05019. MGH2O24-394assess the robustness of the oxCCO changes estimates with or without known tissue scattering. We hypothesize oxCCO concentration will decrease during oxygen desaturation. oxCCO data analysis: The main challenge of using NIRS to measure oxCCO in biological tissues is the low concentration of CCO with respect to the concentration of hemoglobin. While the concentration of CCO in the human brain is higher than in extracerebral tissues and the CCO extinction coefficient in the 750-900 nm spectral range is higher than the extinction coefficient of hemoglobin, the contribution of CCO to the overall absorption in brain tissue is no more than 5-10% of that of hemoglobin. Therefore, to isolate changes in light attenuation due to oxCCO from the much larger changes due to Hb, multi-wavelength approaches have been shown to be necessary. Arifler and colleagues have experimentally demonstrated that while a 3 -wavelength combination leads to mean recovery errors of up to 10%, these errors drop to <4% with 4 or 5 wavelengths and <2% with 8 wavelengths. Others have reported encouraging results when using only 4 LEDs, available in commercial cerebral oximeters. Dr. Gemma Bale will guide us in analysis of oxCCO and discrimination from oxy- and deoxyhemoglobin states.
[0225] To calculate the concentration changes of oxy-, deoxyhemoglobin concentrations, and oxCCO we will use the modified Beer Lambert law, AA(A) = p(A)kAck• fk(A), where A is the measured light attenuation, p is the differential pathlength, Ackis the concentration change of the kthchromophore, and £kis the corresponding extinction coefficient. The changes in chromophores concentration are derived from the change in attenuation spectrum measured at the multiple wavelengths using a classical least-squares process and applying variations in the assumptions as previously described. We will compare results by assuming rather than measuring a differential pathlength factor as is done with the FD-NIRS We will implement multi-distance approaches as in work by Dr. Bale's group and others. We will also consider the oxidized-reduced spectra relationship to improve quantification of oxCCO changes. Finally, in optimizing the processes, we will take into account the LED spectral width and the -2.5 mm maximum difference in separations of the various LEDs in the package. The quantification of oxCCO concentration changes will also be approached theoretically using two-layer Monte-Carlo simulations. For the simulations we will use realistic optical properties of the tissue chromophores, we will input the measured LED spectra and we will use a short and two long separations as in the oxCCO-NIRS device. The crosstalk between the chromophores will be calculated pair-wise as the relative errorAtty. Dtk. No. 125141.05019. MGH2O24-394observed in one chromophore subsequent to the induction of a concentration change in another chromophore.
[0226] Aim 3: Collect oxCCO-NIRS data for 20 neonates with moderate / severe HIE and correlate changes in oxCCO with presence and type of brain injury on post-TH brain MRI Subject recruitment and methods: Concurrent to validation of the oxCCO-NIRS device, as described above, we will recruit neonates with moderate or severe HIE, aiming to enroll 10 neonates in each group. Encephalopathy severity is determined and documented by the medical team based on the clinical neurologic exam (Sarnat exam) prior to treatment with therapeutic hypothermia. Neonates with mild encephalopathy, who may be receiving therapeutic hypothermia, will not be included in this study. Again, parents will consent to the study using an informed consent document approved by the MGH Human Research Committee and MGH's Institutional Review Board. Inclusion criteria: D36 weeks gestation (this is the target population for TH), admitted to the NICU with evidence of perinatal asphyxia and being treated with TH. Exclusion criteria: contraindications to wearable oxCCO-NIRS placement, congenital / genetic abnormalities, early plan for redirection of care. We will bilaterally monitor these neonates during TH treatment (72 hours), during rewanning (-6-12 hours), and after rewarming (1-2 days) and measure SO2, Hb changes, and oxCCO changes continuously during these time periods. The overall monitoring period will be 5-6 days ideally. In the MGH NICU, on average, we treat approximately 40 neonates per year with therapeutic hypothermia. Therefore, with over 1 year of recruitment / enrollment / measurements, we anticipate no challenges in reaching the enrollment goal of 20 neonates. All neonates treated with therapeutic hypothermia have a brain MRI-MRS completed after rewarming. We will subsequently classify neonates with HIE by presence and type of brain injury on post-TH brain MRI-MRS, using an establish MRI-scoring system and will correlate these findings with the concentration of oxCCO. We will compare the metrics (oxCCO, SO2, and Hb concentration changes) between those neonates with moderate and severe HIE and between those with and without brain injury on MRI-MRS and stratified by severity of brain injury determine by the Weeke classification system. We will compare metrics between two groups using the Student's T-test and means between three or more groups using a one-way analysis of variance test (ANOVA) With a standard deviation on the oxCCO measure of 5%, we will be able to detect an effect size of 1.4 between the two groups with a power of 0.9.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0227] Statistical analyses and considerations. All statistics will be performed after examination of the primary variables for di stribution and to assure that all variables meet the assumptions for each statistical model. Additionally, we will perform each test using the primary' variables alone as well as with statistical covariates when appropriate (e g., controlling for demographic factors such as sex, gestational age, illness acuity, presence of seizures).
[0228] i) The performance of the original FlexNIRS and of the current 8-color FlexNIRS design of the oximeter has been evaluated as described in the preliminary' studies and publications, and therefore we believe that the technical development components of this work for optimizing the device are relatively low risk and will be done promptly during the first 3 months of the grant. Any setbacks will be addressed at that time using alternate design principles. Any delay will be addressed by starting the human subjects measurements with the current prototype, ii) Patients will be excluded if we aren't able to monitor during at least two of the three distinct time points (during TH, during rewarming, and 1-2 days after rewamiing). Possible causes include difficulties with sensor contact, inability to reach the brain due to extracranial bleeding / large subarachnoid space, clinical instability as decided by the medical team caring for the neonate. Additional patients will be recruited if necessary, iii) As stated above, given the average number of neonates admitted to the MGH NICU for TH in one year, we don’t anticipate difficulties achieving the enrollment goals. However, if after one year, we are behind schedule, we will add to the study the Brigham and Women's (BWH) NICU. We have collaborated with the BWH NICU and Dr. Mohamed El-Dib on several prior studies (5R01HD091067). Given the low cost and availability of the oxCCO-NIRS devices, both the device and the Android tablet will be easily replicated in the two hospitals if needed, iv) We will need to account for the oxCCO-NIRS device being removed for purposes of obtaining the post-TH MRI-MRS (as the device is not MRI-compatible), so this may interfere with continuous monitoring in the 1-2 days post-TH. However, given the short nature of the brain MRI-MRS (typically 30-45 minutes), this isn't anticipated to greatly impact the monitoring window, v) Given the innovative wearable feature of this device, we don’t anticipate any insult to skin integrity that comes with existing optical sensors that need to be directly applied to the forehead. However, if we come across any issues with skin integrity, we will work to optimize the device design and add hydrogel and other skin barriers as needed as a protective interface between device and the skin, as we have done previously in extremely premature infants.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0229] Sex as a biological variable. Groups will be matched for sex All analyses will be performed across and within sex to determine any potential differences. All analyses will be performed with and without sex as a covariate, and direct comparisons will be performed between females and males to directly examine potential sex differences for each primary outcome. Groups will be balanced for race and ethnicity by targeting enrollment of minorities, as well as balanced for skin types, by enrolling subjects with light, medium, and dark skin pigmentation, based on the Fitzpatrick classification.
[0230] Rigor and reproducibility. This proposal is dedicated to establishing rigorous and reliable measures of the oxCCO-NIRS system in neonates affected by HIE. The generic primary analyses for each aim are: Aim 1: a) seamless hardware / mobile software functionality; b) signal equivalent of better to the quality of the original FlexNIRS system, c) phantom validation of oxCCO changes estimates. Aim 2: a) assess sensitivity of oxCCO changes due to deoxygenation events in infants; b) effect of skin color; c) sensitivity to repositioning and motion; c) effect of scattering on oxCCO quantification. Aim 3: a) significant statistical association between oxCCO changes during treatment and HIE severity and hypoxic-ischemic brain injury on brain MRI with spectroscopy.
[0231] Overall, these Aims would demonstrate the value of the oxCCO-NIRS system for monitoring neonates with HIE as they would demonstrate the feasibility of continuous monitoring in this population and suggest optimal support of cerebral metabolism and possible areas of target intervention.
[0232] We have designed the study based on previous literature reports and supported by the preliminary results. The foundation of the study is a newly developed oxCCO system able to continuously monitor cytochrome c oxidase. A rigorous device development plan has been proposed, and a pilot clinical study has been designed, accounting for statistical considerations, expected achievable enrollment, and patient population characteristics. Comparison with gold standard measures offers a well-accepted comparative predictor and a standard for evaluating advantages of non-invasive oxCCO monitoring in newborns. For each Aim, the limitations of the device will be thoroughly evaluated.
[0233] I. Specific Aim - Blood Pressure Measurement. Intra-arterial blood pressure (ABP) monitoring is essential for the diagnosis and management of hypertension hypotension, and hemodynamic instability. Neonates, particularly those born preterm, exhibit uniqueAtty. Dtk. No. 125141.05019. MGH2O24-394physiological characteristics and vulnerabilities that necessitate precise monitoring of blood pressure for optimal clinical management The ability to detect and prevent fluctuations in blood pressure is crucial in preventing complications such as germinal matrix / intraventricular hemorrhage, which can profoundly impact neurodevelopment Additionally, accurate blood pressure measurement is essential for guiding interventions in neonates with various medical conditions, including respiratory distress syndrome, pulmonary hypertension, congenital heart disease, hypoxic-ischemic encephalopathy (HIE), sepsis, or necrotizing enterocolitis. Medications to treat low blood pressure are used in 10% of infants less than 29 weeks gestation and about 30% of term infants suffering from HIE. However, invasive blood pressure is used less frequently in neonates because of the associated risks. Umbilical artery catheters and peripheral arterial lines, commonly used for ABP monitoring, pose significant risks such as line malposition, vascular accidents, thromboembolism, and other complications. The most commonly used intermittent upper arm oscillometer ABP monitoring, is known to be less accurate and precise, potentially missing dangerous hypotensive or hypertensive episodes. Addressing this gap, there is an urgent need for continuous non-invasive arterial blood pressure (CNAP) monitors tailored to the neonatal population.
[0234] While CNAP technologies are quickly advancing, most methods are cuff-based volume-clamp devices, which continuously partially inflate the finger or wrist cuff to a pressure that minimizes changes in arterial expansion to keep the blood volume constant during each pulse. While some feasibility data exist for young infants and children, this methodology cannot identify absolute blood pressure in preterm infants and the constant pressure from the cuff poses safety concerns for infants’ small and fragile limbs. Cuffless devices using optical sensors (photoplethysmography, PPG) in the finger or the wrist can measure changes in pulsatile blood volume and from that determine pulse transit time (PTT) analysis to extract BP values. While PTT analyses, showed a promising correlation with blood pressure changes, these techniques require multiple PPG sensors, frequent calibrations, and have limited reports in the neonatal population.
[0235] To monitor ABP in newborns continuously and non-invasively, we propose to use a cost-effective wearable device utilizing wireless LED-based near-infrared spectroscopy (NIRS) technology, the FlexNIRS. FlexNIRS acquires high-temporal-resolution NIRS-PPGAtty. Dtk. No. 125141.05019. MGH2O24-394cerebral pulsatile waveforms with a strong signal-to-noise ratio (SNR) at a sample rate of 266 Hz at 3 cm source-detector separation. Preliminary results in adults indicate that the time derivative of the optical pulse waveforms (d / dt(NIRS-PPG)), related to pulsatile blood flow, provides specific morphological features indicative of blood pressure changes. Utilizing machine learning techniques such as linear models and k-fold cross-validation techniques, the group achieved a good correlation (R2=0.83) between true ABP and the estimates in adults. Based on these preliminary7results, we hypothesize that the d / dt(NIRS-PPG) morphological signal collected on the head offers a more reliable measure of blood pressure compared to peripheral signals because it is closely linked to central pressure and less susceptible to environmental factors like touch and temperature. The preliminary data from newborns demonstrate similar morphological features in NIRS-PPG signals, supporting the potential for d / dt(NIRS-PPG) to assess blood pressure in this population The FlexNIRS features low-power LEDs for safe prolonged use, and a wireless design for easier integration into existing care protocols for critically ill infants in the Neonatal Intensive Care Units (NICUs), and post-discharge home care settings.
[0236] The aim is to investigate and fully characterize the relationship between d / dt(NIRS-PPG) and blood pressure in neonates. By developing a reliable, cost-effective CNAP-FlexNIRS monitor tailored for the newborn, this project seeks to improve clinical practice by better monitoring and managing blood pressure in this vulnerable population. The Specific Aims of the Project are: Aim 1. Develop a CNAP-FlexNIRS device tailored specifically for neonatal extended monitoring of blood pressure, capable of collecting high-temporal-resolution, high-SNR, NIRS-PPG pulsatile waveforms. Aim 2. Test the CNAP -FlexNIRS in 20 NICU patients undergoing intra-arterial blood pressure monitoring for clinical care, including infants of different gestational ages, races, ethnicities, and medical conditions. Aim 3. Develop a NIRS-PPG waveform analysis process to recover ABP. Utilize the collected data for training and testing deep learning models for blood pressure estimation. Assess the correlation between ABP estimated with the CNAP -FlexNIRS device and simultaneously measured intra-arterial blood pressure.
[0237] These initial studies are crucial for optimizing the CNAP-FlexNIRS device and refining data analysis processes before scaling up to a broader clinical investigation aimed at translating this innovative monitoring approach to clinical practice.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0238] II. Research strategy. A. Significance Continuous arterial blood pressure (ABP) monitoring is critical for unstable newborn infants to prevent brain injury' caused by dysregulated swings in cerebral perfusion driven by hyper- or hypo-tension episodes in the absence of robust autoregulation. These fluctuations pose a significant risk to the underdeveloped cerebral vasculature, potentially resulting in conditions such as germinal matrix / intraventricular hemorrhage, which can profoundly affect the neurodevelopmental outcomes of preterm infants. Additionally, accurate blood pressure measurement is essential for guiding interventions in neonates with various medical conditions, including those with respiratory distress syndrome, pulmonary hypertension, congenital heart, disease, hypoxic-ischemic encephalopathy (HIE), sepsis, or necrotizing enterocolitis. Indeed, medications to treat low blood pressure are used in 10% of infants less than 29 weeks gestational age (GA) and -30% of term infants suffering from HIE.
[0239] The gold standard for blood pressure measurement in newborns is intra-arterial line BP (A-line BP) monitoring. However, this involves catheterization of the umbilical or a peripheral artery and can lead to complications such as vasospasm, nerve damage, ischemia, thrombosis, and, in severe cases, limb amputation. Due to these risks, A-line BP monitoring is only used when deemed absolutely necessary in critically ill neonates. The alternative commonly used in the neonatal intensive care unit (NICU) is intermittent ABP checks using sphygmomanometers on peripheral sites like the arm or leg through automatic oscillometry' However, such techniques have suboptimal precision and can miss dangerous hypo / hyper-tensive episodes. Currently, we lack a noninvasive method that can be used with complete clinical trust for blood pressure monitoring.
[0240] One potential solution involves a continuous monitoring method using a volume¬ clamp photoplethysmography (PPG) approach. This technique entails applying counter pressure to mirror volume changes in a finger (in adults) or wrist (in newborns), thereby reflecting changes in arterial blood pressure. While some feasibility data exist for young infants and children, this methodology cannot identify absolute blood pressure in preterm infants, and the constant pressure from the cuff poses safety concerns for infants’ small and fragile limbs. Additionally, cuffless techniques employing optical sensors PPG in the finger or the wrist can measure changes in pulsatile blood volume and determine pulse transit time (PTT) analysis to extract ABP values. However, these techniques require multiple PPGAtty. Dtk. No. 125141.05019. MGH2O24-394sensors, frequent calibrations, and have limited reports in the neonatal population Furthermore, recently proposed machine learning approaches leveraging detailed pulse wave features show promising results in inferring ABP but lack validation in infants or larger sample sizes, which is required to be considered for clinical use.
[0241] We recently developed a low-cost, wearable wireless light emitting diode (LED)-based near-infrared spectroscopy (NIRS) device named FlexNIRS, capable of acquiring pulsatile signals at a large source-detector separation (NIRS-PPG) with a high signal-to-noise ratio (SNR) at a sample rate of 266 Hz. Using this device in adults, we have shown that the time derivative of the optical pulse waveforms, d / dt(NIRS-PPG), is related to pulsatile blood flow, and, more importantly, specific morphological features of the d / dt(NIRS-PPG) signal collected at large source-detector separations (2.8-3.3 cm) in the forehead show7a strong correlation with changes in blood pressure (FIG. 20A--B). Conversely, signals acquired at shorter separations (0.8 cm) demonstrate greater inconsistency and weaker correlation with invasive ABP. Building upon these initial findings, we hypothesize that morphological features derived from the time derivative NIRS-PPG signals obtained from the forehead and brain offer a more reliable measure of blood pressure compared to signals acquired from peripheral sites. This is attributed to the closer association of the pulsatile signal in the head with central pressure, as well as reduced susceptibility to vasoconstriction and external factors such as touch and temperature. Furthermore, employing larger source-detector separations enables the investigation of deeper tissue volumes thereby reducing sensitivity to local variations such as superficial veins, which are known to impact PPG signals in the scalp. Initial data collected in newborns showed that the NIRS-PPG signals and their time derivatives exhibited similar morphological features to those observed in adults (FIG. 21).
[0242] Therefore, we hypothesize that NIRS-PPG morphology can be used to effectively assess blood pressure continuously and non-invasively in neonates using Continuous Non-Invasive blood Pressure- FlexNIRS, (CNAP -FlexNIRS). In Aim 1, we plan to modify the FlexNIRS to enable higher sampling rate and improve wearability in neonates, ensuring comfortable continuous monitoring and enhancing signal detection robustness In Aim 2, we will assess the performance of the device in 20 NICU patients undergoing intra-arterial blood pressure monitoring for clinical care. This diverse patient population will includeAtty. Dtk. No. 125141.05019. MGH2O24-394neonates of varying GAs, races, ethnicities, and medical conditions In Aim 3, we will refine the NIRS-PPG waveform analysis process to accurately recover ABP. We will utilize the collected data to train and test deep learning models for blood pressure estimation, optimize calibration procedures and test correlation between ABP estimated with the NIRS-PPG device and the simultaneously measured intra-arterial blood pressure.
[0243] Rigor of the Prior Research: While significant progress has been made in the development of continuous noninvasive ABP monitoring devices for adults, a notable gap persists in technology suitable for neonatal populations. Neonates' small and fragile physiological monitoring sites present unique challenges that have impeded the advancement and commercial availability of infant ABP monitoring devices. Furthermore, investigations into noninvasive ABP monitoring for neonates have predominantly focused on methods that assess peripheral vascular changes usually significantly far from the heart and susceptible to motion artifacts. Such limitations necessitate extensive signal processing and the compilation of large datasets for process training, further complicating the development of viable neonatal ABP monitoring solutions.
[0244] In contrast, the proposal, which is centered around the head-wearable device CNAP-FlexNIRS, seeks to overcome these limitations by employing a novel approach which involves monitoring pulsatile blood volume changes in deeper tissue in the forehead. This method offers distinct advantages, primarily stemming from the inherently stable nature of head vasculature, which is less susceptible to external factors. Encouraging preliminary results in both adults and infants underscore the potential of this approach.
[0245] Investigator Team: The proposal fits seamlessly with the primary expertise of the investigators. Dr. El-Dib’s team has nearly a decade of experience in newborn brain research at the Brigham and Women’s Hospital (BWH) NICU, providing the requisite knowledge, staff, and clinical scientists necessary to conduct the study in a safe environment. Dr. El-Dib is also recognized as an expert in neonatal NIRS and its implications in various complications in preterm infants and those with HIE. This group has maintained a long and fruitful collaboration with Dr. Franceschini. Dr. Franceschini and her team have a strong history of experience developing and translating NIRS devices. For this project, they will be responsible for the technical development of the FlexNIRS device, process development, and data analysis. Leveraging insights gained from a recently completed 5-year NIH project targeting extremelyAtty. Dtk. No. 125141.05019. MGH2O24-394preterm infants, the team is confident in the feasibility of wearable brain monitoring for neonates. Additionally, the preliminary NIRS-PPG data from newborns have shown pulsatile morphology patterns that may mirror blood pressure changes observed in adults.
[0246] CNAP-FlexNIRS, as a noninvasive blood pressure monitor for newborns, has the potential to revolutionize neonatal care by eliminating high-risk invasive procedures and providing real-time data for prompt intervention. This technology will empower clinicians with safe and reliable vital information to personalize management for infants at risk of blood pressure fluctuations, leading to improved outcomes.
[0247] B. Innovation. The CNAP-FlexNIRS will be the first continuous non-invasive blood pressure monitor tailored specifically for neonates, leveraging wearable wireless NIRS technology. This project is innovative in several ways: Reliable noninvasive ABP estimates in newborn infants can assist clinicians in effectively managing conditions related to hypo- and hypertension. Intra-arterial blood pressure monitoring in neonates, particularly those born preterm, is essential for the diagnosis and management of hypertension, hypotension, and altered hemodynamics, all of which impact neonatal development. However, intra-arterial blood pressure monitoring is less commonly used in neonates due to its invasiveness and associated risks, and currently proposed noninvasive methods offer only suboptimal precision and safety in newborns. Leveraging the experience with developing FlexNIRS devices and processes, which have shown good correlation with true ABP in adults, we aim to develop precise CNAP monitors for neonates.
[0248] FIG. 20A-B shows (a) Mean arterial blood pressure (MAP) estimates from FlexNIRS, based on pulsation waveform feature analysis (P2 / P1) during a carotid endarterectomy (CEA) in an adult patient, showing a strong correlation with arterial-line readings (R2=.87) and (b) Accumulative results in 6 CEA patients showing (bl) the prediction vs. the true target blood pressures (R2=0.83) and (b2) corresponding Bland- Altman plot (LoA: 95% Limit of agreement; CI: 95% confidence interval).
[0249] Continuous, non-invasive blood pressure monitoring of newborn infants using forehead NIRS measurements. One of the novelties of the approach lies in the selection of the monitoring site. By focusing on the head, we leverage a region that is both critical for monitoring vital signs in neonates and underutilized in current non-invasive ABP monitoring technologies. This approach is innovative because monitoring the regional brain through theAtty. Dtk. No. 125141.05019. MGH2O24-394head using a wearable sensor serves as a stable measurement site free from common issues like motion artifacts and peripheral influences in limb-based measurements.
[0250] Use of miniaturized wearable and wireless technology is vital for successful implementation of in the NICU. The use of miniaturized wearable and wireless technology is essential for successful implementation in the NICU. Wired devices pose challenges in the NICU setting, especially as neonates require multiple forms of cardiorespiratory and neurologic monitoring. Traditional neuromonitoring probes, such as NIRS and electroencephalogram (EEG), can lose contact or signal when the baby moves or is repositioned. In contrast, wearable wireless devices offer several advantages, including ease of application and better tolerance by neonates, with improved signal quality even during movement or holding, and no risk of wire dislodgement. Moreover, the low-cost, easy wearability, and long battery life of these innovative devices make them suitable for continuous monitoring, even in low-resource settings. This capability opens up new possibilities for home monitoring and telehealth applications, enhancing access to care and improving outcomes for neonates beyond the hospital setting.
[0251] C. Approach. Preliminary Data: With respect to quantifying blood pressure, we have tested the current FlexNIRS device in a) six adult patients undergoing carotid endarterectomy (CEA) surgeries FIG. 20A-B one 32 weeks gestation preterm infant (FIG. 21). The adult CEA study demonstrates a good correlation and agreement between a broad range of arterial-line ABP values and the forehead FlexNIRS ABP estimations. The FlexNIRS was attached to the contralateral side of the CEA surgery, unaffected by the diminished pulsatile PPG due to the clamping of the carotid artery. The FlexNIRS estimation is based on the P2 / P1 ratio (top panel of FIG. 20A-B) obtained by averaging 4 pulsatile signals measured at 2.8 cm
[0252] source-detector separation. Pl represents the maximum flow during systole, followed by P2, which is a reflective wave affected by the compliance and impedance of the vascular network downstream of the artery, factors that are closely linked to changes in blood pressure. Heart rate (HR), which is also relevant for estimating ABP, is included in the machine learning model. In each subject, we use 20 min of P2 / P1 and HR data to train the model, employing a 5-fold cross-validation strategy, and use the remaining 2-3 hours of measurements to predict ABP.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0253] We have also tested the current FlexNIRS on a 32-weeks gestation preterm infant. Although we were limited to periodic ABP sphygmomanometer readings taken approximately every four hours, the P2 / P1 ratio closely mirrored ABP increases and decreases over these intermittent measurements (FIG. 21). These preliminary findings across adults and newborns, demonstrating a close correlation with gold standard blood pressure measurements, strongly indicate the potential of the device and method for CNAP in newborn infants.
[0254] In this proposal, we aim to miniaturize the FlexNIRS device (Aim 1, first 6 months of year 1), test its applicability, and demonstrate its ability to provide continuous, noninvasive blood pressure monitoring in neonates (Aim 2 and 3, months 6-24).
[0255] FIG. 21 shows that consistent with adult data, a higher P2 / P1 ratio found by FlexNIRS in preterm infants indicated elevated blood pressure, as in arm sphygmomanometer readings (purple dots).
[0256] Aim 1: Develop a CNAP -FlexNIRS device tailored specifically for neonatal extended monitoring of blood pressure, capable of collecting high-temporal-resolution, high-SNR NIRS-PPG pulsatile waveforms. The original FlexNIRS device was initially designed to monitor hemoglobin oxygenation (SO2) using a self-calibrated symmetric multi -di stance geometry (FIG. 22). However, this wider geometry, though suitable for adult head sizes, proves too large for premature infants. Additionally, the requirement for two source locations acquired in sequence, slows down acquisition to a maximum of 266 Hz, more than sufficient for measuring pulsatile features in adults but potentially inadequate for infants, whose heart rates can be more than double of adults. Since the primary focus is measuring blood pressure via PPG morphology rather than quantifying SO2, we propose removing the self-calibrating feature and the second illumination point from the FlexNIRS and adopting an in-line sourcedetector geometry (Fig. 3b). This modification will ensure that the CNAP -FlexNIRS can be securely and comfortably attached to a premature infant's forehead, maintaining good contact. These modifications are straightforward and will only require redesigning of the current FlexNIRS printed circuit board (PCB), programming of the Analog Front End (AFE) and Bluetooth microcontroller, and modifications to the data collection and visualization App.
[0257] FIG. 22A-Cshows a currently developed FlexNIRS and (b) modification plan, (c) Small FlexNIRS can be embedded into an infant cap and (d) operated wirelessly via a tablet.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0258] The CNAP-FlexNIRS devices will be based on the latest FlexNIRS generation prototype (FIG. 22). The system will include one 2-color LED (e.g., 760 nm and 850 nm; SMT760 / 850; Marubeni America Corporation), and three sensitive fast PIN photodiodes (PD) (VEMD5060X01, Vishay Semiconductors) arranged in a line at source-detector separations of 2, 2.5, and 3 cm (FIG. 22). This geometry will allow us to test optimal source-detector separation for ABP estimations in extreme premature infants and term-infants, and still provide us with multi-distances estimates of SO2 and hemoglobin concentration changes over time, while assuming scattering coefficients and initial hemoglobin concentration values, similar to what currently done with commercial cerebral oximeters. The CNAP -FlexNIRS will use the same AFE integrated circuit (IC) (AFE4900, Texas Instruments) and a 3-axis accelerometer IC (ADXL344, Analog Devices) currently used. A Bluetooth low energy 5.2 communication (BGM220S, Silicon Labs) will run the measurement routine: programming the AFE, acquiring the sampled light and motion signals, and transmitting data to the external tablet. The CNAP-FlexNIRS device will be manufactured on a flex PCB, allowing maximum flexibility to accommodate for the head curvature of the infant and maximize the optical coupling. The sensor neck and / or battery cable can be extended, enabling a minimal sensor size on infants’ forehead, and facilitating placement of the rechargeable battery away from the baby for easy replacement without disturbance. The sensor portion of the device in contact with the skin will have dimensions of 3.5 cm x 1.0 cm, the driving portion of the device (2.3 cm x 2.3 cm) can be adjusted at any angle or folded on top of the sensor part. The device will be encapsulated in a low-profde medical-grade silicone elastomer (Elkem RTV-4420) vacuum-cast case and embedded in a fabric headband or baby cap as shown in FIG. 22. Finally, a clip-in socket for the battery will be either attached to the baby clothes or positioned on the bed above the head. We will utilize magnetic pogo-pin connectors to facilitate effortless battery replacement. The estimated battery life is anticipated to be 8 hours with the utilization of the same medical-grade Li-Po battery as the current FlexNIRS but potentially longer if larger batteries are employed. The weight of the entire assembly is expected to be around 2-3 ounces (-60-80 grams).
[0259] The plan is to finalize the circuit design and manufacturing by month 3 of year 1, thus completing packaging and characterization by month 6 of year 1. Simultaneously, the current Android App will be modified to control and display the data acquired with the new geometry.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0260] Aim 2: Test the CNAP-FlexNIRS in 20 NICU patients undergoing intra-arterial blood pressure monitoring for clinical care, including infants of different gestational ages, races, ethnicities, and medical conditions. Subject recruitment and methods: We will measure a total of 20 neonates admitted to the BWH NICU who are 2: 24 weeks GA but < 44 weeks postmenstrual age (PMA) and are undergoing A-line BP monitoring for clinical reasons. Parents will consent to participate in this study using an informed consent document approved by the Mass General Brigham (MGB) Institutional Review Board. Each measurement will involve co-registering continuous PPG waveforms of CNAP-FlexNIRS with bedside vital signs, including EKG and ABP, for up to 7 days per patient. During this period, we anticipate mean ABP fluctuations of ±15 mmHg. If available, sphygmomanometer ABP measurements done at every care time will also be noted for cross validation. The CNAP-FlexNIRS sensor will be securely attached to the infants' foreheads using medical-grade tape designed for preterm use (e g., Mepitel Skin-Protective Film Dressing), checking the skin integrity every care period (4-6 hours), and adjusting as necessary. We have experience with continuous monitoring using a similar sensors in this population without any complications Exclusion criteria: Patients with contraindications to wearable NIRS placement in the head, known congenital or genetic abnormalities, skin infection or skin breakdown will be excluded from this study.
[0261] Population distributions & factors: A-line BP monitoring is common in the BWH NICU and typically involves infants born preterm. The BWH NICU admits ~42 infants per year who are less than 28 weeks GA, and of them more than 50% receive A-line BP monitoring. Currently, local clinical practice guidelines recommend A-line BP monitoring for all infants born < 26 weeks GA. In addition, A-line BP is commonly used in term infants with HIE and hemodynamic instability regardless of GA. Hence, the enrollment will be well distributed across different GA populations. While we anticipate large variations in pulsatile features across different age ranges, this should not affect the validation of the method because the main ABP analysis is based on individual calibration.
[0262] In the analysis we will document details like sex, race, skin color, and other medically significant conditions and setups that might affect cardiorespiratory functions. These might include factors such as the presence of a hemodynamically significant patent ductusAtty. Dtk. No. 125141.05019. MGH2O24-394arteriosus, the use of ventilators, cooling therapy, and any other conditions known to influence blood pressure regulation in these infants.
[0263] Aim 3: Develop a NIRS-PPG waveform analysis process to recover ABP. Utilize the collected data for training and testing deep learning models for blood pressure estimation. Assess the correlation between ABP estimated with the CNAP-FlexNIRS device and simultaneously measured intra-arterial blood pressure.
[0264] We will analyze d / dt(NIRS-PPG) waveform data at all source-detector separations and at wavelengths 760 and 850 nm to extract morphological features including Pl and P2. Preprocessing will be used to remove data with substantial motion artifacts detected by the FlexNIRS accelerometer and strong ambient light contamination measured during the device's LEDs-off phases.
[0265] Next, we will apply various settings of bandpass filtering, including a bandpass of 0.6-38 Hz, which is the frequency range ~1.5 times higher than the optimal range determined on the adult data. Noisy pulse signals will be removed by establishing minimum and maximum percentile thresholds on the raw NIRS-PPG. After the signal cleaning, we will determine the minimum pulse averaging necessary to achieve clean d / dt(NIRS-PPG) waveforms. In adults we found that 4 pulses provide sufficient signal quality and the best ABP estimates.
[0266] The P2 / P1 time series, along with other NIRS-PPG and d / dt(NIRS-PPG) features and other biologically important variables, such as heart rate, will be trained and tested using Random Forest and a mix of convolutional neural network (CNN) and recurrent neural network (RNN) featuring long / short-term memory (LSTM) which are known for their efficacy with time series data. We will train the model with different length data periods to determine the minimum requirement, and by using both the A-line and the sphygmomanometer ABP for calibration. Correlation analyses and Bland-Altman plots will be assessed using the ABP estimation in comparison to invasive A-line BP readings.
[0267] Statistical analyses and considerations. All statistics will be performed after examination of the primary variables for distribution to assure that all variables meet the assumptions for each statistical model. Additionally, we will perform each test using the primary variables alone as well as with statistical covariates when appropriate (e.g. controlling for demographic factors such as sex, gestational age, illness acuity, presence of seizures).Atty. Dtk. No. 125141.05019. MGH2O24-394
[0268] Power analysis: The objective is to achieve a limit of agreement (LoA) of -25% (-7 mmHg) between the mean ABP (MAP) estimates and A-line BP readings. Using a Bland-Altman analysis with a 95% confidence interval and an expected standard deviation of the difference between the measurements of -6.5 mmHg reported by Rao et al., we determine a required sample size of 10 patients in order to achieve the goal of -25% limit of agreement (i.e., LoAtarget= 1.96 X requiredsample size). Thus, monitoring 20 patients for up to 7 days ensures a sufficient data set and degrees for evaluating other covariates like age, sex, race and skin color.
[0269] Power analysis: Rigor and reproducibility. This proposal is dedicated to establishing rigorous and reliable measures of the CNAP-FlexNIRS system in newborn, including preterm and term infants. This study, guided by vascular physiology and supported by preliminary findings, centers on a novel, compact, wearable NIRS device for neonates, designed for continuous arterial blood pressure monitoring. A rigorous device development plan has been proposed, and a pilot clinical study has been designed, taking into account statistical considerations, the expected achievable enrollment, and patient population characteristics. Comparing the results with invasive ABP measurements provides a solid benchmark for assessing the CNAP-FlexNIRS monitoring's efficacy in newborn care.
[0270] Aim 1: a) miniaturization of the original FlexNIRS hardware and software into a compact CNAP-FlexNIRS tailored to small infants; b) increasing the sampling rate to -500 Hz to fully capture Pl and P2 features of pulsatile blood flow in neonates. Aim 2: a) streamline patient recruitment, measurement logistics, and data safety; b) make necessary adjustments to accommodate the effects of GA, birthweight, sex, and skin color on signal quality and the accuracy of ABP estimation; c) gather information on how repositioning and head motion influence the signal quality and accuracy of ABP estimation. Aim 3: a) optimize and standardize signal pre-processing routines and parameters tailored for small infants; b) beyond Pl, P2, and heart rate features, identify and assess additional covariates and signal features that can improve ABP estimation; c) adapt advanced deep learning models to improve the accuracy of the ABP process; d) achieve a limit of agreement of 7 mmHg between the CNAP-FlexNIRS ABP estimates and invasive ABP readings.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0271] During each aim, the limitations of the device will be thoroughly evaluated and addressed: Timeliness: Hardware and software modifications will take a maximum of 4-6 months, as they are minimal. Signal to Noise Ratio (SNR): The current FlexNIRS has a sufficient SNR to provide clean d / dt(NIRS-PPG) signals at 2.8 cm source-detector separation in adults at 266 Hz. The hardware modifications will allow us to increase the sampling rate to 500 Hz without compromising SNR. Although we have successfully tested running one channel at 1000 Hz, a 500 Hz sampling rate should be more than sufficient to extract pulsatile features from infants’ signals, especially since the high frequency content is not used for feature estimates. Diversity in recruitment: Given the extensive experience with brain monitoring research in the NICU and based on the projections regarding the availability of potential subjects, we do not foresee major pitfalls with recruitment among different races. One potential challenge might be the limited diversity of the subject pool due to the geographic location in the Northeastern United States. Nonetheless, based on enrollment records from the previous projects in the NICU, we have been successful in recruiting patients from minority populations. Skin integrity: Given the wearable design of the device with flex PCB sensors for other monitor devices already tested on preemies, we don’t anticipate any issue with skin integrity due to the direct contact of the optical sensors on the forehead. Nevertheless, should any skin integrity concerns arise, we plan to further optimize the device design, adding a hydrogel layer between the sensor and the skin. Consideration of sex, race, and other biological variables: Groups will be matched for sex. All analyses will be performed across and within sexes to determine any potential differences. To ensure diversity, groups will be balanced for race and ethnicity by actively enrolling minorities. Additionally, subjects with light, medium, and dark skin pigmentation, classified according to the Fitzpatrick scale, will be recruited evenly to investigate the potential impact of skin pigmentation on the ABP estimates. These analyses will compare the ability of the NIRS-PPG method to recover blood pressure in comparison to the A-line gold standard. While it is anticipated that the darker skin group may exhibit lower SNR, the method calibrates for individual subjects, leading us to expect good agreement regardless of skin tone. These aims and considerations highlight the potential value of the CNAP -FlexNIRS technique for noninvasive ABP monitoring in neonates, demonstrating the technique’s capability for continuous monitoring in this vulnerableAtty. Dtk. No. 125141.05019. MGH2O24-394population as well as assisting clinicians in creating targeted interventions for infants facing hyper-and hypo-tensive episodes leading to improved outcomes.
[0272] Aim 1: a) miniaturization of the original FlexNIRS hardware and software into a compact CNAP -FlexNIRS tailored to small infants; b) increasing the sampling rate to -500 Hz to fully capture Pl and P2 features of pulsatile blood flow in neonates. Aim 2: a) streamline patient recruitment, measurement logistics, and data safety; b) make necessary adjustments to accommodate the effects of GA, birthweight, sex, and skin color on signal quality and the accuracy of ABP estimation; c) gather information on how repositioning and head motion influence the signal quality and accuracy of ABP estimation. Aim 3: a) optimize and standardize signal pre-processing routines and parameters tailored for small infants; b) beyond Pl, P2, and heart rate features, identify and assess additional covariates and signal features that can improve ABP estimation; c) adapt advanced deep learning models to improve the accuracy of the ABP process; d) achieve a limit of agreement of 7 mmHg between the CNAP-FlexNIRS ABP estimates and invasive ABP readings.
[0273] Specific aim - Intracranial Pressure. Traumatic brain injury (TBI) is a major cause of death and disability, accounting for more than 60,000 TBI-related deaths and 225,000 hospitalizations per year in the United States, with vulnerable populations, minorities, veterans, and rural residents at higher risk for worse outcomes. This discrepancy is in part due to the difficulties of accessing appropriate TBI care for people living in rural areas. Current guidelines recommend using invasive intracranial pressure (ICP) monitoring in severe TBI patients at high risk for elevated ICP and to treat elevations in ICP to sustain cerebral blood flow (CBF) and prevent secondary brain injury. A non-invasive, low cost and portable ICP monitor could transform global access to ICP monitoring and enable faster and better triage of trauma patients. Furthermore, such a monitor would extend ICP monitoring to patients whose risk may be substantial but not enough to justify invasive monitor placement, as well as reducing overall ICU stays since ICP monitoring in less acute stages will be possible outside the ICU. Transcranial Doppler ultrasound (TCD), the leading technology studied for non-invasive ICP monitoring, requires specialized operators, is impractical for continuous monitoring and is not sufficiently accurate. In this context, diffuse correlation spectroscopy (DCS) a near-infrared diffuse optical method which directly measures an index of cerebral blood flow (CBFi), has emerged as a promising alternative to estimate ICP from the criticalAtty. Dtk. No. 125141.05019. MGH2O24-394closing pressure (CrCP, the arterial blood pressure at which brain vessels collapse and cerebral blood flow ceases) or the morphology of the pulsatile cerebral blood flow signal (pCBFi). DCS uses lightweight sensors that do not require specialized knowledge to apply. However, a significant barrier for adopting DCS for ICP measurements is the low signal to noise ratio (SNR) of currently technology, which limits source-detector separation (SDS) to 2.5 cm, offering relatively low brain sensitivity and strong scalp contamination. Pulse-gated averages of 40-60 cardiac cycles are required to extract pCBFi waveforms, compromising morphological information. High-end DCS devices while providing improved SNR, are expensive, non-portable, and not yet suitable for clinical use.
[0274] There is a critical need for a clinically useful device with better SNR, higher acquisition rates, greater brain sensitivity, lower cost, and higher portability. Supported by experimental results and Monte Carlo simulations (FIG. 24A-B and 29) we propose to implement speckle contrast optical spectroscopy (SCOS), an emerging camera-based diffuse optical method able to assess CBFi, and further improve it to measure perfusion changes and pulsatile CBF waveforms with higher sampling rate (50 Hz), superior SNR (> I00x SNR gain), increased brain sensitivity (>35% relative increase moving from 2.5 to 3.5 cm SDS FIG. 24A-B) and reduced costs (>40-50%) than with standard DCS. This will be achieved by optimizing light delivery (using a pulsed laser source to maximize photon throughput during camera integration time periods) and detection (use of heterodyne detection, optimize speckle to pixel size ratio, camera on contact with the skin). Moreover, we will synergically integrate SCOS with the recently developed wearable near-infrared spectroscopy (NIRS) device (FlexNIRS). The integration of NIRS with SCOS will provide additional important biomarkers to monitor brain health, like changes in hemoglobin concentration (AHbT) and oxygenation (SO2) and will allow simultaneous measurements of pulsatile blood volume (NIRS-PPG, photoplethysmography at 3.5 cm SDS). As recently demonstrated, the simultaneous pulsatile flow and volume measurement allows separation of arterial inflow and venous outflow, and augments pCBFi SNR and sampling rate (FIG. 28A-D). We believe improved brain sensitivity, ability to resolve pCBFi features with high SNR and minimal averaging, separation of pulsatile inflow and outflow contributions, and quantification of AHbT and CrCP, are keys for an accurate estimate of ICP. The Specific Aims are:Atty. Dtk. No. 125141.05019. MGH2O24-394
[0275] Aim 1: Develop compact SCOS-FlexNIRS systems for bilateral monitoring in TBI patients: la: Design and build a fiber-based dual-channel heterodyne SCOS system; lb: Integrate it with FlexNIRS; 1c: Design and build a direct-contact, miniaturized, low-cost version of the fiber-based SCOS-FlexNIRS system to facilitate broader user adoption. Id: Develop real-time software.
[0276] Aim 2: Optimize system operation and validate SCOS-FlexNIRS in phantoms and 90 healthy subjects: 2a: Optimize hardware and demonstrate increased performance of SCOS vs. DCS; 2b: Validate SCOS CBFi against DCS and ASL MRI. 2c: Demonstrate equivalence of contact and fiber-based SCOS-FlexNIRS devices; 2d: Evaluate the effect of skin tone on the recovered SCOS-FlexNIRS parameters.
[0277] Aim 3: Conduct a pilot study in 60 TBI patients and demonstrate initial clinical utility of the proposed system: 3a: Validate SCOS estimates of CrCP against TCD; 3b: Develop and test novel processes combining pCBFi, NIRS-PPG, AHbT, CrCP and ABP for improved ICP estimates and validate them against invasive ICP; 3c: Correlate the measured neurovascular parameters with neurological exam and imaging findings; 3d: Explore multiparametric predictive models based on these measurements to guide clinical intervention.
[0278] The successful realization of this state-of-the-art, non-invasive, portable, and comprehensive cerebral monitor will have a transformative role in neurocritical care especially in low resource and rural settings by enabling personalized treatments and new monitoring indications to improve outcomes after TBI. Further, the prototype instrument developed in this work will provide a strong basis for future clinical studies and commercialization.
[0279] Significance. Traumatic brain injury (TBI) is a major cause of death and disability. The most recent CDC reports estimate about 60,000 TBI-related deaths and 225,000 hospitalizations per year in the United States alone, and while TBI affects people of all ages, vulnerable populations, minorities, veterans, and people living in rural areas are at greater mortality and morbidity risk after the injury. This discrepancy is in part due to the difficulties of accessing appropriate TBI care for people living far from major neurotrauma centers. The standard of care to prevent secondary brain injury following severe TBI and other brain injury etiologies at risk for brain swelling, includes monitoring intracranial pressure (ICP) and promptly treating elevations in ICP to maintain adequate cerebral perfusion pressure (CPP), cerebral blood flow (CBF) and oxygen saturation (SO2). Unfortunately, ICP is measuredAtty. Dtk. No. 125141.05019. MGH2O24-394invasively by inserting a catheter connected to a pressure transducer through the skull. Because of the complexity of the procedure and the risks of intracranial hemorrhage and infection, ICP sensor placement is performed only by specialized personnel in neurological intensive care (neuro-ICU) settings and limited to the most critically ill patients. The ability to monitor ICP non-invasively would allow the extension of ICP diagnostics to patients in ambulatory and emergency settings, allowing for fast and more effective triage of trauma patients. It will also allow extension of ICP monitoring to patients whose risk may be substantial but not enough to justify the invasive procedure, and reduce overall ICU stays since ICP monitoring in less acute stages will be possible outside the ICU. Current attempts at non-invasive ICP measurement, by measuring optic nerve sheath diameter, tympanic membrane displacements, and fundoscopy, do not achieve sufficient accuracy (an accuracy of ± 5 mmHg is needed to be clinically relevant), are based on a coarse scale of ICP and are difficult to apply continuously. Other modalities, like magnetic resonance imaging (MRI) and computer tomography (CT), while very powerful, cannot be used continuously. One of the most promising non-invasive methods to assess ICP is transcranial Doppler ultrasound (TCD), which can monitor cerebral blood flow velocity (CBFV) in the middle cerebral artery (MCA), and from the pulsatile component can recover the critical closing pressure (CrCP, the pressure below which small brain vessels collapse) and the pulsatility index [(systolic flow velocity - diastolic flow velocity ) / mean flow velocity]. While both parameters have shown some correlation with ICP, CrCP also depends on the vessel wall tension (VWT, also known as vasomotor tone), CrCP=ICP + VWT, and since VWT changes as vascular smooth muscles vasodilate and vasoconstrict, ICP cannot be correctly estimated from CrCP without also measuring VWT. Similarly, the pulsatility index is affected by compliance of the cerebral arterial bed, arterial blood pressure (ABP), and heart rate. Other issues with TCD are: i) it is not clear if the MCA, as one of the largest arteries in the brain, is sensitive to global ICP changes; ii) it cannot be tolerated by patients for a long period of time due to the rigid and bulky transducer and the need for tight headgear for consistent alignment; iii) the accuracy of TCD-measured parameters strongly depends on uniform insonation, requiring measurements to be done by trained operators; iv) ultrasound cannot penetrate the skull in 15% of patients. Diffuse optical methods sensitive to the dynamics of red blood cells, like diffuse correlation spectroscopy (DCS), have emerged as viable methods to non-invasively and continuously measure ICP fromAtty. Dtk. No. 125141.05019. MGH2O24-394pulsatile cerebral flood flow (pCBF) morphology changes or from CrCP. To measure CBF, DCS quantifies the temporal fluctuations in the intensity of speckles formed by coherent light that has been scattered by moving red blood cells. These fluctuations are detected several cm away from the light injection point using single photon counting avalanche photodiodes, and the correlation diffusion equation is used to obtain a blood flow index (BFi) from the temporal intensity autocorrelation function (g2). DCS g2 permits estimation of an absolute index of blood flow, which has been extensively validated against several “gold-standard” techniques, including arterial spin-labeling (ASL) MRI, fluorescent microspheres, bolus tracking timedomain near-infrared spectroscopy (NIRS), phase-encoded velocity mapping MRI and PET. The biggest limitation of DCS is the low signal to noise ratio (SNR) of single speckle detection approaches, even when done with parallel detection, or SPAD arrays. The low SNR of current DCS technology hinders the use of DCS for adult human subject measurements at large (>2.5 cm) source-detector separations (SDS) needed to achieve sufficient brain sensitivity, and at high sampling rates (>20 Hz) needed to resolve pulsatile blood flow for CrCP and ICP estimates. Emerging DCS methods integrating across multiple speckles (interferometric / heterodyne DCS), while providing a substantial SNR improvement, need very fast and expensive cameras at the limits of current technology (>300 kFrames / s, needed to resolve decorrelation at lags of a few pm). Alternatively, as we have recently shown, by moving DCS to 1064 nm illumination and to superconducting nanowire detection (SNSPD), we can achieve a 16x improvement in SNR. But cost and bulkiness of the SNSPD detectors limit their use to very few specialized centers.
[0280] Recently, a technique derived from laser speckle contrast imaging, named speckle contrast optical spectroscopy (SCOS), has been demonstrated. In contrast with DCS, which measures the temporal speckle fluctuations, SCOS measures changes in the spatial pattern of the speckle fluctuations called speckle contrast (K). More specifically, the contrast of the measured speckle pattern, captured at a distance from the source, as opposed to over the illuminated area, as done in laser speckle contrast microscopic imaging, is calculated as K = a( / ) / ( / ), where a( / ) and ( / ) are the standard deviation and the mean of the intensity for a given integration (exposure) time (T) on the order of a several hundred micro-seconds. Dynamics in the tissue (blood flow) blur the speckle pattern, decreasing K, and thus K can be utilized to monitor cerebral blood flow as CBFi=l / (2TK2). Compared to DCS, SCOS uses relatively low-Atty. Dtk. No. 125141.05019. MGH2O24-394cost, low frame-rate (fps), metal-oxide-semiconductor (CMOS) megapixel cameras as detectors, making it feasible to average over millions of speckles / pixels. This massively parallelized detection has been recently shown to enable sufficient SNR measurements in human brain applications when detection is fiber-based or in direct contact with the skin. More recently, further SNR improvements have been demonstrated by implementing heterodyne acquisition to overcome the camera noise.
[0281] Supported by the above literature and the BU collaborators recent advances (FIG. 23A-C), new experimental results and Monte Carlo simulations (FIG. 24A-B and FIG.28A-B), we plan to implement SCOS to resolve CBFi, pCBFi, CrCP and ICP with higher SNR, higher brain sensitivity at high sampling rate than what is achievable with conventional DCS, and at the same time allowing for reduced cost, increased portability and miniaturization. The novel fiber-based SCOS prototype, shown in FIG. 24A-B, despite using a relatively inexpensive CMOS camera (a2A1920-160umPRO - Basler ace 2, <$500) is able to recover pCBFi waveforms at 3.3 cm SDS and ~46 Hz sampling rate with lower noise than with the most expensive and best performing DCS system available in the lab (FIG. 24A-B). The exceptional performances of this SCOS prototype are due to 1) The use of a pulsed laser source, which is ON only during the exposure time of the camera. For a 2 ms exposure time this allows us to deliver a lOx higher power during the camera frame while still delivering to the tissue an average optical power within ANSI safety standards limits due to the reduced duty cycle (FIG. 23A-C) Improved light coupling to the camera sensor through the use of rectangular fiber bundles and projection optics which defocus the image of the fiber array to reduce non-uniform illumination. Further, we achieve shot-noise limited performance by increasing the spatial and temporal integration of laser speckles, i.e., by decreasing the speckle to pixel size ratio (s / p) and increasing the camera integration time T to 1-2 ms. 3) We have developed a robust processing stream to correct for system noise via post processing. The data analysis pipeline we have developed to calibrate and correct for the system noise induced bias is illustrated in FIG. 23A-C. With these factors, the recent modeling work has predicted that the performance of SCOS can surpass that of DCS systems for human brain measurements in terms of contrast to noise ratio (CNR) by two orders of magnitude for a total system cost >40-50% lower. This is illustrated in FIG. 24A-B where the coefficient of variation (CoV) and CNR of DCS and the new SCOS system are compared. While we can expect another lOx improvementAtty. Dtk. No. 125141.05019. MGH2O24-394in CBF CNR when utilizing 1064nm as has been explored in DCS, the camera technology for this wavelength is in the early stages of development and is much more costly. Importantly, the approaches developed here will translate directly to 1064 nm when those cameras become accessible.
[0282] FIG. 23A-C shows a schematic of the fiber-based pulsed source SCOS system with the detection fiber bundle rectangle distributed at the camera end. MMF=multimode fiber, FB= fiber bundle b) Light intensity, K and BFi measured with a sCMOS camera at 2.5 cm SDS using CW illumination (top) and 10% duty cycle pulsed illumination (bottom), both with an average power of ~33mW. Camera T 2 ms and 46 Hz fps. The 10 times higher power of the pulsed source increase SNR lOx. c) Pictorial representation of the data analysis processing stream: The raw intensity image from the detection fiber bundle at time t is defocused to reduce the non-uniform intensity distribution over the fiber bundle ROI. An average of 500 dark images is subtracted from the raw intensity, and a region of interest (ROI) is selected. The raw speckle contrast square is calculated for each sliding window of 7x7 pixels within the ROI, to reduce the noise arising from calculated within each window. We then remove the bias introduced by shot noise K2, read noise K2, spatial non-uniformity noise K2, and digitization error K2for each 7x7 pixel window Finally, we perform a weighted average by I2for all the windows within the ROI to obtain a single value of K2(t) for the camera frame.
[0283] In this project we are planning additional improvements to the current benchtop SCOS system to make it portable and further increase its performance: i) move from the chopper wheel to pulse implementation via either a galvo or a pulsed laser driver, to optimize packaging and portability of the system; ii) implement heterodyne detection to further reduce camera read noise contribution, permitting shot-noise limited performances at shorter T to further improve depth sensitivity, and allowing use of even cheaper and more compact cameras; iii) move from fiber-based to contact SCOS to improve camera-tissue coupling and remove the comparatively heavy multimode detection fiber bundle to increase wearability and decrease system costs.
[0284] Moreover, to further increase the ability to recover pCBFi at high sampling rate and to provide additional parameters for accurate ICP estimates, we plan to add near-infrared spectroscopy (NIRS). Specifically, we will integrate fiber-based and contact SCOS with the recently developed low cost, wireless and wearable cerebral oximeter, called FlexNIRS. TheAtty. Dtk. No. 125141.05019. MGH2O24-394FlexNIRS includes 2 dual-A LEDs and 3 photodiode detectors arranged in a symmetric geometry (SDS 0.8, 2.8 and 3.3 cm), which allows for a self-calibrated multi-distance method for the recovery of hemoglobin oxygenation (SO2) and concentration (Hb) changes at a rate of 266 Hz. The device is powered by a rechargeable battery and uses Bluetooth Low Energy for wireless communication. We have implemented a custom mobile application built on the Android operating system which collects, processes, and displays data in real time. The cost of the FlexNIRS, once produced at high volume, will be ~ $100 / unit. We currently make all circuit drawings available to researchers following the open-source model, and we distribute devices via service contracts through MGH and Neuluce (non-profit). Characterization tests in phantoms and human participants show very low noise (noise equivalent power of 70 fW / VHz) and robustness of SO2 quantification in vivo thanks to the multi-distance approach and innovative quality control metrics. The high
[0285] SNR performance of the FlexNIRS allows us to resolve pulsatile blood volume at large SDS (NIRS-PPG, photoplethysmography) and its temporal derivatives (d / dt(NIRS-PPG)), which carry important information about the propagation of the pulsatile waveform in the cerebral vessels. Morphological features extracted from pulse-oximetry PPG have been long investigated in the literature and used to obtain information about the state of blood vessels and cardiovascular health, like aging, stiffness, and peripheral resistance. Pulsatility index differences, measured with NIRS in cerebral blood vessels, have also been reported in aging, cortical thinning, and coronary artery disease studies. The ability to measure both pulsatile blood flow and blood volume waveforms of cerebral blood vessels, as proposed here, opens a new dimension to studying brain health with diffuse optical methods. For instance, as shown in a recent publication, by measuring both the NIRS-PPG and pCBFi signals, we are able to separate the inflow and outflow contributions to the pulsatile flow (FIG. 28A-D) and potentially use the two components to improve ICP estimates. Moreover, by taking advantage of the higher SNR and sampling rate of the FlexNIRS we can enhance the pCBFi signal. As described in, we have developed a new method that uses NIRS-PPG signals to model pCBFi. As shown in FIG. 28A-D, the fitted pCBFi (pCBFi-fit) achieved a 2x reduction in the coefficient of variation and a 5x increase in temporal resolution with respect to the pCBFi coacquired with SNSPD-DCS. The high temporal resolution in particular is needed to identify ICP spikes and plateau waves and fundamental for the accurate extraction of pulsatile featuresAtty. Dtk. No. 125141.05019. MGH2O24-394used for TCP quantification. As a result, using this approach, we were able to recover CrCP and cerebrovascular resistance (CVRi) with a 4s temporal resolution, a substantial improvement with respect the 1 min reported in published DCS studies.
[0286] The plan is to combine SCOS with the FlexNIRS. The two devices will be integrated to work simultaneously (fast multiplexing), and the data collection and visualization will be unified. The use of a pulsed laser source and low duty cycle for SCOS, will allow for time sharing operation with the FlexNIRS, which will work in between SCOS camera frames to maintain a fast sampling rate. Performance comparisons with FDNIRS and DCS will be done both in phantoms and humans, demonstrating accurate parameters quantification.
[0287] Rigor of the Prior Research. Previous investigations by the group and others have demonstrated the clinical utility of non-invasive DCS CBFi monitoring, validated CrCP estimates against TCD, and shown that detailed morphological features in the pCBFi waveform carry information about ICP that can be extracted using machine learning (ML) methods. However, previous measurements were limited in quality due to the SNR shortcomings of existing DCS technology. Thus, data in adults was obtained at 2.5 cm separation, where the measurement is substantially driven by extracerebral perfusion, limiting signal fidelity. Further, even at 2.5 cm, averaging over a significant number of heartbeats was required, obscuring fine features needed to achieve robust ICP estimation (FIG. 27). As shown by the preliminary results, SCOS is able to achieve pCBFi measurements with higher SNR and larger SDS (FIG. 24A-B and 29), and we can further denoise the pCBFi measurement by synergically complementing DCS (or SCOS) with FlexNIRS (FIG. 28A-D). As we have previously demonstrated, using the diastolic runoff of pCBFi we can accurately estimate CrCP, and with the addition of NIRS-PPG, we are now able to separate inflow and outflow contribution to the pulsatile signal (FIG. 28A-D). These are all key factors that we believe will allow to accurately estimate ICP non-invasively with optical measures. As detailed above, an improved, more robust, cost-effective, and comprehensive, neuromonitoring method is needed to improve TBI patient management and outcomes. An optical device that integrates spectroscopy and speckle dynamics, while maintaining wearability, ease of use, and low cost is now feasible, and we have all the required expertise to build it (Aim 1). The device will be fully validated in healthy subjects (Aim 2) and initial clinical utility will be tested in TBI patients (Aim 3). Dr. Franceschini and Dr Carp teams have an extensive track record in diffuse optics devices andAtty. Dtk. No. 125141.05019. MGH2O24-394methods development and clinical translation, Dr. Boas is a leading expert in speckle dynamics, and Dr. Chung has the clinical knowledge to guide the development of the device from the application perspective and to validate it in patients. The successful realization of this state-of-the-art, non-invasive cerebral perfusion monitor will have a transformative role in patient management in neuro-critical care. Specifically, the availability of a nearly riskless, direct, and continuous measure of a patient’s ICP, CrCP, CBFi, SO2 and pulsatility will enable personalized treatments and improved outcomes after brain injury and will open the door to novel monitoring indications.
[0288] Innovation. The proposed SCOS-FlexNIRS system is a leap forward in neuromonitoring with diffuse light since it provides high SNR, high acquisition rates and operation at large separations, leading to more robust measurements of CBF changes, ICP, and CrCP in adults. The proposed device design offers a path to miniaturization and the synergy with NIRS provides additional clinically relevant parameters and further increases pCBFi SNR and sampling rate. This device will offer to the medical community a high-performance, low cost, bedside, non-invasive cerebral perfusion monitor with the potential to transform clinical practice. It will meet the monitoring needs in intensive care units whenever there is concern about a patient’s cerebral health, promote personalized care, and at the same time, because of its ease to use, reduce need for specialized personnel. While wearable SCOS devices have been recently developed none have demonstrated the performances of the prototype. First wearable, fully integrated SCOS-NIRS device providing a comprehensive array of vascular cerebral biomarkers at a competitive cost and performance. Long-separation measurements (3.5 cm) allowing superior brain sensitivity to current DCS technology. Automatic data quality control metrics, including motion artifact detection enabled by a built-in accelerometer. NIRS augmentation to provide noise-free pCBFi measurements with minimal averaging. Validation of non-invasive assessment of CBFi, ICP and CrCP in healthy subjects and neuro ICU patients. Generalized machine learning estimations of ICP derived from pCBFi waveform features, AHbT and CrCP. Additional information about changes in cerebral perfusion, oxygenation, vessel compliance to increase ICP diagnostic value and to tailor medical treatment to the individual characteristics of each patient. Real-time comprehensive bilateral monitoring without the need for users special training.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0289] Approach. Timeline: This project is divided into three aims whose timeline is detailed in Table 1. In Aim la we will build a fiber-based SCOS system and in Aim lb integrate it with generation 2 FlexNIRS. Modification to the current SCOS system will include heterodyne detection, use of galvo driven pulsed lasers, and bilateral acquisition. Because of the working prototypes and the straightforward modifications, we expect to execute this aim in just 6 months and fully characterize the fiber-based SCOS-FlexNIRS within the first year. In Aim 1c we will work on further miniaturizing the SCOS system by moving from fiber-based to contact SCOS, by mounting a CMOS camera directly into the FlexNIRS board, and by electronically pulsing the laser. We believe this configuration will further increase SNR while providing higher wearability and lower costs. At the same time, we will develop the unified control and real-time analysis and visualization software. In Aim 2 we will characterize the SCOS-FlexNIRS systems and demonstrate device operation and performances with respect to FDNIRS-DCS, SNSPD-DCS and ASL-MRI (Aim 2b) using standard phantoms and healthy volunteer tests. Concurrently, in Aim 3, we will conduct measurement in the clinic monitoring TBI patients, validate SCOS estimates of CrCP against TCD (Aim 3a), develop and test novel processes, including machine learning, combining pCBFi, NIRS-PPG, CrCP and other slow varying NIRS signals (as AHbT) for improved ICP estimates and validate them against invasive ICP (Aim 3b). Finally, in Aims 3c and 3d we will correlate the measured neurovascular parameters with neurological exam and imaging findings and explore multiparametric predictive models based on optical measurements to identify threshold for clinical intervention. By the end of the proposal, we will have demonstrated initial clinical utility of the SCOS-FlexNIRS monitoring system. The main outcome of this project will be the delivery of a comprehensive wearable, low-cost, optical neuromonitoring system providing brain perfusion and intracranial pressure measures.
[0290] Aim 1 SCOS-FlexNIRS development. Preliminary results: Following progress demonstrated by the BU collaborators we have built the fiber-based SCOS prototype (FIG. 23 A-C). With respect to the BU system, the prototype includes a low-cost Basler camera (a2A1920-160umPRO,) 12 bit, 2.3 MP, up to 160 Hz fps, pixel size 3.5 um, in place of the high-end scientific CMOS camera (Orca Fusion BT, Hamamatsu). After careful experimental comparisons, we have determined this Basler camera has equivalent SCOS performances at strongly reduced costs (<$500 vs. $20,000). A fiber bundle (2.5 mm diameter, -2000 strandsAtty. Dtk. No. 125141.05019. MGH2O24-394of 50 pm core diameter multimode fiber, 0.66 NA, rectangular shape at the camera end) is used to deliver collected light from the tissue to the camera. The light from the fiber bundle is imaged into the camera and slightly defocused to fill the pixels in the gaps between individual fibers with a more uniform light distribution. With this configuration we achieve an s / p ratio of 0.7. For the illumination we use a long coherence length free-space VHG laser at 850 nm (LD852-SEV600, Thorlabs) coupled to a multi-mode fiber (0.22 NA, 200 pm-core diameter) emitting up to 600 mW CW power, pulsed via a chopper wheel with adjustable blades (Thorlabs, MC2000B+ MC1F10A) set to a 10% duty cycle, providing 0.5-5 ms laser pulses. For any given exposure time, duty cycle and frame-rate, we adjust the power so that we do not exceed 38 mW average power (ANSI limit at 850nm for skin illumination). In FIG. 24A-B we report the pCBFi measured on the forehead of a subject with this SCOS system operating at 50Hz fps and 2 ms exposure time, analyzed following the processing stream described in FIG. 23A-C. For comparison in FIG. 24A-B, we also report two pCBFi data segments measured on the forehead of different subjects with two DCS systems at the same 50 Hz sampling rate: (middle panel) a SNSPD-DCS system operating at 1064 nm which includes 4 colocalized detectors (OpusOne, Quantum Opus, -$160,000) and dual 100 mW illumination (4X SNR improvement with respect to); (bottom panel) a conventional DCS system operating at 785nm acquired at only 2.5 cm SDS (3 Excelitas detectors, -$6,000 each, data from). The pulsatile CBFi waveforms are best recovered with the SCOS system, highlighting the enhancement of CBFi measurements through multi-speckle acquisition. As shown in FIG. 24A-B, these experimental results are in agreement with noise estimations made with Monte Carlo simulations enhanced with noise models developed by. While depth sensitivity is lower for SCOS, the improved CoV allows to operate SCOS at larger SDS that with DCS, hence improving depth sensitivity. At 3.5 cm SCOS CNR is 600 times better than with conventional DCS and 30 times better that with a 1 channel SNSPD-DCS system at 1064 nm.
[0291] FIG. 24A-B shows a comparison of pCBFi (4 heartbeats) acquired on the forehead of different subjects with the fiber-based light-pulsed Basler SCOS prototype (top), a dual illumination 4 detectors SNSPD-DCS at 1064 nm (middle), and a conventional 3 channel DCS system (bottom). In all cases light power is within ANSI standard limits, signals are processed at -50Hz, no smoothing a verages applied. Despite being at a 1cm shorterAtty. Dtk. No. 125141.05019. MGH2O24-394source-detector separation the pCBFi resolved with the conventional DCS is much noisier pCBFi with SCOS is comparable to the one acquired with the most expensive and best performing DCS system we have in the lab. b) Simulated cerebral perfusion measurement performances of the Basler fiber-based light-pulsed SCOS (T 1ms, fpslOOHz, peak power 380 mW, duty cycle 10%, s / p=0.7), vs. SNSPD-DCS at 1064 nm and conventional DCS at 850 nm, both with 1 source and 1 detector and illumination at ANSI limits. In the panels we report cerebral sensitivity (top), CoV (standard deviation of the fit results / average value of the fit results) (middle) and CNR=sensitivity / CoV (bottom), for detecting a change in CBF at 10 Hz using a 3D head model (extracerebral thickness of 1.5 cm) with typical tissue properties. Pulsed fiber-based SCOS achieves orders of magnitude lower CVR and higher CNR than DCS across all SDS. SCOS CNR benefits are achieved despite the lower brain sensitivity of SCOS.
[0292] Fiber-based SCOS implementation: The new fiber-based SCOS system (FIG. 25A-C) will be based on the current prototype, with the following new implementations: i) The chopper wheel set will be replaced by a galvanometer mirror. This design enables us to rapidly angle the laser beam into and away from launching into a 200 pm diameter multimode fiber achieving switching transition times of <0.5 ms with >70% coupling efficiency as we have achieved previously. This will allow us to use the same high-power long coherence length laser source for bilateral illumination. The illumination will be slightly longer than the exposure to prevent any intensity jitter while the laser beam is moving to or from the chosen fiber, ii) A dedicated clock network (see Sync in FIG. 25A-C) will be designed to adequately synchronize the galvanometer with the CMOS Basler detectors, iii) We will implement heterodyne interferometric detection. As demonstrated by this strategy will help overcome the camera read noise and allow for shorter integration times, boosting the measurement SNR and further increasing brain sensitivity. The heterodyne strategy will also help reduce the influence of ambient light noise, which is important in the clinical setting where room light cannot be controlled. To make the Mach-Zehnder interferometer module as compact and mechanically stable as possible, we will use a 99 / 1 fiber coupler to split the laser light in two beams, two computer-controlled variable attenuators (VA) to adjust reference arm intensity, and a custom 1-99% cube beam splitter directly coupled to the CMOS camerato interfere the reference arm with the sample arm. A second detector will be used to acquire data at short SDS (5-8 mm).Atty. Dtk. No. 125141.05019. MGH2O24-394Either a SPAD or a second CMOS camera will be used for the short separation detection. In the latter case, a small multimode fiber (up to 400um-core diameter) will be sufficient at this separation, without the need for interferometric detection. Two identical bilateral SCOS systems will be built during the first 6 months of the project to support the healthy volunteer and patient studies. Optimal SCOS operating parameters will be determined with measurements in phantoms and healthy subjects, in terms of speckle per pixel ratio, exposure time and frame-rate, and will be used in patients.
[0293] FIG. 25A-C shows a fiber-based SCOS FlexNIRS implementation a) Simplified block diagram (shown unilateral); b) Acquisition scheme for bilateral acquisition at 100Hz sampling rate; c) Schematic drawing of the set up. The SCOS fibers have pictorially rendered into an actual FlexNIRS probe. Generation II FlexNIRS and integration with SCOS: The gen II FlexNIRS system (bottom part of block diagram in FIG. 25A-C) will be based on an improved version of the original FlexNIRS. For the gen II FlexNIRS we will use a new photodiode (VEMD5160X01, Vishay semiconductors) which filters out room light below 700nm, and to avoid photon losses, we will move the short wavelength from 735nm, for which we estimate filter attenuation to be
[0294] ~3dB, to 760 nm (SMT760 / 850, Marubeni, negligible attenuation). We will continue to use the self-calibrating geometry with SDS of 30 and 35 mm (currently 28 and 33 mm). The symmetric geometry allows for compensation of for inter-device variability without need of external calibration, provides quality control metrics, and the two distances increase the accuracy in estimating cerebral SO2, and changes in hemoglobin concentration (AHb). The short SDS photodiode (8 mm) enables the acquisition of superficial scalp signals, for subtraction from the long SD channels to enhance depth sensitivity. A 3-axis accelerometer and gyroscope integrated circuit (IC) will be embedded on the probe to allow tracking patient movements and detect and reject motion artifacts in the optical data. The analog-front-end integrated circuit (AFE4900, Texas Instruments) is fully programmable and allows us to optimize the sequence and timing of 4 LEDs and 3 detectors (+ ambient). A dedicated clock network, controlled by the microcontroller, will be used to sync the device with a second FlexNIRS unit and with the bilateral fiber-based SCOS. The microcontroller will run the measurement routine, programming the AFE, acquiring the sampled light, motion and temperature signals and transmitting data to the external tablet or computer through a USB linkAtty. Dtk. No. 125141.05019. MGH2O24-394or Bluetooth low energy 5.2 communication. The system will be manufactured on flexible printed circuit boards (PCBs) as shown in FIG. 26A-C, offering maximum flexibility to accommodate for the head curvature of the patient and maximize the optical coupling, while allowing the hosting of narrow-pitch ICs. The control module will be embedded on the same PCB to reduce the device bulk, interconnections and production costs. As shown in FIG. 26A-C, 3D-printed fiber holders for the SCOS fibers (terminated with prisms for minimizing probe profile) will be incorporated in the assembly. The SCOS and FlexNIRS will operate in a fast time-multiplexed regime to avoid optical crosstalk between the two and maximize the overall sampling rate. An example of acquisition scheme for bilateral SCOS-FlexNIRS acquisition at 100Hz is shown in FIG. 26A-C. The two FlexNIRS will be turned on simultaneously (we have verified lack of crosstalk when positioned across the forehead with the sources furthest from each other) after completion of the two SCOS camera periods (Texp=l ms) (to prevent the >300mwSCOS pulsed laser from interfering with the contralateral FlexNIRS). A full FlexNIRS sampling cycle is typically done at 266 Hz, hence 5ms is plenty to achieve good SNR. Because of the very low cost we plan to build several gen II FlexNIRS. The fiber-based SCOS and FlexNIRS devices will be integrated during the first 9 months of year 1, using a modified FlexNIRS Android app and SCOS computer software to control and display data.
[0295] Contact SCOS development: To increase wearability we will test smaller monochromatic CMOS detector, like the Basler dart daA1440-220um 5mm x 3.7mm, 227 fps, 1.6 MP (same Sony IMX family), the Basler dart daA1280-54um 4.8mm x 3.6mm 54fps, 1.2MP or the Arducam OV9281, 3.6mm x 2.7mm, 120fps, IMP, suitable for SCOS measurements, to be mounted directly on the optical sensor. While we are looking for smaller camera with less pixels and lower power requirements, the fact they will be in contact with the skin means they will collect from an area 2-3 times larger than the one of the 2.5 mm diameter fiber bundle. Furthermore, a direct coupling with the tissue will help minimize the photon losses introduced by the fiber-bundle coupling optics. These factors will improve the photonharvesting capability of the detection system, and because of the increased number of speckles detected and the smaller s / p ratio achievable with contact SCOS it is possible interferometric detection will not be needed. Tests will be done to assess the optimal detection strategy. The use of a smaller number of pixels and lower frame-rate (50-100 Hz) will reduce the power consumption and simplify heat management strategies. The contact SCOS camera will beAtty. Dtk. No. 125141.05019. MGH2O24-394integrated into the FlexNIRS sensor using a two stacked-PCB assembly. The camera will be hosted above the FlexNIRS flex PCB, centered over a window over the patient’s skin. As shown in FG. 26 this is feasible even when using the large Basler dart camera daA1920. If needed, integration of heterodyne detection will be done using custom miniaturized optics following the lead of the BU colleagues on implementing contact camera for wearable microscopes. As in the fiber-based SCOS-FlexNIRS, the optical sensor will host 3D-printed fiber holders for the short separation camera and for the SCOS light source fiber. The plan to start by using the camera own driving circuit and interface, and then modify them to improve wearability. Heat management will be addressed by either adding spacers between the camera sensor and the FlexNIRS window and heath sinks over the FPGA and CPU, or by moving the FPGA and CPU away from the optical sensor in the wearable control module. A thermistor will be added in the FlexNIRS sensor in proximity of the camera, (the AFE already supports driving and reading a temperature sensor) to continuously read temperature. Once we find the optimal camera, we will attempt replacing the FPGA and CPU with a custom readout architecture to operate the CMOS detector, acquire each frame, and performing on-board data processing (e.g., computing speckle contrast). This will reduce the overall data size and ease communication requirements with the external computer. In the control module, as shown in FIG. 26A-C, we will add dedicated on-board timing circuitry (sync) to synchronize the laser and FlexNIRS with the CMOS cameras. An external module will host the SCOS laser, the laser driver electronics, the fiber-based short separation camera, the power supply, and related circuitry. For the laser we will develop a custom driving circuit to allow pulse operation. Considering design and manufacturing, we expect the contact SCOS-FlexNIRS device will be available for human testing at the beginning of year 4. The final device will be fully optimized (camera frame-rate, exposure time, laser peak power, multiplexing sequence), calibration and noise subtraction will be performed accordingly, and NEP, SNR, temperature and long-term stability will be measured. In phantoms we will also verify the robustness of the entire software ecosystem. Phantom tests will be performed repeating the standard experiments described in Aim 2.
[0296] i) There are low risks in the Aims la and c hardware and software development, as the current fiber-based SCOS and FlexNIRS prototypes have been characterized and shown to work very well. With respect to aim lb, we have built interferometric DCS systems, and weAtty. Dtk. No. 125141.05019. MGH2O24-394do not foresee complication on adding a reference arm to illuminate the free space camera. There is some risk involved with the development of the contact SCOS system (Aim Id), but successful development of this device will not impact execution of Aims 2 and 3, which we plan to execute with the fiber-based system. If the contact SCOS device will result too bulky or inferior in performances to the fiber-based system, the effort will still be an important step toward a wearable device for this and other applications, ii) in the contacts SCOS we may still consider using the galvanometer if the laser performance in the pulsed regime does not meet the requirements (e.g., low coherence output, insufficient output power), iii) We do not plan to integrate the short separation detector in the flex probe. The detection fiber for the SS camera will be:5400 um-core, much smaller and lighter than the 2.5 mm fiber bundle needed for the large separation. Still, we will evaluate if for the SS channel we can use a photodiode mounted in the probe, or if the NIRS SS channel is sufficient to regress scalp signals.
[0297] Aim 2 Human subject validations and performances optimization Preliminary results. FIG. 27 shows DCS measurements in the neuro ICU with current technology, and its limitations. This DCS system includes double illumination (76 mW total) at 850 nm which provides a factor 2 SNR increase with respect to. Still, in most patients, as the one in the figure, to obtain pCBFi waveforms at 2.5 cm SDS we need to gate average at least 40 cardiac cycles (averaging multiple g2 at the same point in the cardiac cycle). This limits cerebral sensitivity and the ability to use pCBFi morphology to estimate ICP in adults. As shown in FIG. 26A-C the SNSPD-DCS system at 1064 nm is much better at resolving pCBFi waveforms, but the current system, which includes a cryostat to reach the 2 Kelvin operational temperature of the superconducting detector, has a footprint, acoustic noise, and heat dissipation not adequate for the clinic. We regularly use the SNSPD system in the lab in healthy subjects as the gold standard best performing DCS, and in a recent study we have coupled it with the FlexNIRS to investigate the relationship between pCBFi and NIRS-PPG when collecting data at large SDS (FIG. 28A-D). With this study, taking advantage of the high SNR (NEP 70 fW / / Hz) and sampling rate (266 Hz) of the FlexNIRS we have developed a new method that uses the NIRS-PPG waveform to model pCBFi. The PPG signal, i.e., the negative changes in detected light intensity during a heartbeat, is proportional to pulsatile blood volume changes underneath the probe. As a result, the PPG first temporal derivative represents the changing rate of bloodAtty. Dtk. No. 125141.05019. MGH2O24-394volume, which is the difference between the blood inflow and the blood outflow during a heartbeat. This is also true for the NIRS signal acquired in reflection at large SD separations:
[0298] BFin— BFout=Yi^ NIRS — PPG). As suggested by, if we assume fluid drains out passively, BFoutcan be modeled to be proportional to the volume of blood within the tissue bed: BF0Ut= [3 ( IRS — PPG) + C. On the other hand, pCBFi represent the total flow through the vasculature: BFin+ BF0Ut= k ■ pCBFi with k the scaling factor converting pCBFi (cm / s2) to blood flow units (mL / (100gxmin). By combining these equations, we have derived pCBFi as a function(NIRS - PPG) + - C = Ci — (NIRS - PPG) + C2(NIRS - PPG) + C3and we are also able to separate the inflow (BFin) and outflow (BFout) contributions to the pulsatile flow:PPG)). We applied this method to the data and determined Cl, C2, and C3 by performing a non-linear fit to minimize the residuals. These parameters do need to be fit for each subject and different condition, like hyperventilation or breath-holding, underlying the need of a pCBFi waveform grand-truth, SCOS will be able to provide time varying pCBFi at high SNR. Then we will further enhance the SCOS pCBFiusing the Flex -NIRS after determining Cl, C2 and C3 on minimally cardiac gate averaged data. Using this approach, averaging only 4 cardiac cycles, across all subjects and conditions we found CoV of pCBFi-iit 2.25±0.99 times lower than the CoV of pCBFi acquired with SNSPD-DCS at 1064 nm, with a fit R2of 0.98±0.01 (FIG. 28A-D). We were able to separate the arterial inflow and venous outflow contribution to the pulsatile signal, which we believe can be used to improve the estimates of ICP. The Pl, P2 and DP cardiac peaks visible in the pCBFi-fit, d / dt (NIRS-PPG), and flow-in signals correspond to the Pl, P2, P3 peaks in the ICP cardiac wave. With the high SNR achieved we realized automatic pulsatile features extraction and in the proposal these features will be used to evaluate ICP and vascular compliance. Finally, by using pCBFi-fit the method allows us to d pABPestimate CrCP and cerebrovascular resistance(CV7?j =d^pCBF.^) changes with a temporal resolution 5 times higher than with the SNSPD-DCS 1064 nm system alone.
[0299] Phantom tests: For the hardware developments we will use homogeneous and 2-layer liquid phantoms to optimize, characterize and compare systems, an example of which isAtty. Dtk. No. 125141.05019. MGH2O24-394shown in FIG. 28A-B. We will start with the fiber-based SCOS and repeat the same tests for the contact SCOS-FlexNIRS. For the system parameters optimization, we will evaluate the SNR dependence on speckle / pixel ratio and exposure time and perform camera noise corrections as done in FIG. 23 A-C. We will perform comparisons between SCOS systems and vs. the 4 channel SNSPD-DCS available in the lab. The SNSPD-DCS system will use the same laser source as the SCOS system and 2 co-localized detector fibers (5um core) centered inside the SCOS detection bundle (FIG. 25A-C). When performing phantom experiments with the contact SCOS, the SNSPD-DCS will not be colocalized and the probes will not be immersed into the liquid but separated by a thin clear plastic layer. Phantoms comparisons with a non-co-localized hybrid frequency-domain NIRS DCS system (MetaOx, ISS Inc) will be done as was done for the original FlexNIRS. The liquid phantoms will be made of Intralipid, ink, and water, which concentrations will be varied to mimic a range of tissue optical properties, for the two-layer phantoms we will use a custom-made two-compartment tank to mimic the layered structure of the human head, and, by varying thickness of the top layer, we will assess depth sensitivity. To vary the dynamic contrast of the bottom layer we will use a magnetic stirrer at various speeds. With these tests in addition to quantifying SCOS SNR improvement and the equivalence in performances of the various FlexNIRS prototypes, we will verify correct recovery of changes in the Brownian diffusion coefficient (equivalent to BFi in liquid phantoms) with SCOS and absorption coefficients with FlexNIRS.
[0300] FIG. 26A-C shows the implementation of the contact SCOS FlexNIRS a) Simplified block diagram; b) photograph of a FlexNIRS system integrated with a Basler camera (Sony IMX392 sensor, 6.6mm x 4.2 mm, larger than what we plan to use) to help visualize final device. The SCOS source and short separation detector fibers have pictorially added, c) Preliminary results of contact SCOS, SDS 3.5 cm, fr 100Hz, T 1ms.
[0301] FIG. 27 shows an example of DCS data collected on the forehead of a TBI patient at a SDS of 2.5 cm, together with arterial blood pressure and intracranial pressure. A 6 sec average and 0.05 Hz low pass filter is applied to CBFi, ICP and ABP data. CrCP is calculated gate averaging 40 s of data (100Hz g2) as in. CBFi in this patient strongly correlate with ABP (R=0.88), indicating loss of cerebral autoregulation. The correlation between ICP and DCS CrCP is statistically significant (R=0.18), but the strong CrCP averaging makes it difficult to identify the intracranial pressure spikes. In the bottom inserts we show the zoomed-in raw ICPAtty. Dtk. No. 125141.05019. MGH2O24-394data (250 Hz) during a spike (left) and during normal TCP (right). The 3 cardiac peaks in the pressure wave are very distinct, with P2 always higher than Pl, indicating poor vascular compliance, which gets considerably worse during the pressure spikes. The DCS pCBFi shown in green is not able to resolve the 3 cardiac peaks because the large averaging needed dampen and corrupt the cardiac peaks.
[0302] FIG. 28A-D shows a) NIRS-PPG, d / dt( NIRS-PPG), b) Dt pCBFi and pCBFi-fit, and c) blood inflow and blood outflow waveforms obtained averaging 4 heartbeats (subject 10 of the lower left image). Dashed lines in panels a and b are the standard deviation over 1 minute rest, e) Coefficient of variation of pCBFi and pCBFi-fit across 10 subjects. Pl = systolic peak, P2 = tidal peak, DN = dicrotic notch, DP = diastolic peak.
[0303] FIG. 28A-B shows BFi comparison of fiber based SCOS at 3 cm and conventional DCS at 2.5 cm (average of 4 detectors) on a intralipid solution during temperature titration, a) 5 s averaging vs. time, b) 1 min averaging vs temp. Exposure time 5 ms, pa=0.1 cm'1, ps’=9.4 cm'1. DCS is much noisier than SCOS despite the shorter SDS, SCOS BFi changes agree with DCS.
[0304] Healthy subject validation studies: We plan to comprehensively test the SCOS systems in humans to ensure performances and correct quantification of CBFi and the other measured parameters. For this study we plan to enroll 90 subjects, age ranging from 18 to 75 years old, with different skin pigmentation, 30 each with light, medium and daks skin, following the Fitzpatrick classification and colorimeter measure (DSM-4, Cortex Technology), and equal number of male and female. The 90 subjects will all be tested with the fiber-based SCOS and a comparator technology. Sixty subjects will perform breathing and posture changes tasks (Hl group), and 30 subjects will undergo hypercapnia challenges (H2 group) (see FIG. 30 for subgroups subject distributions). Full descriptions of the human protocols are detailed in the human subject section. The Hl study will be done with unilateral fiber-based SCOS colocalized with SNSPD detectors (20 subjects) or integrated FlexNIRS (40 subjects), and include the following tasks: i) Torniquet manipulation will be used to decrease scalp blood flow (>50%) by compressing the superficial arteries, without changing the cerebral blood flow. By comparing changes at short and large separation, this task will allow us to determine SCOS vs DCS depth sensitivity, ii) A series of repeated brief breath holding will be used to increase blood pCO2 and thus CBF by about 20-30%. CrCP and blood pressure will also slightlyAtty. Dtk. No. 125141.05019. MGH2O24-394increase as shown in FIG. 28A-D. iii) Conversely a hyperventilation task will be used to decrease pCO2 and thus CBF, cerebral blood volume, and ICP, while increasing heart rate, iv) Changes in posture (from supine to 20° head up and to 90° head up) are expected to decrease CrCP and ICP. Blood flow is expected to mostly change in the scalp, v) Squatting for 1 minute and then suddenly standing causes a relevant drop in blood pressure and cerebral blood flow, probing dynamic cerebral autoregulation. Tasks will be repeated after swapping the SCOS and comparator optical probes on the forehead. The H2 study will be done with bilateral fiberbased SCOS (+SNSPD-DCS) without the FlexNIRS component (not MRI compatible), using custom hybrid probes with 10 m long fibers symmetrically attached to the subject’s forehead, and will include 2 graded hypercapnia cycles (up to lOmmHg end-tidal carbon dioxide (ETCO2) increase), which is expected to increase CBF 30- 40%. Ten subjects will be simultaneously measured with TCD (dual channel Doppler Box, DWL) to optimize the graded hypercapnia protocol and assess initial correlation of SCOS CBFi and TCD CBFV and compare CrCP with the two modalities in preparation for the measurements in patients. In years 3, SCOS CBFi will be validated against ASL-MRI in 20 subjects. These measurements will be done on a 3T scanner at the Martinos Center and, as done in the past, for ASL we will use a pCASL sequence. In all subjects we will continuously monitor systemic physiology, including ECG, pulse oximetry, respiration rate, blood pressure (Finapres® NOVA, not for H2b study), ETCO2 (only for the H2 study), with devices available in the lab. For all measured parameters baseline values and changes with tasks will be quantified. These experiments will allow us to address all Aim 2 sub-aims: 2a) To demonstrate increased performance of SCOS vs. DCS in vivo, we will use 60 sec of rest data in each subject FIG. 30 subjects). We will quantify SNR from the CBFi time series, after removing motion artifacts and physiological “noise” (HR, resp rate, etc.), and from the pulsatile signals at the diastolic point over the same rest period as done in, repeating calculation for the fiber-based SCOS, FlexNIRS, FDNIRS-DCS and SNSPD-DCS data. We expect the SCOS SNR to be >100 times higher than the SNR of the commercial DCS Metaox system and >10 times better than the SNR of the SNSPD-DCS at 850 nm. A paired design will be used to test whether the paired mean difference is greater than zero. The SNR comparison will be made using a one-sided paired t-test with a type-I error of 0.05. Assuming an underlying standard deviation of 3, to detect a paired mean difference of at least 1.73 with a sample size of 20 pairs (FDNIRS-DCS, smaller group), the power will beAtty. Dtk. No. 125141.05019. MGH2O24-39480%. 2b) SCOS CBFi estimates and changeswill be compared against the colocalized SNSPD-DCS, and against ASL-MRI. We will initially calculate CBFi with SCOS as CBFi=l / (2TK2), which is valid for Texp » r. If we observe some deviation with respect to DCS, we will use the more general relation, Kf — — JQ(1 — -) |^1(T) | with the gl(r) obtained from the analytic equations known for DCS. In this case we will also consider using multiple exposure times to quantify beta. A paired design will be used to test the agreement of two measurement using the Bland-Altman range of agreement method. Based on literature values we expect an -4% CBF increase per mmHg, and a standard deviation of the differences in the paired ASL vs DCS measurements of - 1%. Further, we do not expect a bias between the ASL-MRI and SCOS CBF increase metrics. Under these conditions, a Bland-Altman power analysis indicates that with confidence levels of 95% for both the limits of agreement and the confidence intervals of the limits of agreement, with 20 subjects we can detect agreement within a limit of 3.3% change with 81% power. 2c) With these measurements we will test equivalence of contact and fiber-based SCOS-FlexNIRS by comparing SNR, changes in CBFi, pCBFi waveforms, and FlexNIRS (SO2, AHbT and PPG) during Hl 2c tasks. A paired design will be used to test whether the mean SNR (primary metric) values are equivalent with equivalence limits of ± 2. The comparison will be made using two one-sided paired-difference Z-tests, with an overall type-I error rate of 0.05. Assuming a standard deviation of the paired differences of 3 and a mean difference of 0 with a sample size of 20 the power is 80%. 2d) To evaluate the effect of skin tone on the measurements, the 60 subjects enrolled on study Hl will be divided in 3 groups based on their skin pigmentation (light, medium and dark). Data from the fiber-based SCOS-FlexNIRS will be considered. We expect lower SNR in the darker skin group, but similar CBFi, AHbT and SO2 responses, since SCOS is a light scattering based technique less impacted by absorption, and FlexNIRS is based on a multi-distance approach which is insensitive to thin superficial layers. To establish equivalence in CBFi (primary metric) we will perform pairwise comparisons (n=3; light vs medium, light vs dark, and medium vs dark) and apply a 2-group equivalence test adjusting the type-I error to 0.05 / 3=0.017 to account for the multiple comparisons. For each comparison we will have 20 subjects per group. Assuming an equivalence region of ±0.8, a mean group difference of 0 and standard deviation of 0.75, then we will be able to detect one standard deviation worth of bias. Similarly, using the same dataAtty. Dtk. No. 125141.05019. MGH2O24-394we will compare performances on male vs. female (2 groups, 30 subjects per group) and young (<39), middle (40-59) and older (>60) age individuals (3 groups, 20 subjects per group).
[0305] i) If the CBFi changes at 3.5 cm are lower than with TCD and ASL-MRI, because of scalp contamination, we will apply short separation regressions and use Monte Carlo as done in previous studies with DCS at 2.5 cm SDS to improve CBFi estimates, ii) When using the SCOS pulsed laser at 850 nm for the SNSPD-DCS measurements SNR. will be lower than what achievable at 1064 nm. If the noise is too high and we cannot recover clear pCBFi waveforms with the SNSPD-DCS, we will use non-colocalized SNSPD-DCS at 1064 nm for comparisons with SCOS and repeat tasks.
[0306] Aim 3 Pilot study in 60 TBI patients and demonstration of initial clinical utility of SCOS-FlexNIRS The goal of this final aim is to demonstrate feasibility of SCOS-FlexNIRS clinical measurements and quantify accuracy of the CrCP and ICP estimates against TCD and invasive ICP. To achieve these goals together with the MGH clinical collaborators we will enroll 60 TBI neuro ICU patients hospitalized in the unit who, for clinical reasons, are undergoing intraparenchymal ICP monitoring. We will preferably enroll patients with a cranial bolt but will also consider patients with an external ventricular drain (per standard of care the drain is closed and ICP measured every 30-60 min) to ensure targeted enrollment is reached in < 4 years (years 2-5 of the project). We plan to monitor the first 50 patients with fiber-based SCOS-FlexNIRS bilaterally (Group Pl) and the last 10 subjects with a fiber-based and a contact SCOS-FlexNIRS, respectively on each side (Group P2, year 5). We will monitor patients for a minimum of 1 hour to a maximum of 24 hours per day for up to 3 days, while the invasive ICP monitor is on. The optical sensors will be positioned symmetrically in the outer left and right forehead, which, based on the recent study of cerebral perfusion sensitivity guided by MRI anatomical scans, consistently provides higher brain sensitivity. Optical probes will be kept in place with Tegaderm and other medical grade tape, repositioned every 8-12 hours, and removed as needed during patient care or imaging. In a sub-group of 30 patients, TCD research exams will be performed on each monitoring day, data will be continuously collected for 10 min while also acquiring SCOS-FlexNIRS data. Clinically monitored ICP will be recorded with a Moberg System (CNS) and co-registered with the optical data together with ABP and all other clinically available physiological parameters. In addition to ICP, we expect several patients will have an invasive Licox® probe in place for clinical monitoring ofAtty. Dtk. No. 125141.05019. MGH2O24-394intracranial oxygen tension (PbtO2). These data will be used for validation of SO2. Moreover, all relevant metadata, laboratory, imaging, and clinical assessments will be extracted from the patient’s Epic charts. For Aim 3a we will evaluate the per-patient correlation of SCOS-FlexNIRS CrCP estimates with TCD CrCP values. CrCP will be calculated as in using CBFi, pCBFi (or pCBFi fit) and ABP. With the increased SNR and sampling rate we will test and improve the process. A paired design similar to the one used for Aim 2b will be used to test the agreement of two measurement using the Bland-Altman range of agreement method. With a sample size of 30 subject pairs, a mean sample difference between CrCP estimates by TCD vs SCOS equal to 2.5 and a standard deviation of 0.8, the power to detect agreement is 90%. For Aim 3b we will develop new processes to estimate and validate SCOS-FlexNIRS ICP against invasive ICP. Several studies have demonstrated that ML processes, based on preidentified waveform features may be able to estimate ICP from pCBFi measurements. The main limitation of previous studies was that to correctly estimate ICP they had to be trained on a portion of the same patient’s data. We are confident that the substantial increase in signal quality provided by SCOS and NIRS-PPG fits will more accurately extract Pl, P2 and P3 features in pCBFi, and by imputing CrCP and an estimate of vessel wall tension based on flow in and flow out, cerebrovascular resistance, blood volume changes (AHbT), and available systemic parameters as arterial blood pressure, heart rate and arterial oxygenation, we will be able to accurately estimate ICP across patients without need of initial calibration. Three approaches will be taken across two levels of prediction: detecting a clinically meaningful ICP change (5 mmHg) and estimating the actual ICP value: i) based on pulsatile waveform feature sets, BFin, BFout, CrCP, AHbT and ABP we will train classifiers using decision trees and random forest approaches for binary classification and regression trees / forests for ICP estimation; ii) we will use a simple artificial neural network architecture with a combination of recurrent neural network (e.g. GRU) and fully connected layers, the number of which will be tuned as hyperparameters, with the output layer either a binary classifier or a quantitative predictor; iii) we will use a convolutional neural network where data segments will be used as an input to a convolutional layer followed by a sequence of convolution and max-pooling layers the extent of which we will tune, then flowing into two fully connected layers, the last providing either a binary output or a quantitative prediction. For all these approaches we will explore using from 1 to 60 seconds of data and evaluate output stability. We will use k-foldAtty. Dtk. No. 125141.05019. MGH2O24-394cross-validation to efficiently train these models and evaluate their accuracy using 40 patient’s data. Then we will use data from the remaining 20 patients to test out-of-sample prediction performance and repeating it permutating patients. From 40 patients with a few hundred ICP datapoints each we expect -10,000 total samples for training which should be sufficient for all 3 ML approaches. For Aim 3c and d to evaluate the clinical utility of the combined information obtained with SCOS and FlexNIRS, we will use the Aim 2 healthy subjects CBFi, pCBFi, CrCP, AHb and SO2 values and changes as normative, and from these establish thresholds indicating abnormal events in patients. Due to vasospasms, delayed cerebral ischemia, elevated ICP, and other clinical instabilities, we expect in average patient will experience intracranial hypertension event 2.5% of the time during the monitoring sessions. We will evaluate if abnormal events in CBFi, pCBFi, CrCP, etc. correspond to these clinical indications. Based on these data we will attempt to build a model for predicting the incidence of neurological sequelae. To validate it, we will include neurological tests done clinically during the patient’s hospital stay; EEG signals abnormalities, and CT or MRI indications as additional regressors in building the model. We will evaluate the model performance using ROC curves for either median or lowest CBFi and SO2 drops and ICP and CrCP spikes, seeking to define non-invasive values that may serve as thresholds for intervention. Characterization of these signal thresholds is needed to design future, adequately powered, clinical studies. Another important parameter we will attempt to extract from the SCOS-FlexNIRS data is cerebral autoregulation. Autoregulation indices will be computed over sliding 5-minute intervals after averaging CBFi and ABP values at a 10 second time resolution. The SCOS autoregulation index will be computed as the Pearson correlation coefficient between CBFi and MAP. With a sample size of 60 observations when performing a logistic regression of a binary response variable (a binary neurologic outcome) on a continuous normally distributed variable we will achieve a power of 80% at a 0.05 significance level to detect an odds ratio of 2.5 when the prevalence is 0.25.
[0307] : i) Based on admissions of TBI patients requiring invasive ICP (-60 per year) and the previous experience on enrolling patients at the MGH neuro ICU, we are confident we will have no problems enrolling 60 patients in a 3.5-year period, ii) It is possible that because of current electronic component shortages, we will be delayed and only able to start the clinical study later into year 2, but we do not see this as a problem because of the high volume ofAtty. Dtk. No. 125141.05019. MGH2O24-394patients at MGH. iii) In some patients because of EEG, bolt and other clinical monitors, it may be difficult to find the space for 2 SCOS-FlexNIRS sensors in the forehead. In these cases, either unilateral monitoring or other head locations will be considered, iv) We do not expect any issues arising from the optical measurements. We routinely monitor patients with devices developed in the lab.
[0308] Consideration of Sex and Other Biological Variables: Statistical analyses will include sex as a variable, and possible significant differences will be investigated. This is needed because females, in general, have thinner extracerebral layer, thus optical measurements in females have higher brain sensitivity. With this analysis we want to confirm the method works equally well in male and female. In patients, the ability to enroll non-white subjects and test the effect of skin tone on the measured parameters is limited since the Massachusetts population is predominantly white and TBI incidence is similar across races. To overcome this issue, in healthy subjects we will balance race and ethnicity by targeting enrollment of minorities, and as described in Aim 2d, we will be sufficiently powered to evaluate possible effects of skin color on the measured parameters.
[0309] Scientific Rigor: We have designed the study based on previous literature reports and supported by the preliminary results. The foundation of the study is a newly developed SCOS-FlexNIRS system able to acquire signals at large separations with high SNR and unprecedented sampling rate, which facilitates the execution of the study in adults, and its combination with NIRS-PPG to extract additional relevant parameters. A rigorous device development plan has been proposed, comprehensive healthy subject validation studies and a pilot clinical study have been designed, taking into account statistical considerations, the expected achievable enrollment, and patient population characteristics. Comparison with gold standard measures offers a well-accepted comparative predictor and a standard for evaluating advantages of SCOS-FlexNIRS monitoring over other methods. Comparison with DCS, FDNIRS, MRI, TCD, ICP, PbtO2, and other clinical data will allow us to assess accuracy of the recovered parameters. The leap forward in signal quality of combined NIRS and SCOS measurements is expected to enable effective training of ML processes for generalized ICP predictions. At the end of the project, we will have both proof of feasibility for clinical monitoring with SCOS-FlexNIRS and proof-of-concept demonstrations of non-invasiveAtty. Dtk. No. 125141.05019. MGH2O24-394perfusion and pressure monitoring to initiate larger multicenter clinical studies and make available out technology and data to interested researchers.EXAMPLE 3
[0310] FIG. 31 shows a schematic illustration of PPG to access vascular age.
[0311] FIG. 32 shows a schematic illustration of PPG to access cerebrovascular health.
[0312] FIG. 33 shows a schematic illustration of the pulsatile signal as an index to perfusion.
[0313] FIG. 34 shows a schematic illustration of PPG for non-invasive arterial pressure monitoring.
[0314] FIG. 35 shows a schematic illustration quantifying closing pressure (“CrCP”) with diffuse correlation spectroscopy (“DCS”).
[0315] FIG. 36 shows a schematic illustration illustrating the relationship between NIRS-PPGand pCBFi.
[0316] FIG. 37 shows a schematic illustration detaining CrCP changes at four second temporal resolution.
[0317] FIG. 38 shows a schematic illustration detaining measuring cytochrome c oxidase using NIRS and specifically the FlexNIRS.
[0318] FIG. 39 shows graphs detailing the glymphatic system activity during sleep using NIRS and specifically the FlexNIRS.
[0319] FIG. 40 shows graphs detailing glymphatic system oscillations during sleep using NIRS and specifically the FlexNIRS.
[0320] FIG. 41 shows a schematic illustration of a system for FlexNIRS validation.
[0321] FIG. 42 shows a schematic illustration of a combined SNSPD-DCS and FlexNIRS system.
[0322] FIG. 43 shows a graph detailing the relationship between NIRS PPG and pCBFi.
[0323] FIG. 44 shows graphs detailing the enhancement of pCBFi with NIRS-PPG (e.g., from FlexNIRS).
[0324] FIG. 45 shows graphs detailing FlexNIRS data during sleep.
[0325] FIG. 46 shows a schematic illustration of the NIRS multi -di stance self-calibrating method.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0326] FIG. 47 shows another schematic illustration of the NIRS multi -di stance selfcalibrating method.
[0327] EXAMPLE 4
[0328] A MW-FlexNIRS, a wearable, low-cost, LED-based, multi-wavelength nearinfrared spectroscopy (NIRS) system is presented, which is designed for continuous monitoring of cerebral oxygenation and metabolic dynamics in neonates. The device extends the original FlexNIRS platform by integrating custom eight-wavelength LED sources, a new analog front-end, and Bluetooth low energy 5 communication onto a single flexible printed circuit board encapsulated in medical-grade silicone. To support quantitative analysis with broadband LED sources, we introduce a wavelength-weighted fitting (WWF) algorithm that explicitly accounts for LED emission spectra, photodiode responsivity, and tissue optical properties. System performance was characterized through spectral calibration, stability testing, and noise equivalent power measurements, yielding an average NEP of 123 ± 25 fW / √Hz. Validation experiments using solid silicone phantoms demonstrated accurate recovery of effective attenuation coefficients with a mean error of 2.1% using a calibrated multi-distance approach. Liquid phantom studies incorporating blood, Intralipid, and yeast were used to evaluate recovery of hemoglobin and to investigate cytochrome c oxidase (CCO) related spectral sensitivity, parameter coupling, and model limitations. Finally, a pilot measurement in a term infant with transient respiratory distress demonstrates the preliminary feasibility of stable, continuous cerebral oxygenation monitoring in the neonatal intensive care unit. Together, these results support MW-FlexNIRS as a promising platform for multiwavelength wearable NIRS in neonates and highlight both the potential and current challenges of oxCCO quantification.
[0329] Hypoxic Ischemic Encephalopathy (HIE) is one of the leading causes of mortality in term infants in the developed world with an incidence rate of 1-2 per 1000 live births. While various factors contribute to HIE, the primary pathophysiology involves a decrease in the infant’s cardiac output, leading to reduced cerebral perfusion and oxygen supply. In mild to moderate cases this leads to anaerobic metabolism and mitochondrial dysfunction, and in severe cases, it causes irreversible cell necrosis. Survivors of HIE are at high risk for long-termAtty. Dtk. No. 125141.05019. MGH2O24-394neurological disabilities, including cerebral palsy, cognitive impairment, epilepsy, and motor deficits.
[0330] Therapeutic hypothermia (TH) is the standard treatment for moderate or severe HIE, where the infant’s body temperature is lowered 2-4 °C to reduce the metabolic demands and minimize the risk of reperfusion brain injury. However, TH is applied uniformly, despite considerable variability in the timing, severity, and metabolic response to injury. Personalized therapy, guided by real-time physiological monitoring, could better tailor the intensity and duration of TH to the individual infant, potentially reducing secondary injury and improving neurodevel opmental outcomes.
[0331] Magnetic resonance imaging (MRI) to assess brain injury after HIE is typically performed 72 hours after treatment with therapeutic hypothermia, limiting its value for early, tailored interventions. While MRI and magnetic resonance spectroscopy (MRS) provide key metabolic insight by measuring brain metabolites (e.g., lactate, N-acetyl-aspartate, glutamate), their clinical utility is constrained by delayed acquisition, logistical challenges. Continuous, real-time monitoring of cerebral metabolism is needed to overcome these limitations. Such capability could guide dynamic optimization of hypothermia therapy, support early interventions, and ultimately improve outcomes for neonates with HIE.
[0332] Optical imaging modalities such as near infrared spectroscopy (NIRS) and diffuse correlation spectroscopy (DCS) are established neuromonitoring tools, which have been extensively vali-dated and are increasingly used to assess cerebrovascular health in both research and clinical settings. NIRS provides continuous, noninvasive monitoring of changes in hemoglobin concentration, blood oxygen saturation (SO2), as well as functional activation, whereas DCS enables measurements of changes in cerebral blood flow. Standalone NIRS devices have been used to assess metabolism via changes in the oxy- (HbO) and deoxyhemoglobin (HbR), as well as fractional tissue oxygen extraction; however, these measurements are a proxy for oxygen utilization and can be affected by factors such as blood flow, blood pressure, and hematocrit. Because of this, changes in HbO and HbR alone cannot disentangle contributions from oxygen supply and utilization. To obtain more robust measures of oxygen consumption hybrid NIRS-DCS systems have been used to estimate the cerebral metabolic rate of oxygen (CMRO2), a key indicator of cerebral metabolism, in both adults and neonates. While CMRO2 is a valuable metric, its noninvasive estimation is indirect and reliesAtty. Dtk. No. 125141.05019. MGH2O24-394on physiological modeling, typically using Fick’s principle. Current implementations often depend on fiber-based NIRS-DCS systems that, while effective, can be costly and cumbersome, limiting feasibility for continuous, long-term bedside monitoring in sensitive populations such as neonates with HIE. Furthermore, because CMRO2 NIRS-DCS estimation relies on hemoglobin signals, it may be influenced by extracerebral tissues, reducing specificity to brain metabolism. These limitations motivate targeting oxidized cytochrome c oxidase (oxCCO) as a more direct, brain-specific indicator of mitochondrial oxygen utilization, beyond what can be inferred from hemoglobin-based measurements alone.
[0333] Cytochrome c oxidase (CCO) is a key enzyme in the electron transport chain inside mitochondria (Complex IV) and plays a central role in the production of adenosine triphosphate (ATP) during aerobic glucose metabolism. It is responsible for the final step in cellular respiration by transferring electrons from cytochrome c (Cyt c) to molecular oxygen, reducing it to water. As a mitochondria-bound signal, CCO serves as a direct biomarker of cellular metabolism. It is relatively abundant in the brain compared to extracerebral tissue, making it a more brain-specific indicator of metabolic activity. CCO contains multiple metal centers, including copper sites essential for efficient electron transfer and oxygen reduction. Among these, the CUA site is of particular interest due to its strong, broad absorption peak in the near-infrared range, centered around 830 nm, when oxidized. The contribution of CuA to NIR absorption changes depending on redox state of CCO, oxidized (oxCCO) or reduced (redCCO). By exploiting this property, NIRS can provide real-time, non-invasive monitoring of CCO redox dynamics, offering valuable insights into cerebral metabolism under both physiological and pathological conditions.
[0334] Clinically, oxCCO becomes particularly informative in situations where cerebral oxygen deliv-ery appears adequate by NIRS (stable SO2, HbO / HbR) but mitochondrial oxidative metabolism is impaired — for example, in infants with HIE during and after therapeutic hypothermia, in sepsis, or under sedative / antiepileptic drugs with mitochondrial toxicity. In these cases, hemoglobin-based measures can show preserved or even increased oxygenation due to reduced extraction, whereas a fall in oxCCO would indicate impaired oxygen utilization at the mitochondrial level. Conversely, during transient desaturations or hemodynamic instability, stable oxCCO could help distinguish benign fluctuations in oxygen delivery from episodes that materially compromise oxidative metabolism. In the context ofAtty. Dtk. No. 125141.05019. MGH2O24-394respiratory distress and HIE, this added specificity may improve risk stratification by identifying infants whose mitochondrial recovery lags behind normalization of perfusion and SO2.
[0335] To translate this potential into practice, however, instruments must reliably separate the relatively small oxCCO signal from the dominant hemoglobin background. While the extinction coefficient of CCO is significantly higher than that of hemoglobin, its relatively low concentration in tissue contributes only 10% of total absorption in the NIR range. Recent work by Arifler et al. demonstrated that changes in hemoglobin and oxCCO concentrations could be recovered in newborn piglets undergoing transient hypoxic ischemia using a broadband NIRS system with 1 nm spectral resolution. Using a genetic algorithm for wavelength selection, they showed that an 8-wavelength discrete system, uniformly spanning the near-infrared spectrum, could estimate oxCCO changes with less than < 2% error relative to the broadband reference. Since the first demonstration of CCO measurements by Jobsis in 197, interest in monitoring this biomarker has grown significantly within the NIRS community. Over the past two decades, advances in optical spectroscopy, computational modeling, and hardware miniaturization have further enhanced the feasibility of CCO measurement in both research and clinical settings. As a direct indicator of mitochondrial function, CCO measurement holds significant promise for improving our understanding of cerebral metabolism in health and disease.
[0336] The group previously developed the FlexNIRS, a low-cost, wireless, wearable cerebral oximeter, which has demonstrated ease to use, reliability, and low-noise performance. Since its introduction, the device has been applied in various settings, including enhancing DCS pulsatile blood flow measurements, estimating blood pressure, and assessing muscle oxygen metabolism, to name a few. The system has also undergone continuous improvements in both performance and form factor. Building on these results, and leveraging recent advancements in electronics and LED packaging, we have built a new eight-wavelength NIRS system specifically designed for non-invasive monitoring of hemoglobin and CCO changes in infants at risk of HIE. In this work, we report the system design, along with experimental characterization on both solid and liquid phantoms, and preliminary in-vivo data from a neonate with transient respiratory distress. Altogether, our system represents a significant step toward integrating multi-wavelength (MW) optical spectroscopy into a compact, wireless andAtty. Dtk. No. 125141.05019. MGH2O24-394wearable platform, the MW-FlexNIRS, suitable for continuous metabolic monitoring in neonatal intensive care.
[0337] The MW-FlexNIRS device builds on the previous generation of the FlexNIRS, maintaining a similar architecture with several key improvements. A diagram of the system is shown in Fig. 1. To achieve near-uniform coverage across the near-infrared spectrum, necessary for recovery of changes in oxCCO concentration, we used two custom 8-wavelength (8k) light emitting diode (LED) packages (Marubeni America Corporation) with emission peaks at 735, 760, 780, 810, 830, 850, 870, and 910 nm. The LED package has a small footprint of 2.9 × 2.2 mm and low profile of 0.9 mm. For detection we used 4 fast PIN photodiodes (PDs) (VEMD5060X01, Vishay Semiconductors) with a responsivity > 57% in the 700-950 nm range. To expand both the number of multiplexed sources and detectors sampled, we adopted a new analog front-end (AFE 4460, Texas Instruments). This module can drive up to 16 individual sources at its standard maximum current setting (250 mA) and simultaneously sample up to four PDs in parallel using four internal transimpedance amplifiers (TIA).
[0338] FIG. 48 shows a block diagram for the MW -Flex NIRS device.
[0339] In addition to the optical components, the system integrates a 3-axis accelerometer (ADXL344, Analog Devices Inc), which we use for motion detection and data quality purposes. The MW-FlexNIRS device additionally features a newer micro controller unit (MCU) (BGM220SC22HNA, Silicon Labs) that supports Bluetooth Low Energy (BLE) 5 communication allowing us to utilize significantly higher data transfer rates, relative to the older generation of the FlexNIRS. The device is powered by a single 3.7 V, 600 mAh lithium-ion battery (LP503040JH, Jauch Quartz; 42.0 × 30.5 × 5.4 mm; 14 g). At the maximum optical power setting (all LEDs active at 250 mA drive current, 40 Hz sampling, continuous BLE streaming), the average current consumption during the 10-hour stability test was approximately 40 mA, corresponding to an average electrical power of 0.15 W.
[0340] The new AFE enables sequential acquisition of up to 16 active LED phases-8 per side of the probe- where a phase is defined as a period during which all four PDs are read in parallel while no more than one LED is active. In addition to the 16 active phases, the AFE performs automatic ambient light measurements, resulting in a total of 17 acquisition phases. Each phase has an integration time of 117 μs. Once an acquisition cycle is complete, the MCU reads out the data from the AFE and the onboard accelerometer and transmits it to a dedicatedAtty. Dtk. No. 125141.05019. MGH2O24-394Android application for real-time display and storage. The MW-FlexNIRS operates at sampling frequency of 40 Hz.
[0341] Unlike the previous iteration of the device, which consisted of two separate rigid printed circuit boards (PCBs), one for the AFE and one for the BLE MCU, along with an optical probe mounted on a flexible PCB the MW-FlexNIRS integrates all components onto a single flexible PCB, shown in Fig. 2(a). This design reduces the overall size of the device and minimizes mechanical and electrical failure points due to wear and detachment. To further improve durability and ensure user-friendly attachment, we incorporated a magnetic pogo-pin battery connector, which minimizes the risk of mechanical breaks during setup. The board is coated with an acrylic conformal coating to reduce the impact of moisture and sweat and housed within a medical-grade silicone elastomer case (Elkem RTV-4420, Factor II, Inc.). To block ambient light from reaching the detectors, the silicone elastomer was rendered black using a medical-grade dye (FI-205, Factor II, Inc.).
[0342] The fully assembled system, shown in Fig. 2(b) has a footprint 8.2 × 3.2 cm², a height of 0.6 cm, and a weight of 9.4 g (23 g including the battery). The device is placed inside a headband, as shown in Fig. 2(c), to ensure stable contact with the subject’s skin without applying pressure, which could cause discomfort during prolonged use, especially in infants. The new acquisition sequence with more wavelengths and detectors, along with the integration of a new AFE module, required the development of a revised data structure, and the design of a new app to seamlessly operate the device. The app was built in Android Studio for the Android operating system and includes several new features to enhance user experience compared to the previous version.
[0343] To improve the spectral accuracy of this LED-based device, we characterized both the LED sources and PD detectors under controlled conditions. LED characterization was performed by mounting each LED on a prototype board, which was placed on a thermoelectric cooler (TEC) to maintain a constant temperature of 38°C, approximating real-use conditions. An additional set of measurements was performed at 19°C to support phantom validation. Light from the LEDs was coupled into a 62.5 μm optical fiber and delivered to a miniature spectrometer (STS-NIRS, Ocean Insight). Emission spectra were recorded across a range of drive currents up to the maximum of 250 mA. Each spectrum was then normalized to its peak intensity to allow comparison across diodes.Atty. Dtk. No. 125141.05019. MGH2O24-394
[0344] Photodiode characterization was carried out by mounting a single PD on a prototype PCB and placing it on the TEC. To evaluate the temperature dependence of the detector responsivity, we used three lasers at 785, 808, and 853 nm (iBeam Smart, TOPTICA Inc.) and measured the detected intensity over a temperature range of 20 - 38°C. The measured temperature dependence was then combined with manufacturer-reported responsivity data at 835, 865, and 940 nm. We interpolated these data to estimate the detector responsivity at 38°C across all relevant wavelengths.
[0345] To assess system stability during prolonged operation, effects of self-heating and battery voltage on LED intensity were investigated. The MW-FlexNIRS device was placed on a solid phantom with known optical properties and powered using a fully charged battery. Light intensity data were continuously acquired until the battery was fully depleted. In addition to light intensity, battery voltage, LED temperature, and ambient temperature at a sampling rate of 1 Hz were monitored. Temperature was measured using a fiber optic temperature sensor (Reflex, Neoptix) with one probe positioned directly on the backside of the flexible PCB, immediately beneath the 8-wavelength LED package, and a second probe placed near the phantom to record room temperature. To monitor battery voltage, a modified lithium-ion battery with a secondary BNC-connectorized output was measured by a digital acquisition device (USB-6003, National Instruments). All LEDs were driven at the maximum current of 250 mA. Prior to turning on the LEDs, the system was allowed to reach equilibrium with room temperature.
[0346] To evaluate the sensitivity of the MW-FlexNIRS system, the noise equivalent power (NEP) was measured. Light from each LED was coupled into a fiber bundle and passed through two sets of neutral density filters mounted on a filter wheel. All LEDs were driven at their maximum current (250 mA), and PD gain was set to its maximum to replicate real-use conditions. To estimate the optical power incident on the PD for each LED / filter combination, the wavelength-dependent attenuation of the filters using a power meter (843 -R power with 818-SL detector, Newport Corporation) was characterized. From these measurements, an incident power matrix was generated based on the known filter combinations and the unattenuated optical power for each LEDs. For each attenuation level, light intensity was measured using MW-FlexNIRS for 10 seconds, and the signal was averaged in 500 ms intervals. NEP was determined by plotting the detector signal as a function of incident power and identifying theAtty. Dtk. No. 125141.05019. MGH2O24-394point where the extrapolated sensitivity curve intersected the noise floor, measured with the LED off. The noise floor was defined as the standard deviation of the dark measurement. The intersection of the two lines represents the minimum detectable optical power above the system’s intrinsic noise.
[0347] Accurate quantification of chromophore concentrations in LED-based NIRS systems requires careful consideration of both the broad spectral emission profiles of LEDs and the wavelength-dependent responsivity of photodiodes (PDs). Unlike laser-based systems with narrowband sources, LED-PD systems, such as the MW-FlexNIRS, must account for spectral overlap and signal distortion across a wide wavelength range (700-950 nm). To address this, we developed a novel wavelength-weighted fitting (WWF) algorithm which includes the measured spectral output of each LED and the PD responsivity profile into the optical model for both single-channel and multi-distance measurements. This approach enables more accurate mapping between changes in optical density and underlying chromophore dynamics by correcting for spectral weighting at each wavelength. For multi-detector measurements, we can estimate the effective attenuation coefficient ( / / ) of tissue, which accounts for light absorption and scattering (jiaand psrespectively), using the model of continuous wave light diffusion and light collected in a calibrated multi-distance measurement as follows:3.a( ')(jis'( ') + ia(A)), P isthe source-detector separation for a given LED-PD pair, I is the measured intensity incident on the PD, and pais the optical absorption coefficient. For an LED-PD system the measured / will be a weighted average across the LED emission profile. To model the light intensity of our system, we combine the LED emission profile (I EI). ), photodiode responsivity (RPD 2, and the attenuation of light inside tissue (A, p)) in the fitting approach for a given set of tissue chromophores. A (,p) is modeled using a solution to the light diffusion equation for a semi-infinite geometry as follows: i4(A,p) = 3(^W+paW)p) = / p2+ ( / Q2r2(A, p) = V when the refractive index n= 1.37, and the K term is slightly modified to account for the simplified case where r = 0, such that A (A) = j 3a(.s'(X) + pa(A)). Optical absorption is defined as the sum contribution of chromophores pa(A) = q(l) • Q where a A is the extinction coefficient and G is theAtty. Dtk. No. 125141.05019. MGH2O24-394concentration for a given chromophore i and the reduced scattering coefficient was modeled using the empirical powerbFrom here we can compute the estimatedintensity incident (4 / 1^° ( ))on aPD given a single LED using ILEDp) = f^maxILED(A) ■ RPD( ~) ■ A(, p) dA and express the effective attenuation coefficient as In [p2■ ILEDp)'] = —p.eff • p + ILED, O(P=0, A). We then minimize the loss function, which calculates the difference between the modeled effective attenuation coefficient and the measured attenuation coefficient obtained using calibrated intensities, using a least-squares non-linear fitting the index of each LED diodeand N is the total number of LEDs. For measurements where calibration is not possible or the use of a multi-distance approach is prevented, we designed an alternative algorithm for a single-channel AOD approach that models changes in optical density (\O / )'' °) as the LED lightintensity measured at some time t, / foEDis the light intensity measured at baseline, and DPF is the wavelength dependent differential pathlength factor expressed as: DPF(A,p) =with reference measurements or reasonable estimates ofbaseline optical properties, we can estimate the detected intensity using the model in above (e g., 0DLEDequation) and estimate changes in chromophore concentrations.
[0348] This WWF method differs from standard approaches in literature, which often rely on peak wavelengths or centroid approximations, potentially introducing bias when using broad-spectrum LEDs. Comparative performance between our WWF approach and conventional discrete peak-based methods is reported in the result section.
[0349] Two sets of phantom validation experiments were conducted to evaluate key performance metrics of the MW-FlexNIRS system. First, solid silicone tissue-like phantoms with known optical properties were used to assess the accuracy of effective attenuation coefficient u<) recovery using the multi-distance approach. Second, blood-Intralipid-yeast liquid phantoms were used to test the system’s ability to resolve dynamic changes inAtty. Dtk. No. 125141.05019. MGH2O24-394chromophore concentrations under physiologically relevant conditions using the singlechannel AOD approach.
[0350] To evaluate the accuracy of,eff recovery with the MW-FlexNIRS device, measurements were performed on four silicone phantoms with known optical properties. One phantom was used to calibrate the device, while the remaining three were reserved for validation. For the low attenuation phantom, neutral density filter films (Wratten 2, Kodak) were used to prevent detector saturation. Since the manufacturer provided optical property values only at the wavelengths of the reference FD-NIRS system, absorption was interpolated and scattering values to the MW-FlexNIRS wavelengths using a linear and power-law approach, respectively.
[0351] fieff was estimated using two methods: (1) a per-source multi-distance WWF approach, applied independently to each source (SI and S2) using detectors DI, D2, and D4 (at 3.3, 2.8 and 2.4 cm source-detector separation); and (2) a self-calibrating, cross-symmetric source-detector geometry using SI, S2 and detectors DI and D2. In the latter case, centroid wavelengths derived from the combined LED emission spectra and PD responsivity were used due to the method relying on discrete wavelengths. Since the WWF algorithm calculates the weighted effective attenuation coefficient across the entire emission range for a given LED, recovered values were compared to the / / interpolated from manufacturer-reported values at the centroid wavelength of each LED.
[0352] To validate the MW-FlexNIRS system’s ability to recover changes in HbO...
Claims
Atty. Dtk. No. 125141.05019. MGH2O24-394CLAIMSWhat is claimed is:
1. A near infrared spectroscopy (NIRS) device, the NIRS device comprising:a substrate including a body and an arm extending from a side of the body, the arm being bendable relative to the body;a light source coupled to the arm; anda light detector coupled to the arm.
2. The NIRS device of claim 1, wherein the arm includes a base, a first finger extending from the base, and a second finger extending from the base and further comprising:a first group of light sources, the first group of light sources including the light source; anda second group of light sources; andwherein the first group of light sources is substantially similar to the second group of light sources;3. The NIRS device of claim 2, wherein the first group of light sources have the same number and type of light sources as the second group of light sources.
4. The NIRS device of claim 3, wherein the first group of light sources include two different colored light sources; andwherein the second group of light sources include two different colored light sources.
5. The NIRS device of claim 4, wherein the first group of light sources includes four different colored light sources, eight different colored light sources, or nine different colored light sources.Atty. Dtk. No. 125141.05019. MGH2O24-3946. The NIRS device of claim 5, wherein the first group of light sources includes nine different colored light sources; andwherein the second group of light sources includes nine different colored light sources.
7. The NIRS device of claim 5, wherein at least one light source of the first group of light sources is an about 735 nm light source, an about 765 nm light source, an about 785 nm light source, an about 805 nm light source, an about 830 nm light source, an about 855 nm light source, an about 870 nm light source, or an about 910 nm light source.
8. The NIRS device of claim 2, wherein the arm includes a third finger extending from the base of the substrate and a fourth finger extending from the base of the substrate and the light detector is a first light detector;further comprising a second light detector;wherein the first light detector is coupled to the third finger and the second light detector is coupled to the fourth finger;wherein the first finger is substantially parallel to the third finger;wherein second finger is substantially parallel to the fourth finger.
9. The NIRS device of claim 1, further comprising a group of light sources coupled to the arm of the substrate at a distal end thereof, the group of light sources including the light source.
10. The NIRS device of claim 9, wherein the light detector is a first light detector and further comprising a second light detector coupled to the arm;wherein the first light detector, the second light detector, and the group of light sources are aligned with each other.
11. The NIRS device of claim 10, further comprising a computing device coupled to substrate at the base.Atty. Dtk. No. 125141.05019. MGH2O24-39412. The NIRS device of claim 10, wherein the substrate is a printed circuit board (PCB); andwherein the PCB is flexible.
13. A near infrared spectroscopy (NIRS) device, the NIRS device comprising: one or more light sources;one or more light detectors; anda computing device in communication with the one or more light sources and the one or more light detectors, the computing device being configured to:cause the one or more light sources to emit light towards tissue of a subject; receive, from the one or more light detectors, optical data in response to the emitted light;extract a photoplethysmogram (PPG) waveform from the optical data; receive a derivative PPG waveform as the time derivative of the PPG waveform; anddetermine an intraarterial blood pressure measurement, based on the derivative PPG waveform.
14. The NIRS device of claim 13, wherein the computing device is further configured to at least one of:present the intraarterial blood pressure measurement on a display; ornotify a user of the intraarterial blood pressure measurement of the subject.
15. The NIRS device of claim 13, wherein the computing device is further configured to:provide the derivative PPG waveform to a machine learning model trained to determine an intraarterial blood pressure measurement from derivative PPG waveforms; and receive, from the machine learning model, the intraarterial blood pressuremeasurement.Atty. Dtk. No. 125141.05019. MGH2O24-39416. A near infrared spectroscopy (NIRS) device, the NIRS device comprising: one or more light sources;one or more light detectors; anda computing device in communication with the one or more light sources and the one or more light detectors, the computing device being configured to:cause the one or more light sources to emit light towards tissue of a subject; receive, from the one or more light detectors, optical data in response to the emitted light;extract a photoplethysmogram (PPG) waveform from the optical data; determine that the user has a cognitive impairment, based on the PPG waveform.
17. The NIRS device of claim 16, wherein the computing device is further configured to:determine a first time to a systolic peak for the PPG waveform;determine a second time to a diastolic peak for the PPG waveform; and determine that the subject has the cognitive impairment, based on at least one of: the first time being above a first time threshold; orthe second time being above a second time threshold.
18. A near infrared spectroscopy (NIRS) device, the NIRS device comprising: one or more light sources;one or more light detectors; anda computing device in communication with the one or more light sources and the one or more light detectors, the computing device being configured to:cause the one or more light sources to emit light towards tissue of a subject; receive, from the one or more light detectors, optical data in response to the emitted light;extract a photoplethysmogram (PPG) waveform from the optical data; determine that the user has ischemia or an occluded blood vessel, based on the PPG waveform.Atty. Dtk. No. 125141.05019. MGH2O24-39419. The NIRS device of claim 18, wherein the computing device is further configured to determine that the user has ischemia or an occluded blood vessel, based on a second amplitude of the PPG waveform at a second time being less than a first amplitude of PPG waveform at a fist time; andwherein the second time occurs after the first time.
20. A system comprising:a pulsed light source;an image sensor;one or more near infrared spectroscopy (“NIRS”) light sources;one or more NIRS light detectors; anda computing device in communication with the pulsed light source, the image sensor, the one or more NIRS light sources and the one or more NIRS light detectors, the computing device being configured to:cause the one or more light sources to emit light towards tissue of subject; receive, from the one or more light detectors, optical data in response to the emitted light;extract a photoplethysmogram (PPG) waveform from the optical data; emit pulsed light from the pulsed light source towards the tissue of the subject;receive, from the image sensor, a plurality of speckle image in response to the pulsed light;determine, from the plurality of speckle images, a blood flow waveform; modify, suing the PPG waveform, the blood flow waveform to generate a modified blood flow waveform; anddetermine an internal pressure, based on the modified blood flow waveform.
21. The system of claim 20, wherein the blood flow waveform is a cerebral blood flow waveform and the internal pressure is an intracranial pressure.Atty. Dtk. No. 125141.05019. MGH2O24-39422. The system of claim 21, wherein the computing device is further configured to determine a presence of a brain injury, based on the intracranial pressure.
