A system and method for measuring intracranial pressure using optical interferometry

Abstract

A method and system for estimating intracranial pressure using a regression model. Optical interferometric measurements are generated that relate to cerebral blood flow. Intracranial pressure is estimated using the regression model, based on the optical interferometric measurements and one features of extracerebral blood flow or extracerebral blood pressure.

Description

A System and Method for Measuring Intracranial Pressure Using Optical Interferometry Technical Field

[0001] The present disclosure relates to a system and method for measuring intracranial pressure using optical interferometry. Background

[0002] Various issues may exist with conventional solutions for systems and methods for measuring intracranial pressure. In this regard, conventional systems and methods for measuring intracranial pressure may be costly, cumbersome, and / or inefficient. Brief Summary of the Disclosure

[0003] Shown in and / or described in connection with at least one of the figures, and set forth more completely in the claims are a system and method for measuring intracranial pressure using optical interferometry.

[0004] These and other advantages, aspects and novel features of the present disclosure, as well as details of illustrated embodiments thereof, will be more fully understood from the following description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] The various features and advantages of the present disclosure may be more readily understood with reference to the following detailed description taken in conjunction with the accompanying drawings, wherein like reference numerals designate like structural elements. FIG.1 illustrates an interferometric measurement system to measure intracranial pressure (ICP). FIG.2 shows exemplary cerebral blood flow as a function of time.FIG.3a shows three exemplary processes to estimate ICP. FIG.3b shows an exemplary CBFi vs. ABP plot. FIG.4a shows an exemplary measurement setup to obtain extracerebral blood flow measurements. FIG.4b shows a histogram of received photons as a function of time-of-flight. FIG.4c shows photon time of flight for two different source-detector separations. Detailed Description

[0006] The following discussion provides various examples. Such examples are non-limiting, and the scope of the appended claims should not be limited to the particular examples disclosed. In the following discussion, the terms “example” and “e.g.,” are non-limiting.

[0007] The figures illustrate the general manner of construction, and descriptions and details of well-known features and techniques may be omitted to avoid unnecessarily obscuring the present disclosure. In addition, elements in the drawing figures are not necessarily drawn to scale. For example, the dimensions of some of the elements in the figures may be exaggerated relative to other elements to help improve understanding of the examples discussed in the present disclosure. The same reference numerals in different figures denote the same elements.

[0008] The term “or” means any one or more of the items in the list joined by “or”. As an example, “x or y” means any element of the three-element set {(x), (y), (x, y)}. As another example, “x, y, or z” means any element of the seven-element set {(x), (y), (z), (x, y), (x, z), (y, z), (x, y, z)}.

[0009] The terms “comprises,” “comprising,” “includes,” and / or “including,” are “open ended” terms and specify the presence of stated features, but do not preclude the presence or addition of one or more other features.

[0010] The terms “first,” “second,” etc. may be used herein to describe various elements, and these elements should not be limited by these terms. These termsare only used to distinguish one element from another. Thus, for example, a first element discussed in this disclosure could be termed a second element without departing from the teachings of the present disclosure.

[0011] Unless specified otherwise, the term “coupled” may be used to describe two elements directly contacting each other or describe two elements indirectly connected by one or more other elements. For example, if element A is coupled to element B, then element A can be directly contacting element B or indirectly connected to element B by an intervening element C. Similarly, the terms “over” or “on” may be used to describe two elements directly contacting each other or describe two elements indirectly connected by one or more other elements.

[0012] Measurements of intracranial pressure (ICP) occur during the diagnosis and treatment of a variety of medical conditions including traumatic brain injury, stroke, and hydrocephalus. Current measurement solutions are invasive, and can take several forms including a pressure gauge inserted directly into the brain, or a catheter inserted directly into the brain’s ventricles. These invasive measurements of ICP carry a substantial infection risk and require highly trained staff, and are thus costly in terms of health, hospital resources, and money. As a result of its risk and cost, ICP measurement is not applied to many conditions where its value is not outweighed by its risks, and cannot be used outside of the operating theater or intensive care unit. For these and other reasons, non-invasive ICP measurements may be desirable, for example using near-infrared (NIR) light in the approximate wavelength range of 700 nm to 2500 nm. The use of NIR light may be advantageous because NIR light is absorbed less than e.g., visible light by skin, bone and / or other human tissues.

[0013] At a high level, the present disclosure relates to an interferometric optical sensing system and method to measure intracranial pressure (ICP). Such a system may comprise two or more paths of light: a first reference path of light, and a second path of light that may travel through an object that is the subject of study. The object for ICP measurements typically may be a person's head. The light that may have traversed the second path may be referred to as sample light and is combined with the reference light. Often, the sample light may becombined with the reference light by interfering the two light signals. The combined light field is then directed towards one or more detectors that may suitably process the received light and convert it into an electrical signal for further processing.

[0014] A signal so obtained may exhibit properties related to blood flow, for example a pulsatile waveform of blood flow, which in turn may be related to ICP, as will be further elaborated upon below.

[0015] To further improve such a system, the interferometric measurements obtained may be combined with measurements of extracerebral blood flow.

[0016] Such a system may be enhanced by using a data driven model based on machine learning.

[0017] Referring now to FIG.1, there is shown an interferometric measurement system 10, to measure intracranial pressure (ICP) in a subject’s head 2.

[0018] The iNIRS system 10 may include a light source 20, a light source modifier 22, and a light splitter 24. There is further shown a light delivery probe 25’, a light receiving probe 35’, a sample delivery channel 25, a reference channel 26, and the sample receiving channel 35. There is also shown a light detector 30 and a controller 40. Inset A shows a more detailed view of the light detector 30 shown in the dashed box shown in Fig.1. Inset B shows a more detailed view of an alternative camera-based light detector 30.

[0019] There may be multiple sample delivery channels 25 and probes 25’, and multiple sample receiving probes 35’ and associated sample receiving channels 35. The sample receiving probe 35’ may be operable to receive light. The sample receiving channel 35 may be a single-mode fiber (typically in conjunction with a photo-diode detector 30 arrangement as illustrated in Inset A), or a multi-mode fiber (typically in conjunction with a camera-based detector 30 arrangement illustrated in Inset B). In accordance with various embodiments of the patent, the system may comprise one or more detectors 30 and / or controllers

[0020] The iNIRS system 10 is shown coupled to an object / head 2 to be imaged / monitored. The light source 20 may comprise a laser enabled to emit light conforming to certain properties, some of which may be controlled by the light source modifier 22. In accordance with various embodiments of the patent, the laser may be a high coherence laser. For example, the laser may be a Distributed Feedback laser (“DFB”) or a MEMS – Vertical Cavity Surface Emitting laser (“MEMS-VCSEL”). Other types of suitable laser include a Distributed Bragg Reflector laser (‘DBR’), a Fourier Domain Mode Locking laser (‘FDML’), a Vertical Cavity Surface-Emitting laser (‘VCSEL’), an external cavity diode laser (ECDL), feedback-stabilized laser, or line-locked laser. In accordance with various embodiments, a Master Oscillator Power Amplifier (MOPA) configuration may be used. As will be clear to the person skilled in the art, various laser configurations may be used for the present patent, comprising e.g., LiDAR lasers, lasers for coherent telecommunications or Optical Coherence Tomography (OCT) lasers. Additionally, or alternatively, a pulsed supercontinuum laser may be used in combination with a pulse stretching mechanism, such as a grating or GRISM pulse stretcher or length of dispersive optical fiber. For example, such an arrangement may be configured to temporally separate the wavelengths in the pulse such that a frequency chirped pulse is created (e.g., for ultimately providing an interferogram when sample and reference pulses are compared).

[0021] The light source modifier 22 may comprise a source for providing a variable electrical control signal (e.g., a variable current or voltage provider) to the light source 20. The light source modifier 22 is coupled to the light source 20. The light source modifier 22 may be electrically connected to the light source 20 to provide a variable current / voltage thereto, enabled to control the light source 20 to desirably adjust properties of the light emitted by the light source 20. Specifically, the light source modifier 22 may interact with the light source 20 to vary the wavelength of the light emitted from light source 20. This may be referred to as wavelength swept emission of light or frequency sweeping.

[0022] The light source 20 is coupled to the light splitter 24. The light splitter 24 may have an input for receiving light from a light source and comprise suitable circuitry and / or hardware that is enabled to output light on two or more outputs.Light splitters 24 of the present disclosure may comprise fiber-optic splitters or free-space beam splitters. The light splitter 24 outputs, for example, are coupled to the reference delivery channel 26 and to the sample delivery channel 25. The splitter may be configured so that the majority of the light is directed towards the subject’s scalp via sample delivery channel 25. For example, the splitter may be a 90:10 splitter, or a 99:1 splitter. Each light detector 30 may be connected to the light source 20 indirectly to receive reference light there from a reference delivery channel 26 and one or more sample delivery channel 35. The sample delivery channel 25, reference delivery channel 26, the sample receiving channel 35, the first sample receiving channel 351, and the second sample receiving channel 352 are enabled to communicatively couple optical devices by enabling light transmission, typically implemented as optical fibers. Some or all of the optical channels of the iNIRS system 10 may be provided by optical fibers. The iNIRS system 10 may include optical devices such as lenses, reflection and / or refraction devices for beam steering, as relevant.

[0023] The sample delivery channel 25 couples the light splitter 24 to the sample delivery probe 25’. The sample delivery probe 25’ will be placed at a location on the scalp of subject’s head 2. The sample delivery probe 25’ is enabled to couple light received from the light splitter 24 to the object / subject’s head 2. Similarly, sample receiving probes 35 are enabled to couple a subject’s head / object 2 to receive light. For example, the sample delivery probe 25’ may include one or more lenses for spatially distributing sample light from the sample delivery channel 25 towards the subject’s brain tissue. As another example, one or more of the sample receiving probes 35 may include a lens for focusing received light into its associated sample receiving channel 35 (as connected to that sample receiving probe). As another example, the probes 25’, 35’ may be bare optical fibers which have been cleaved and / or polished.

[0024] Each sample receiving channel 35 may comprise a single-mode fiber (SMF), a few-mode fiber (FMF), or a multi-mode fiber (MMF). In systems using camera sensors instead of photodiodes, the sample delivery channels 25, the reference channel 26, and / or the sample receiving channels 35’ may often bemulti-mode fiber (MMF). The reference channel(s) 26 may be provided by an SMF, MMF or an FMF.

[0025] The object 2 may be a subject’s head, for example for general interferometric neurobiological imaging. The object 2 may generally be any biological object or biological tissue / matter, for example skin, bone, muscle, fat, and / or brain, and comprise e.g., body fluids such as blood. For ICP measurements, the object 2 may typically be a subject’s head.

[0026] Each sample receiving probe 35’ may be placed on the scalp of the subject head 2. Each sample receiving probe 35’ may be coupled to an associated sample receiving channel 35, that is coupled to an associated light detector 30. In other words, each light detector 30 is connected for indirectly receiving sample light from the light source 20, where the sample light has travelled from the sample delivery probe 25 through the subject head 2 to the (one or more) sample receiving probes 35.

[0027] The detectors 30 are enabled to receive and process optical input signals. Further, the detectors 30 may comprise suitable logic, circuitry and / or code that is enabled to suitably convert and process the received optical input signals to electrical signals. The detectors 30 may generate an electrical output signal that may be coupled to an input of a controller 40. For example, the detector 30 may receive light that has travelled through the object 2 via the sample receiving probes 35’ and the sample delivery channel 35. The detector 30 may also receive light from light source 20 that has not travelled through the object 2 via the reference delivery channel 26, reference receiving channel 36 and the light splitters 24. Further details of detectors 30 will be discussed with reference to Inset A and Inset B below.

[0028] The controller 40 may comprise suitable logic, circuitry and / or code with data receiving and processing functionality. For example, the controller 40 may comprise one or more Application Specific Integrated Circuit (‘ASIC’). Other examples for the controller 40 may include a Field Programmable Gate Array (‘FPGA’) and / or a Data Acquisition module (‘DAQ’). The controller 40 may also comprise a microcontroller or microprocessor. In accordance with variousembodiments of the patent, the controller 40 may comprise machine learning logic, circuitry and / or code.

[0029] The controller 40 is coupled to each of the detectors 30 (shown in FIG. 1 is one detector 30) electrically, as illustrated by the dashed line. The controller 40 may be connected to each detector 30 via a wired connection (for receiving electrical signals indicative of detection therefrom), and / or the connection may be wireless (for receiving transmitted data indicative of detection therefrom). The controller 40 may also be coupled to the light source modifier 22. This connection may be wired or wireless. Correspondingly, based on processing of input signals received from one or more detectors 30, the controller 40 may control the function of the light source modifier 22. In accordance with various embodiments of the patent, the controller 40 may also control the function of the light source modifier 22 independently from the signals received from light detectors 30.

[0030] With reference to Inset A of Fig.1, there is shown an exemplary configuration for a detector 30, which according to various embodiments of the invention, may be used to convert and process optical signals to electrical signals. The exemplary detector 30 may be configured to convert three optical inputs into a discrete-time / digitized, i.e., sampled, electrical signal. The same approach may be used with only two optical inputs, wherein one input may be coupled to a reference receiving channel and one input may be coupled to one sample receiving channel.

[0031] Referring to Inset A of Fig.1, a detector 30 may comprise a light combiner and splitter 301, a balanced photodetector 303, and analogue to digital converter (‘ADC’) 306. An exemplary balanced photodetector 303 may comprise detection photodiodes 310, 312, a transimpedance amplifier (TIA) 304, and an amplifier 305. The ADC 306 is arranged to provide a digital signal output 307, which may correspond to an electrical output of detector 30, as illustrated by a dashed line in Fig.1. The digital signal output 307 may also be referred to as an interferogram, as will be explained below.

[0032] The optical inputs to the light combiner and splitter 301 of detector 30 are coupled to reference receiving channel 36 and exemplary sample receivingchannels 351, 352 (which may be coupled, for example, to one or more sample receiving channels 35, only one sample receiving channel is illustrated).

[0033] One example of an arrangement for converting received light signals into digital data is shown in Inset A of Fig.1. Inset A shows an arrangement of components that may be used as part of a light detector 30 of the present disclosure. As shown in the iNIRS system 10 of Fig.1, the detector 30 is arranged to receive three inputs: (i) reference light which has travelled along reference delivery channel 26 and reference receiving channel 36, (ii) first sample light which has been received through the first sample receiving channel 351, and (iii) second sample light which has been received through the second sample receiving channel 352.

[0034] The light combiner and splitter 301 are arranged to additively combine the electric fields of the light signals received via channels 36, 351, 352. For example, the channels may couple into a beam combination element. This combination could be achieved using e.g., fused fiber couplers, beam splitter cubes, diffraction gratings or more other splitting / combining optical elements such as multiplexing optics. The combined optical signal, sometimes referred to as the mixed signal, is then optically split in light combiner and splitter 301 and output to a first light channel 302a and a second light channel 302b. The mixed signal energy may be split 50:50, for example, between the first light channel 302a and the second light channel 302b. In accordance with various embodiments of the invention, the proportions of how the mixed signal energy is split between the channels 302a, 302b may be adapted to the specific configuration of the detector 30.

[0035] For example, the light combiner and splitter 301 may receive light signals at its inputs that may be represented as electric fields E36(t), E351(t), and E352(t) for each of the channels 36, 351, and 352, respectively. The light combiner and splitter 301 may be enabled to generate output signals proportional to: ^^^^^(^)= ^^^(^)+ ^^(^), ^^^^^(^)= ^^^(^)− ^^(^)

[0036] Where * denotes the complex conjugate operation and Re[.] may denote the real part of a complex quantity. The detector 30 is arranged to combine reference light E36(t) with sample light ^^(^)= ^^^^(^)+ ^^^^(^)(as part of an interferometer). The iNIRS system 10 may be arranged to determine one or more properties of the subject’s brain tissue based on this combination of reference light and sample light (as will be described in more detail below).

[0037] The transimpedance amplifier TIA 304 comprises suitable logic, circuitry and / or code to convert an input current I(t) to a proportional output voltage, i.e., the transimpedance amplifier is arranged as a current to voltage converter. The voltage output of the TIA 304 may be coupled to amplifier 305 for further amplification. In some embodiments, an amplifier 305 may not be necessary, depending on the particular configuration of the photodetector 303.

[0038] The amplifier 305 may be used to scale the output signal to the full range of the ADC 306 and limit the electronic frequency of the circuit to further maximize the SNR. This amplified voltage is then provided to the ADC 306 to be digitized. The ADC 306 comprises a digitizer having sufficient bandwidth so that the full signal bandwidth containing time of flight information may be digitized without attenuation. The ADC 306 may be configured to convert a continuous- time electric input signal to a discrete-time electric output signal, sampled at a desirable sampling rate. This discrete-time signal is output over digital signal output channel 307 and is coupled to controller 40, for suitable processing.

[0039] Each detector 30 may provide part of an interferometer in system 10, such as a Mach-Zehnder interferometer (when receiving sample and reference light from the light source). Each of the different light detectors 30 may be coupled to the same light source 20 (each via one or more reference channels 36). The light detectors 30 may be spatially separated from the light source 20. The sample receiving probes 35’ may also be spatially separated from one another or they may be co-located on a sufficiently similar region of tissue that the received signals can be averaged together. For reference light to reach a light detector(s) 30 from the light source 20, the reference light will travel along one ormore reference channels 26 / 36. For sample light to reach a light detector 30 from the light source 20, the sample light will travel indirectly via the subject’s head 2 brain tissue. The sample light is directed towards the scalp of the subject’s head 2 via one or more sample delivery channels 25’. The sample light may then pass through the subject’s brain tissue and travel into a receiving channel and into the optical detector 30. A first optical path may be the optical path from e.g., splitter 24 via sample delivery probe 25’, through the subject’s head 2, sample receiving probe 35’, and sample receiving channel 35 to detector 30. When there are multiple sample receiving probes, the multiple sample receiving channels 35 generally may be of different (optical path) length, in accordance with various embodiments of the patent. The illumination of the subject’s brain tissue may thus occur using a different light channel to the detection of light from the subject’s brain tissue.

[0040] In accordance with various embodiments of the invention, the iNIRS system 10 may be completely or partially housed within a garment (not shown) for the subject’s head 2. For example, the iNIRS system 10 may be provided in a hat / cap that may be worn by the subject on their head 2. The garment may also be a headband with attachment, or any other suitable fixture to couple at least parts of the iNIRS system 10 to the subject head 2. The head garment may be arranged to hold the light source 20 and detectors 30 in a fixed arrangement relative to the scalp of the subject head 2. For example, the head garment may include a plurality of sample receiving probes 35’ and / or sample receiving channels 35 and / or detectors 30. The controller 40 may be separate to the head garment (e.g., and connected wirelessly or by wire) or it may also be provided as part of the head garment (e.g., by an ASIC within the head garment which may be wire-coupled to the detectors 30 and / or light source modifier 22). For example, the garment may be configured to comprise the sample delivery probes 25’ and channels 25, and the sample receiving probes 35’, and channels 35 with the other components of the system 10 located elsewhere.

[0041] The sample delivery probe 25’ are arranged to be positioned on the subject’s head 2 to provide imaging of a selected region of their brain. At leastsome of the probes 25’ may be arranged to be spatially separated from the light source 20.

[0042] A portion of the light that is delivered to the subject’s head 2 via the sample delivery probe 25’ is received at the sample receiving probes 35’. The light will have travelled through the subject’s head 2. The light will be scattered by the brain tissue, resulting in delay and / or attenuation. The scattering transmission channel for the subject’s head 2 may be modelled as a sum of delayed and attenuated signals that may be received at the sample receiving probes. For example, ^(^)=∑^^^^(^ − ^^)

[0043] where y(t) may be an exemplary received signal at a probe, x(t) may be the transmitted light signal, i may denote the i-th delay path with delay τiand wi may be the i-th attenuation coefficient.

[0044] In some instances, the sample receiving probes may be spatially proximal to each other. In other instances, the sample probes are not substantially collocated and may be separated by a distance.

[0045] The detector 30 may be coupled to the light source 20 via the reference receiving channel 36 (and reference delivery channel 26). The detector 30 may be also coupled to the subject’s scalp via two separate optical channels: the first sample receiving channel 351 and the second sample receiving channel 352 (Inset A). Each of these sample receiving channels may be coupled to the subject’s head 2 via sample receiving probes 35 (a first and second sample receiving probes respectively, only one probe shown in FIG.1). The first sample receiving channel 351 is generally of a different optical length to the second sample receiving channel 352. Multiple source - receiving probe distances may generally be separated by a different distance.

[0046] From here on in, sample light which is received at the detector 30 from the object 2 to be imaged / monitored (e.g., the subject’s brain), and which travelled along the first sample receiving channel 351 will be referred to as ‘firstsample light’. Similarly, sample light which is received at the detector 30 but which travelled along the second sample receiving channel 352 will be referred to as ‘second sample light’. The second sample light will be assumed to take a longer time to travel from the subject’s scalp to the detector 30 than the first sample light due to the delay line (not shown) for the second sample receiving channel 352. Reference light is light which does not travel through the subject head 2 but is coupled to the light source 20 via a reference receiving channel 36.

[0047] The light source 20 may be configured to provide wavelength swept emission of light, for example. For this, the light source 20 may be configured to produce a series of emissions of pulses of light. During each pulse, the wavelength of light may be “swept” through a range of wavelengths. For example, the sweeping may be in the form of a chirped pulse. Light will be emitted at a plurality of different wavelengths during one pulse. For example, the wavelength may continually increase or decrease during one pulse (the rate of change of wavelength may be constant, or it may be variable). The series of chirped pulses may be contiguous (e.g., with a zero inter-pulse time interval). The light source 20 may be configured to successively emit a series of pulses, with each pulse having a wavelength sweep. However, it will be appreciated that the light source 20 need not provide continuous sweeping, or no sweeping at all may be desirable. For example, the light source could be tuned in steps rather than continuously, such that the light source 20 emits light at different wavelengths in different time intervals (e.g., discrete time intervals for emission at each of a plurality of wavelengths). The light source 20 may sweep unidirectionally (e.g., only increasing or decreasing in wavelength during one wavelength sweep), or it may sweep bidirectionally (e.g., both increasing and decreasing in wavelength during one wavelength sweep). Unidirectional sweeping can be beneficial as it increases the number of detected photons per sweep.

[0048] The controller 40 may be configured to selectively control the wavelength sweeping of the light source 20 via the light source modifier 22. The wavelength sweeping of the light source 20 may be controlled by using the light source modifier 22 to apply a corresponding electrical signal to the light source 20. The controller 40 may be arranged to control application of a current / voltageto the light source 20 using the light source modifier 22 to provide a selected pattern for the wavelengths of light emitted by the light source 20.

[0049] The light source 20 may be controlled to wavelength sweep according to a selected pattern for the sweeping. For example, the light source 20 may sweep through a selected range of wavelengths of light and / or the light source 20 may sweep through wavelengths of light according to a selected sweep profile (e.g., linear increasing, sinusoid, triangular etc.). For example, the light source 20 may sweep according to a selected sweeping rate, or a selected total sweeping time. The light source 20 is configured to wavelength sweep light so that during one wavelength sweep, light will be directed towards the subject’s brain tissue through the sample delivery channel (and to the detectors via the reference channels) at each of a plurality of different wavelengths. The wavelength of light emitted by the light source 20 will vary over time. As such, an indication of the time at which light was emitted from the light source 20 may be determined based on a wavelength of that light.

[0050] For example, if the light emitted at sample delivery probe 25 may change its frequency linearly over some period of time as ^(^)= ^^+ ^^ , then observing a frequency of light f1 received via a sample receiving probe (e.g.35’) may be used to estimate the delay introduced by the channel through the subject head 2 from ^, for this particular frequency sweeping pattern.

[0051] Such a frequency difference may be obtained from the balanced photodetector 303. As was explained above, the balanced photodetector 303 generates a signal proportional to ^(^) = 4 ^^[^^^(^)^^∗(^)]

[0052] For example, at a particular instant in time, the reference signal E36(t) may be ^^^ (2^^^^) and the sample received signal may be frequency shifted by ∆^≪^^, i.e. ^^(^)= ^^^ (2^(^^+ ∆^)^) due to a delay introduced by the channel through the subject head 2 in combination with the sweeping in frequency, then by trigonometric identity, ^(^)∝^^^ ^^^(2^∆^^)+ ^^^ (2^(2^^+ ∆^)^). The measured intensity I(t) comprises a high frequency termthat may be low-pass filtered out and a frequency component at the offset frequency ∆^. This may be referred to as beat frequency.

[0053] The light source 20 may be configured to sweep through a selected wavelength range. For example, the light source 20 may be configured to sweep in optical frequency over a range of 50 GHz. For example, this may enable the light source 20 to emit modulated light at a plurality of different wavelengths between e.g., 829.94 nm and 830.06 nm when centered on 830nm for example or between 1309.857 nm and 1310.143 nm when centered on 1310 nm for example. The light source 20 may be configured to sweep through a wavelength range of at least 0.025 nm, such as at least 0.05 nm, such as at least 0.075 nm, such as at least 0.1 nm, such as at least 0.11 nm (e.g., about a wavelength on which it is centered). The light source 20 may have a high output power, a long coherence time, and broad mode-hop free wavelength tuning. The light source 20 may have a relatively narrow linewidth and a longer coherence length, e.g., because the light source 20 will not sweep over particularly large bandwidths.

[0054] Light sources 20 of the present disclosure may be configured to provide emission of high coherence light, e.g., substantially coherent light. It will be appreciated that the light source 20 may not both emit perfectly coherent light and also provide wavelength swept emission of light, e.g., because light at different wavelengths will change phase at different rates. Light sources of the present disclosure may be controlled to sweep through a wavelength range which is relatively narrow compared to their absolute wavelength. Correspondingly, typically ∆^ ≪^^. In other words, the difference between the maximum and minimum wavelengths for one wavelength sweep will be relatively small compared to those absolute wavelengths. Each light source 20 may be configured to emit light (i.e., an electric field) which does not have much change in its phase over time.

[0055] The iNIRS system 10 of the present disclosure will receive e.g., a first sample light, a second sample light, and a reference light, all of which originated from the same light source 20. In other words, one or more samples of light and a reference light may be processed, in accordance with various embodiments of the patent. The light sources of the present disclosure are configured to provide wavelength swept emission of sufficiently coherent light, such that sample and reference light, as received at the optical detector 30, will be in relatively constant relative phase to each other. In other words, the coherence length of the light source may be such that the multiple scattering in the tissue will not reduce the coherence or fringe contrast below a noise floor for the measurement.

[0056] For example, the iNIRS system may be configured to have a coherence length or range of approximately 50 m in air – e.g., the light sources may be selected which have a coherence length of between 50 and 100 m (a coherence time period of between 166 ns and 333 ns). It will be appreciated that this particular range is not intended to be limiting, rather it is illustrative of the approximate range for the light source. The light source may be selected so that it has a coherence length which is two or more times greater than the maximum expected optical path length difference, e.g., the coherence length may be three or four or more times greater. Having a light source with a coherence length which is much greater than the optical path length may increase accuracy for measuring sample light photons which have undergone a large number of scattering interactions within the subject’s brain tissue.

[0057] The iNIRS system 10 may be arranged so that the source-detector path lengths for reference and sample light are different. In other words, the iNIRS system 10 is arranged so that an average, or expected, optical path length for light travelling from the light source 20 to each detector 30 via the subject’s brain tissue will be different to the optical path length for light traveling from the light source 20 to said detector 30 via reference channel(s) 36.

[0058] As will be appreciated in the context of the present disclosure, photons of sample light which are directed towards the subject’s brain tissue may travel from the light source 20 to a light detector 30 via a practically infinite number of different paths through the object 2. A photon of sample light may undergo alarge number of scattering events between the sample delivery probes 25’ and the sample receiving probes 35’. The iNIRS system 10 may be arranged to provide neuroimaging and analysis based at least in part on activity in the subject’s brain tissue, which may affect the sample light received. The time of flight for a sample light photon from light source 20 to light detector 30 will of course increase as the path length it takes increases. As such, a photon which travels a longer path, and penetrates deeper into the subject’s brain tissue, will take longer to arrive at the light detector 30 than a photon that has traveled a shorter path. The longer the time of flight for a sample light photon, the deeper that photon is likely to have penetrated into the subject’s brain tissue. Sample light photons received via the sample receiving probes 35’ will have longer times of flight than reference light via reference channels 26 / 36.

[0059] As will be appreciated in the context of the present disclosure, the path which each individual sample light photon travels through the object 2 to be imaged (between sample delivery probe(s) 25’ and sample receiving probe(s) 35’) cannot be predicted. However, where there are a great number of these sample light photons, the overall time of flight distribution for such sample light photons may be modeled statistically. As such, for a given source-detector probe pair, one or more expected properties for a time-of-flight distribution for sample light may be known. For instance, for each source-detector probe pair, an expected time difference between the shortest time of flight photons and the longest time of flight photons may be known. This difference may be referred to as delay spread of the channel through the object 2. For example, this may be based on previous observable signals for the earliest arriving detectable photons and the latest arriving detectable photons. In other words, for any given source- detector pair, there may be a known maximum expected delay for a resulting time of flight distribution for sample light photons.

[0060] The iNIRS system 10 may be configured such that the light source 20 may be wavelength-swept and the interference pattern measured by detector 30 may be sampled at sufficiently high speed to permit the measurement of the temporal interferogram created with each sweep of the light source 20.

[0061] The resulting temporal interferogram obtained by the detector 30 may comprise a plurality of different beat frequencies due to the various differences in wavelength contained in signals 36, 351, 352, due to the differences in times-of-flight through any associated delay lines, and due to the different times-of-flight of the detected photons through the sampled object 2. The frequency content of such an interferogram may encode the times-of-flight of the detected light through the sampled object for each sample light signal 351 and 352. The higher beat frequencies may correspond to photons with longer times-of-flight.

[0062] The iNIRS system 10 may be configured to obtain a digital representation of each resulting interferogram at the digital signal output 307. For example, the detector 30 may include an analogue-to-digital converter (‘ADC’) 306 configured to obtain discrete- time interferogram data from each interferogram provided by the detector 303. Each obtained interferogram may be Fourier-analyzed (e.g., using an FFT or iFFT) to obtain an indication of a distribution of times-of-flight (‘DTOF’) for sample light photons incident on the light detector. This operation may be implemented in the controller 40, for example. Each determined DTOF may provide a distribution showing the time-of-flight for sample light photons which may be incident on the light detector 30 at a given moment in time.

[0063] When the detector 30 processes multiple received sample signals with different delay lengths, then a single interferograms may be processed to obtain multiple DTOFs.

[0064] When the detector 30 processes multiple received sample signals, multiple interferograms and multiple DTOFs may be obtained.

[0065] As is known to the person skilled in the art, the DTOFs that may be obtained by particular configurations of the iNIRS system 10 may provide time-of-flight data that may relate to absolute optical scattering and absolute optical absorption characteristics of the sampled object 2. The absolute optical scattering and optical absorption coefficients may be computed from the DTOFs. These computations may be performed in the controller 40.

[0066] As is known to the person skilled in the art, the iNIRS system 10 may be configured to provide repeated measurements of DTOFs over time, at a given rate. Changes in the DTOF over time may relate to changes in the absolute optical scatteringand absolute optical absorption coefficients, and thus changes in these coefficients over time, which may be computed in the controller 40.

[0067] By measuring the temporal autocorrelation of a signal derived from the DTOFs over many DTOFs, a measure of blood flow may be obtained. The faster the autocorrelations of the DTOF decrease in amplitude (i.e., the faster the signal decorrelates), the higher the observed blood flow. The obtained data of autocorrelation decay may be referred to as G1, and may be a function of the time-of-flight τs, the time between successive DTOF measurements td, and the autocorrelation lag τd.

[0068] Referring now to FIG.1, Inset B, there is shown a reference receiving channel 26, a sample receiving channel 35, the beam splitter 60, optical assemblies 50, 55, camera sensor 70, and the signal output 307.

[0069] A detector 30 may comprise a camera sensor 70. For example, a reference signal via reference delivery channel 26 may be delivered to a beam splitter 60 through optical assembly 50, which may comprise a collimating lens. A sample signal via sample delivery channel 35 may be delivered to beam splitter 60 through optical assembly 55, which may comprise a collimating lens. In such a case, the sample signal received via sample receiving channel 35 may comprise multiple modes and a larger beam width representing multiple speckles. The image received by a sample receiving channel 35 may thus be interfered with a reference signal via reference delivery channel 26 in beam splitter 60. The interfered signal output by beam splitter 60, may be received on camera sensor 70. The camera sensor 70 may comprise a plurality of photosensitive sites, for example an m x n array of pixels. Each pixel may be operable to record a speckle received from beam splitter 60. As will be clear to the person skilled in the art, the physical arrangement of the photosensitive sites on camera sensor 70 may equally be a square, circular, elliptical or any other suitable arrangement of photosensitive sites.

[0070] In a camera-based approach illustrated in inset B, multiple modes of detected light (via 35) may be interfered with a same reference light (via 26) for different pixels of the camera sensor 70. Correspondingly, we may process signals from each sensor pixel in a similar fashion to a signal received at a photodiode. For each pixel p, we may thus obtain a light intensity by:and where <.> may be ensemble averaging. The intensities obtained for different pixels may then be averaged to improve the signal-to-noise ratio. A camera-based approach as illustrated in inset B, may be thought of as a relatively simple implementation of having a number of parallel photodiode detectors in parallel.

[0071] A camera sensor-based system may be implemented both with a tunable laser as discussed above for inset A, or also with a laser that is not modulated, i.e. that may not use a specific sweeping pattern and may be emitting at a constant wavelength.

[0072] In the case where the iNIRS system 10 may comprise a camera sensor 70, the temporal sampling speed of the camera sensor may not be sufficient to generate a temporal interferogram. The iNIRS system 10 may then be configured such that the interference pattern may be sampled repeatedly by a camera sensor at sufficient speed to permit the measurement of the autocorrelation decay of the electric field measured at each pixel due to the movement of scatterers within the sampled object for each camera pixel (or subset of camera pixels) without first measuring the temporal interferogram. The obtained data of autocorrelation decay may be referred to as G1.

[0073] In the case where the iNIRS system 10 may comprise a camera sensor 70, it may be the case that the temporal sampling speed of the camera sensor is not sufficient to measure temporal speckle intensity fluctuations and thus generate the autocorrelation decay measurements. The iNIRS system 10 may then be configured such that the interference pattern may be sampled repeatedly by a camera sensor and an index of blood flow may be determined from the spatial speckle statistics of the captured camera frames. This approach may be termed speckle contrast spectroscopy.

[0074] The speckle contrast may be proportional to the lag-time integrated autocorrelation, which may provide a measure of blood flow: speckle contrast

[0075] As mentioned above, a camera sensor-based system may be implemented without a modulated / tuned laser and in such a case, the system may not provide time-of-flight resolution. We may then obtain G1(p,^) for each pixel p of the camera sensor at delay ^ by autocorrelation of I(p,^, t) and then average G1(p,^) over all pixels p to obtain a G1(^). The speed of decorrelation of G1(^) may be directly related to the rate of blood flow (volume per time).

[0076] As set out above, the controller 40 may be configured to determine a Cerebral Blood Flow index (‘CBFi’) for the subject’s brain tissue. The CBFi may provide an indication of movement occurring within the subject’s brain. For example, the CBFi may provide a relative measure that describes movement occurring within the subject’s brain. The movement may be attributed to blood moving (e.g., flowing) through the subject’s brain. The CBFi may provide a relative measure. The CBFi may contain an indication of mean squared displacement of scatterers in bulk tissue which is measured in units of cm^2 / s. For example, the CBFi for a given volume within the subject’s brain may provide an indication of movement occurring within that volume. That movement may be attributed to movement of blood, and so the CBFi may provide an indication of mean blood flow in a way that relates to Cerebral Blood Flow Velocity (CBFV, cm / s) and to blood tracer clearance (ml / 100g / min). For example, an increased value for CBFi may indicate an increased velocity occurring within the volume of the brain.

[0077] The controller 40 may be configured to obtain cerebral blood flow data (i.e., data containing an indication of blood flow within the subject’s brain tissue) that may contain an indication of one or more pulses of blood flow through the subject’s brain tissue. The controller 40 may determine the CBFi for the subject’s brain tissue, wherein the pulses of blood flow through their brain tissue may be observable in the determined CBFi for their brain tissue. An example of such a determined CBFi may be illustrated in FIG.2.

[0078] Referring now to FIG.2, FIG.2 shows an example of a determined CBFi evolving over the time of one cardiac cycle for a subject’s brain 2 tissue. In FIG.2, there is one pulse shown. The pulses may occur in a cyclical and repetitive manner (although each individual pulse may vary in size and shape, and the frequency at which pulses occur may also change). Typically, each pulse may have a diastolic (lower) CBFi value and a systolic (higher) CBFi value, as illustrated. During one pulse, the CBFi may start (arbitrarily chosen) at its diastolic value, and then rapidly increase up to its systolic value (P1). The CBFi will then drop back towards its diastolic value from the systolic value, but the CBFi may temporarily increase (or at least begin to decrease at a slower rate) at least onceas it drops from systolic to diastolic values. In other words, there may be local peaks, as illustrated in FIG.2 by P2 and P3. Once back at the diastolic value, the process may repeat again with the subject’s heartbeat.

[0079] The iNIRS system 10 may be configured to use optical interferometry to obtain measurement signals from which the controller 40 of the iNIRS system 10 may obtain cerebral blood flow data indicating this behavior for the blood flowing through the subject’s brain tissue (e.g., the controller 40 may obtain the temporal evolution of the CBFi values as shown in Fig.2).

[0080] The controller 40 may be configured to process the cerebral blood flow data to determine an indication of ICP for the subject’s brain tissue. The controller 40 may be configured to determine ICP based on one or more properties of the pulsatile waveform for the blood flow. For this, the controller 40 may use any of a number of suitable features. For example, the features may describe the shape of the pulsatile waveform, or other measures relating to the pulsatile waveform.

[0081] The volume of a subject’s cranial tissues may be constrained by their skull. Cranial tissues here may refer to the tissues within the skull and underneath the dura mater, and may comprise brain, cerebral spinal fluid, and blood. The skull may typically be very rigid, and so the subject’s cranial tissues may be constrained in a fixed internal volume of the skull. This internal volume does not change, and so, as the pressure within the skull increases, blood flow through the matter in that internal volume may become restricted. For blood to flow through the subject’s brain vasculature, either blood, cerebrospinal fluid or brain tissue must be displaced from the skull’s fixed volume, or else the pressure within the skull’s fixed volume must increase. At short time scales, this may result in flow within the brain that fluctuates with the heart-beat. The shape of this flow waveform may depend on the opposing pressure within the skull, which is ICP.

[0082] Thus, increasing ICP may restrict blood flow within the skull. In circumstances where ICP may be too high for too long, blood flow may be so severely restricted as to result in insufficient oxygenation of brain tissue, leading possibly to brain damage.

[0083] The controller 40 may be configured to determine an indication of the influence of ICP on the pulsatile waveform for blood flow through the subject’s brain tissue using the cerebral blood flow data. The controller 40 may accomplish this by looking for changes in specific features of the pulsatile waveforms of CBFi and / or Arterial Blood Pressure (ABP).

[0084] Arterial Blood Pressure (ABP) may specifically refer to blood pressure and blood flow data that may be obtained from elsewhere than within the skull. Because ABP may be measured outside of the skull, it is not subject to ICP and may indicate properties of blood flow and pressure within the subject, but not under the influence of ICP. The ABP data may thus serve as additional and / or reference data relevant to ICP and / or blood flow.

[0085] The controller 40 may be configured to identify features as changes in the pulsatile waveform, e.g., its shape changing and / or its maximum / minimum values changing (e.g., P1, P2, P3), and may use this information to determine an indication of the subject’s ICP. Pulse flow waveforms may contain a characteristic peak representing the maximum cerebral flow during the cardiac cycle (systole, FSYS, P1), and a trough representing the minimum (diastole, FDIA). Within the cardiac cycle, pulse flow waveforms may comprise several identifiable cardiac features, most obviously with three peaks P1, P2 and P3 that may correspond to the ‘percussive wave’, the ‘tidal wave’, and the ‘dicrotic wave’ of the cardiac cycle. In some conditions when the brain may be unable to regulate its own volume effectively (i.e., disrupted cerebral autoregulation), ICP might be very high (e.g., >20mmHg) and the shape of the pulse flow waveform features may change as a function of the level of ICP. One specific change in the CBFi flow waveform may be a normalized difference between diastolic minimum and the systolic maximum (FSYS - FDIA / FBAR as illustrated in FIG.2). This relationship may be called the Pulsatility Index (PI) and it may often exhibit a linear relationship with ICP. Other specific features that may be expected to change with ICP are the slopes and integrals of a CBFi waveform features as illustrated in FIG.2, or the amplitudes of individual waveform samples at any particular phase of the pulse cycle. The controller may also be configured to extract features related to ICP using a comparison of ABP measured using an external device and CBFiobtained through interferometric measurement. These features may include the absolute values of CBFi, ABP and any combination of those two waveforms including the phase relationship between them, or the amplitude, prominence, width and statistical features of peaks within the pulse cycle, including P1, P2 and P3. Features may exhibit over more than one pulse cycle; for instance, P1 may be modulated at the respiratory frequency and the strength of this modulation may also be used as a feature. Depending on the specific application, it may be desirable to synchronize a ABP and CBF measurements for further processing.

[0086] The controller 40 may also be configured to extract features from the CBFi waveform alone, or in combination with the ABP waveform using ‘unsupervised’ methods that do not require a human to explicitly identify features.

[0087] Referring to FIG.3a, there is illustrated exemplary types of ICP estimation, specifically types 1 - 3. All exemplary processes may comprise CBFi, ABP, and extracerebral blood flow inputs. Type 1 may illustrate a more traditional process, as described above. A type 2 process may be based on unsupervised feature selection from the input data. The features selected may then be mapped onto an ICP estimation using a regression model. A type 3 process may be enabled to skip explicit feature extraction, as will be further described below.

[0088] Unsupervised methods for feature extraction may include any form of matrix decomposition, or any neural network that engages in so-called representation learning, and may be thus designed to learn the common structure within these time-series inputs. An example of how matrix decomposition may be used for feature extraction is by applying Principal Component Analysis (PCA) to a matrix (m,n) where ‘m’ may be a dimension whose length equals the number of pulse waveforms (samples) and ‘n’ may be a dimension whose length equals the number of waveform features. PCA may result in a lower dimensional embedding of new features ‘k’ where each ‘k’ may be an affine transformation of the original features ‘n’, and where k <= n. Each new feature ‘k’ may be ranked by the variance it explains of the original dataset of ‘m’ samples, and the top new features ‘k’ may be used with a regression model that maps ICP onto these features. In other words, PCA may provide a lower dimensional approximation to the input data. Features ‘k’ found using a matrix decomposition such as PCAmay not easily be described in relation to any physical characteristic of the CBFi pulse waveform because they may exist in a dimensional space that may be rotated from the original space which may have dimensions labeled by each feature. PCA feature vectors may be, therefore, dimensionless and abstract, but may be mapped back to the original feature space. In this way, the most significant PCA feature might take the form of a vector of weights that are applied to each input feature such as, for example, a1*prominence(P1), a2*time(P2), a3*(width(P1) - time(P3)), and where the set {ai} may be a set of weighing factors. An example of a neural network architecture that may implement representation learning capable of learning these features may be an autoencoder network. Such a network may transform input features into a typically lower dimensional embedding space via one or more network layers, whilst minimizing an error when predicting its own inputs. If the pulse waveform is, for example, represented by 100 time points, an autoencoder network might embed these 100 features to a space with e.g., 8 dimensions, where each 8-dimensional vector may comprise sufficient information for the network to recreate the original inputs with acceptable accuracy. Then, the 8-dimensional representations may be used as feature inputs to a regression model that maps ICP onto these features. As will be clear to the person skilled in the art, any number of features and any number of dimensions may be used in practical applications and the present disclosure is not to be construed as limited. In accordance with various embodiments of the patent, the iNIRS system 10 and specifically controller 40 may be implemented using any one of numerous machine learning technologies. As an alternative to using simple linear regression to relate ICP to CBFi features, machine learning based regression models enable more complex mappings between ICP and CBFi and may provide better estimation accuracy.

[0089] A desirable machine learning approach may predict ICP based on various predictor features, such as those mentioned above. The prediction model may aim to minimize an error between the predicted ICP and combinations of the features during training and validation, enabling the trained model to then predict ICP based on data inputs only. These models take the form of regression models that relate a set of continuous features onto a continuous target ICP values, or that may classify inputs onto a set of discrete but numerical targetvalues of ICP and thus perform regression using multi-class classification. Regression models may vary in complexity and may also take the form of ensemble models that accumulate the predictions from many individual regressors or classifiers. An example of an ensemble model is a Random Forest Regressor which may be an ensemble of many decision trees that each may relate one or more data features to a target variable, ICP in this application. In this example, a Random Forest Regressor would be constructed to contain e.g., 100 decision trees, which each may be given a random e.g., 50% subset of e.g., 153 input features which in this example may be the prominence, width, and timing of P1, P2, P3, measured from both CBFi and ABP, and pairwise relationships between them. At each node in each decision tree, the value of an input feature may be ‘split’ such that the next node either does or does not receive the input feature. The value of this ‘split’, or decision boundary, may be adjusted iteratively over the set of training data such that the final set of tree nodes classifies the training data to target classes with a minimum error. In this example, target classes may be each possible value of ICP in a range of e.g., 0- 60 mmHg in increments of e.g., 1mmHg. The prediction of each tree may be averaged and the ensemble prediction may become the model output.

[0090] The controller 40 may also be configured to infer ICP through a pre- trained model that may not require a separate feature extraction step, as illustrated in the type 3 process of FIG.3a. Such a process may be referred to as a sequence-to-sequence regression model approach. This class of models may directly learn the correspondence between pulse waveforms (CFBi alone, or CBFi and ABP) and ICP targets. These forms of model may be called ‘supervised representation learning’, and directly map the CBFi values from a time-series of input pulse waveforms onto ICP values without an explicit intermediate representation in feature space. An example of this sort of model is a Long Short-Term Memory (LSTM) neural network that may accept e.g., a continuous sequence of 10 pulse waveforms and may predict from it an ICP value. In this example, the model inputs may be two vectors, one containing e.g., 1500 CBFi values (e.g., 150 values for each of e.g., 10 pulse waveforms), and the other e.g., 1500 ABP values, and the target may be e.g., a single ICP value which may be the average ICP measured over the same duration. LSTMnetworks are a class of recurrent neural networks that may be able to process long sequences of data and maintain and update a representation of the history of those sequences, which may be called the LSTM’s ‘cell-state’. The LSTM network may accomplish this by reading in each individual time-point and updating a ‘hidden-state’, which may impact the ‘cell-state’ through a series of filters that may be implemented as recurrent neural networks. In this fashion, an LSTM network may be able to simultaneously learn a statistical structure of its sequence inputs (such as, e.g., its periodicity), and learn to map those inputs in a way to predict a target output with minimal error (in some sense).

[0091] Another example for determining ICP based on cerebral and extracerebral blood flow data is shown in FIG.3b. In FIG.3b, the cerebral blood flow data may be plotted against the extracerebral blood flow data, which may be shown as CBFi on the y-axis and ABP on the x-axis. The two data plots (i.e., data for the cerebral and extracerebral blood flow) may be aligned on the graph, e.g., so that their linear regions substantially overlie each other (or at least are close to one another in the event that the data plots do not perfectly overlie each other). As may be seen in Fig.3b, there is a linear region from where the line crosses the x-axis (labeled ‘CrCP’ for critical closing pressure) up to a region (labeled ‘dicrotic notch’) where the data diverges and there is a looped region. As can be seen in Fig.3c, the curve crosses the x-axis (i.e., CBFi value of 0) with a positive ABP value. The controller 40 of the iNIRS system 10 may be configured to process the cerebral and extracerebral data to identify the value of this point, the CrCP.

[0092] The CrCP may represent a critical closing pressure for blood vessels in the subject’s brain tissue. As set out above, the CBFi may provide information relating to movement of blood flow through vessels in the subject’s brain tissue. As the CBFi value decreases, this may indicate that there is less movement of blood within the subject’s brain tissue (e.g., a zero measurement for CBFi may represent an absence of any blood flow). The intersection on the x-axis in the graph of Fig.3d represents the value for arterial blood pressure at which there is a zero CBFi (i.e., at which no blood is flowing through the vessel in the subject’s brain tissue). This is referred to as a critical closing pressure, as it is a pressure at which small blood vessels in the brain ‘close’. In other words, the value forCrCP represents the conditions under which the pressure in that blood vessel may be sufficiently low that the blood vessel may be forced to close (due to external pressure, ICP, being applied thereto).

[0093] It will be appreciated that the controller 40 need not produce graphs such as FIG.3b, these are shown to illustrate the method. The controller 40 may be configured to compare the extracerebral blood flow data (e.g., blood pressure data for one or more of the subject’s veins / arteries outside of their brain tissue) and the cerebral blood flow data (e.g., CBFi values from the subject’s brain tissue). Based on this comparison, the controller 40 may be configured to identify a corresponding pressure (e.g., within the subject's skull) at which the blood vessel in their brain tissue closes. The controller 40 may be configured to determine an indication of ICP based on this CrCP value (e.g., by determining a required pressure in the subject’s brain which would cause the cerebral vessels to close). This CrCP value may also be used as a feature in any of the Type 1 and Type 2 ICP estimation methods, or it may be used as a continuous input into any Type 3 ICP estimation method.

[0094] In the methods where extracerebral blood flow data is available, it will be appreciated that in addition to deriving such data from interferometric data, any suitable other device may be used to obtain the data. For example, the iNIRS system 10 may be provided in combination with a separate sensing element which may be configured to obtain such data. The separate sensing element may be configured to measure blood pressure for a separate region of the subject’s body – such as on one of their limbs – e.g., at an arm or on their hand / fingers. The sensing element may be configured to obtain data containing an indication of one or more properties of the pulsatile waveform in that region of the subject’s body (e.g., including diastolic and / or systolic values, and / or average values). For example, the sensing element may be configured to obtain an indication of how the blood pressure changes over time during the pulsatile waveform for blood through the subject’s vein / artery. This data may be referred to as ABP data throughout this patent disclosure.

[0095] Referring now to FIG.4a, there is shown a system similar to that shown in FIG.1. Same reference numbers shall refer to similar elements as inFIG.1. FIG.4a may be used to describe one way of obtaining ABP data from an interferometric system, such as iNIRS system 10.

[0096] As set out above, the iNIRS system 10 may be arranged for light from a light source 20 (not shown) to be split (by light splitter 24) onto a sample delivery channel 25 and a reference delivery channel 26. The optical detector 30 may then receive reference light (from reference delivery channel 26) and sample light (from sample receiving channel 35). A sample delivery probe 25’ and a reference delivery probe 35’ are also shown.

[0097] FIG.4a may show different layers of the subject’s head. The top layer may be scalp skin surface 101. The second layer, which is beneath the scalp skin surface, may be the scalp tissue 102. The scalp tissue may comprise a plurality of arteries and veins (not shown). Beneath the scalp tissue may be the subject’s skull 103, and located within the skull may be brain tissue 104. The veins and arteries within the subject’s scalp tissue are relatively unconstrained by the skull and by ICP. That is, in this region of the subject’s body, the veins / arteries are not impeded from expanding in volume by the skull (whereas blood vessels within the subject’s brain tissue are).

[0098] For photons of light to reach the subject’s brain tissue from the light source 20 (and through sample delivery channel 25 and sample delivery probe 25’), they must penetrate through the subject’s scalp skin surface, the scalp tissue and the skull. For photons of light to reach the veins / arteries in the subject’s scalp tissue, they must only penetrate through the subject’s scalp skin surface (and relevant parts of the scalp tissue). Two example photon paths are shown for sample light reaching the detector 30 from the light source 20: shallow photon path 202 and deep photon path 204. The deep photon path between light source 20 and light detector 30 is much longer than the shallow photon path (as it travels deeper into the subject’s brain tissue). For both photon paths, there are a number of scattering events which occur within the subject’s head. In this example, a photon traveling along the shallow photon path will interact with at least one blood carrying region of the subject’s scalp tissue (e.g., a vein or artery), and a photon traveling along the deep photon path will interact with at least one blood vessel within the subject’s brain tissue.

[0099] The light source 20 will emit a much larger number of photons, and the paths they take through the subject’s head from source 20 to detector 30 will vary. The sample delivery probe 25’ and the sample receiving probe 35’ are spatially arranged on the subject’s scalp so that at least some of the sample light photons arriving at the detector 30 from the light source 20 will have traveled along shallow photon paths, and some will have traveled along deep photon paths. For example, the probes will be sufficiently spatially separated so that some deep photon paths occur.

[0100] FIG.4b shows a time-of-flight distribution for the sample light photons received at the detector 30. As shown, the peak number of sample light photons may occur for a relatively short time of flight, and then for longer times of flight, the number of photons arriving decreases. As described above, the time of flight for sample light photons approximately corresponds to their penetration depth. That is, deeper penetrating photons will have longer times of flight than shallower penetrating photons.

[0101] Correspondingly, the controller 40 may, for example, measure superficial blood flow data by separating shallow penetrating photons corresponding to extracerebral blood flow from deeper penetrating photons corresponding to blood flow in the skull by time-of-flight.

[0102] The controller 40 may be configured to process the cerebral and extracerebral blood flow data separately. For example, the controller 40 may be configured to obtain time of flight data (e.g., which contains measured time of flight distributions), and to separate this time-of-flight data into different data streams: (i) an extracerebral data stream (for the shorter time of flight photons likely to be associated with the scalp tissue), and (ii) a cerebral data stream (for the longer time of flight photons likely to be associated with the brain tissue). Each data stream may be processed independently in the manner set out above – to obtain blood flow index data containing an indication of one or more pulses of blood through the relevant region of the subject’s body (e.g., through veins / arteries in their scalp or blood vessels in their brain tissue). In this manner the extracerebral data stream may be used in conjunction, or as a replacementto, the Arterial Blood Pressure data stream used in type 1,2,3 ICP estimation methods.

[0103] In the examples described above in relation to the figures, the iNIRS system 10 uses one light source and one detector (i.e., there is only one source- detector channel present). However, this should not be considered limiting. The iNIRS system 10 may include a plurality of light sources and / or a plurality of light detectors. Each light source may be coupled to a plurality of other detectors (e.g., via a plurality of reference channels). The plurality of light detectors may be collocated (e.g., they may be provided in the same region on the subject’s scalp / or in close proximity to each other on the subject’s scalp). The different light detectors may each then be arranged to detect photons of sample light from the same light source which have traveled through similar regions of the subject’s brain tissue. The controller 40 may be configured to combine data from the different light detectors (e.g., to average them to provide a single combined read- out for the subject’s brain tissue). For example, the controller 40 may be configured to determine cerebral blood flow index data (and optionally also extracerebral data) for the subject’s brain tissue based on data obtained using a plurality of optical detectors. This arrangement may yield improved signal to noise for measurements obtained using the iNIRS system 10.

[0104] The iNIRS system 10 may include a plurality of source-detector channels which are associated with different regions of the subject’s brain tissue. For example, there may be a plurality of optical detectors in different locations on the subject’s scalp (these may be coupled to the same light source or different light sources). The controller 40 may be configured to use the plurality of detectors of the iNIRS system 10 to obtain a plurality of ICP measurements (as described above, but for each of a plurality of detectors). The controller 40 may then determine an ICP value for the subject based on the plurality of ICP measurements from the different detectors. This arrangement may increase reliability for measurements, e.g., because ICP may be relatively constant for different regions of the subject’s brain tissue.

[0105] In examples described above, the iNIRS system 10 may use one source-detector channel to obtain both the cerebral and extracerebral blood flowdata. However, the iNIRS system 10 need not obtain the extracerebral blood flow data – for example, an additional sensor may be provided for obtaining extracerebral blood flow or blood pressure values. Additionally, or alternatively, a different iNIRS source-detector channel could be used for obtaining the extracerebral data, such as where the source and detector are arranged closer to each other on the subject’s scalp (e.g., to increase selection of shorter traveling sample light photons from source to detector). For example, the iNIRS system 10 may comprise an extracerebral source-detector channel (one source and one detector arranged for measuring extracerebral blood flow, and one or more (e.g., a plurality of) optical detectors arranged for measuring cerebral blood flow.

[0106] In examples described above, the extracerebral blood flow data may be obtained using the iNIRS system 10 to measure properties of blood flow in the subject’s scalp tissue. However, this should not be considered limiting, as other regions could be used for the extracerebral blood data. For example, the iNIRS system 10 may be configured to measure properties for the subject’s neck, ears, forehead etc., or even for regions further away from their brain tissue (e.g., where more than one source-detector channel is used). The iNIRS system 10 may be configured to obtain blood flow data for a region of the subject’s body (not within their skull) where the blood flow being monitored does not have significant external constraints (i.e., where the blood is traveling through a region which is not sufficiently compressed). For example, the extracerebral blood may be flowing in veins / arteries beneath their skin in a region not under great compression by surrounding material.

[0107] FIG.4c illustrates two different photon time of flight distributions, for a short source-detector separation 405 and a long source-detector separation 410.

[0108] In a system 10 setup with a camera sensor and e.g., without modulating the light source 20, as illustrated in inset B, for example, the received sample light may not be easily separable by time of flight. However, by using multiple source-detector pairs and / or adjusting the source-detector separation distance, it may be possible to select for longer or shorter photon paths to dominate the received signal measurements.

[0109] Referring to FIG.4C, there is shown a short source-detector separation 405 time of flight distribution of photons. In this case, a separation between a source 25' and a detector 35' may be 5 mm, for example. There is also shown a long source-detector separation 410 time of flight distribution of photons, where the separation between the source 25' and the detector 35' may be 20 mm, for example. The plots of 405, 410 are not to scale and merely exemplary, as are the source-detection separation distances for 405, 410.

[0110] For a short source-detector separation 405, most of the photons will travel short, shallow paths between source 25' and detector 35'. Correspondingly, most of the photons will travel through extracerebral tissue, as illustrated by shallow photon path 202 in FIG.4A. For a long source detector separation 410, many more photons will travel along deeper photon paths, as illustrated by deep photon path 204 in FIG.4A. As illustrated in FIG.4C, the mean travel time of photons and thus the depth of the mean path may increase with the source detector separation. Correspondingly, by selecting a suitably short source- detector separation, a measurement may be dominated by extracerebral photon paths even if there is no direct time of flight selectivity available.

[0111] One system 10 setup, for example, may be to use a single source probe 25' and two sample light receiving probes 35', a first one at a short source- detector separation distance whose measurements may be dominated by extracerebral photon paths. Another detector with a long source-detector separation distance would be used for measurements dominated by longer photon paths and thus intracerebral blood flow.

[0112] The iNIRS system 10 may include a plurality of light sources. The light sources may emit light in different wavelength ranges. The iNIRS system 10 may be configured to determine the ICP based on measurements obtained using light in each of a plurality of different wavelength ranges. The iNIRS system 10 may include a first light source configured to emit light in a wavelength range above an oximetry isosbestic wavelength and a second light source configured to emit light in a wavelength range below an oximetry isosbestic wavelength. The controller 40 may be configured to determine ICP based on received light signals from both the first and second light source. The provision of such a first and second lightsource may enable greater reliability for measurements where blood oxygenation values vary (e.g., because one of the light sources will be emitting light in a wavelength range more suited for absorption by oxygenated hemoglobin, and another light source will be emitting light in a wavelength range more suited for absorption by de-oxygenated hemoglobin.

[0113] It will be appreciated in the context of the present disclosure that examples described herein are not intended to be limiting. Instead, examples describe certain potential ways of implementing the claimed technology. For example, the iNIRS system 10 is described with a series of optical cables providing channels and probes for coupling those channels to the subject’s scalp. However, it will be appreciated that the probes themselves may be part of the optical channels, or probes may not be provided at all. Similarly, the arrangement of reference channels is just intended to show that reference light is delivered from the light source to the light detectors via optical channels (rather than via the subject’s brain tissue). For example, each light source may include one reference channel for each light detector, where that reference channel directly connects the light source to the light detector. In which case, there may be no reference connections in the system at all. Alternatively, the reference light may be transmitted on a common reference optical channel, where some of that reference light is taken from the common reference optical channel to each of the optical detectors. The light source may also be arranged to deliver light to one of a plurality of different locations on the subject’s scalp. For example, the light source may be coupled to a plurality of different sample delivery channels, each extended towards the subject’s scalp (e.g., from a light splitter).

[0114] It will be appreciated from the discussion above that the examples shown in the figures are merely exemplary, and include features which may be generalized, removed or replaced as described herein and as set out in the claims. With reference to the drawings in general, it will be appreciated that schematic functional block diagrams are used to indicate functionality of systems and apparatus described herein. In addition, the processing functionality may also be provided by devices which are supported by an electronic device. It will be appreciated however that the functionality need not be divided in this way, andshould not be taken to imply any particular structure of hardware other than that described and claimed below. The function of one or more of the elements shown in the drawings may be further subdivided, and / or distributed throughout the apparatus of the disclosure. In some examples the function of one or more elements shown in the drawings may be integrated into a single functional unit.

[0115] As will be appreciated by the skilled reader in the context of the present disclosure, each of the examples described herein may be implemented in a variety of different ways. Any feature of any aspects of the disclosure may be combined with any of the other aspects of the disclosure. For example, method aspects may be combined with apparatus aspects, and features described with reference to the operation of particular elements of apparatus may be provided in methods which do not use those particular types of apparatus. In addition, each of the features of each of the examples is intended to be separable from the features which it is described in combination with, unless it is expressly stated that some other feature is essential to its operation. Each of these separable features may of course be combined with any of the other features of the examples in which it is described, or with any of the other features or combination of features of any of the other examples described herein. Furthermore, equivalents and modifications not described above may also be employed without departing from the invention.

[0116] Certain features of the methods described herein may be implemented in hardware, and one or more functions of the apparatus may be implemented in method steps. It will also be appreciated in the context of the present disclosure that the methods described herein need not be performed in the order in which they are described, nor necessarily in the order in which they are depicted in the drawings. Accordingly, aspects of the disclosure which are described with reference to products or apparatus are also intended to be implemented as methods and vice versa. The methods described herein may be implemented in computer programs, or in hardware or in any combination thereof. Computer programs include software, middleware, firmware, and any combination thereof. Such programs may be provided as signals or network messages and may be recorded on computer readable media such as tangible computer readable mediawhich may store the computer programs in non-transitory form. Hardware includes computers, handheld devices, programmable processors, general purpose processors, application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), and arrays of logic gates.

[0117] Any controller of the present disclosure may be implemented with fixed logic such as assemblies of logic gates or programmable logic such as software and / or computer program instructions executed by a processor. The controller may comprise a central processing unit (CPU) and associated memory, connected to a graphics processing unit (GPU) and its associated memory. Other kinds of programmable logic include programmable processors, programmable digital logic (e.g., a field programmable gate array (FPGA), a tensor processing unit (TPU), an erasable programmable read only memory (EPROM), an electrically erasable programmable read only memory (EEPROM), an application specific integrated circuit (ASIC), or any other kind of digital logic, software, code, electronic instructions, flash memory, optical disks, CD-ROMs, DVD ROMs, magnetic or optical cards, other types of machine-readable mediums suitable for storing electronic instructions, or any suitable combination thereof. In particular, any controller of the present disclosure may be provided by an ASIC.

[0118] Other examples and variations of the disclosure will be apparent to the skilled addressee in the context of the present disclosure.

Claims

What is claimed: Claims 1. A method for estimating intracranial pressure using a regression model, the method comprising: generating optical interferometric measurements relating to cerebral blood flow; estimating intracranial pressure using said regression model based on said optical interferometric measurements relating to cerebral blood flow and one or more features of extracerebral blood flow or extracerebral blood pressure.

2. The method according to claim 1, wherein said estimating intracranial pressure comprises unsupervised feature selection.

3. The method according to claim 2, wherein said unsupervised feature selection comprises a lower dimensional matrix decomposition of an input data matrix.

4. The method according to claim 3, wherein said lower dimensional matrix decomposition is a principal component analysis.

5. The method according to claim 1, wherein said estimating intracranial pressure comprises a representation learning neural network.

6. The method according claim 5, wherein said representation learning neural network is an autoencoder network.

7. The method according to claim 5, wherein said representation learning neural network comprises a random forest regressor.

8. The method according to claim 1, wherein said estimating intracranial pressure comprises a sequence-to-sequence regression model approach.

9. The method according to claim 8, wherein said sequence-to-sequence regression model approach is implemented as a Long Short-Term Memory (LSTM) neural network.10.The method according to claim 1, wherein said generating optical interferometric measurements is using a plurality of sensors and / or a camera sensor. 11.The method according to claim 1, wherein said one or more features extracted from a measure of extracerebral blood flow is an estimate of arterial blood pressure (ABP) data. 12.The method according to claim 1, comprising using near-infrared wavelengths for generating said optical interferometric measurements. 13.The method according to claim 1, comprising varying near-infrared wavelengths in a sweep pattern for generating said optical interferometric measurements. 14.The method according to claim 1, comprising selecting short time of flight optical interferometric measurements to estimate said one or more features of extracerebral blood flow. 15.The method according to claim 1, comprising selecting long time of flight optical interferometric measurements to estimate said plurality of features of said cerebral blood flow.

16. The method according to claim 1, comprising estimating said intracranial pressure using supervised learning.

17. The method according to claim 16, comprising using a random forest regressor for said supervised learning. 18.A system for estimating intracranial pressure using a regression model, the system comprising one or more circuits to: generate optical interferometric measurements relating to cerebral blood flow; estimate intracranial pressure using said regression model based on said optical interferometric measurements relating to cerebral blood flow andone or more features of extracerebral blood flow or extracerebral blood pressure. 19.The system according to claim 18, wherein said one or more circuits are configured to estimate intracranial pressure comprising unsupervised feature selection. 20.The system according to claim 19, wherein said unsupervised feature selection comprises a lower dimensional matrix decomposition of an input data matrix. 21.The system according to claim 20, wherein said lower dimensional matrix decomposition is a principal component analysis. 22.The system according to claim 18, wherein said one or more circuits are configured to estimate intracranial pressure comprising a representation learning neural network. 23.The system according claim 22, wherein said representation learning neural network is an autoencoder network.

24. The system according to claim 22, wherein said representation learning neural network comprises a random forest regressor. 25.The system according to claim 18, wherein said one or more circuits are configured to estimate intracranial pressure comprising a sequence-to-sequence regression model approach. 26.The system according to claim 25, wherein said sequence-to-sequence regression model approach is implemented as a Long Short-Term Memory (LSTM) neural network. 27.The system according to claim 18, wherein said one or more circuits are configured to generate optical interferometric measurements using a plurality of sensors and / or a camera sensor.28.The system according to claim 16, wherein said one or more features of extracted from a measure of extracerebral blood flow is an estimate of arterial blood pressure (ABP) data. 29.The system according to claim 16, wherein said one or more circuits are configured to use near-infrared wavelengths for generating said optical interferometric measurements. 30.The system according to claim 18, wherein said one or more circuits are configured to vary near-infrared wavelengths in a sweep pattern for generating said optical interferometric measurements. 31.The system according to claim 18, wherein said one or more circuits are configured to select short time of flight optical interferometric measurements to estimate said one or more features of extracerebral blood flow. 32.The system according to claim 18, wherein said one or more circuits are configured to select long time of flight optical interferometric measurements to estimate said plurality of features of said cerebral blood flow.

33. The system according to claim 18, wherein supervised learning is used for estimating said intracranial pressure.

34. The system according to claim 33, wherein a random forest regressor is used for said supervised learning.

Images