Mask providing positive airway pressure and measuring changes in blood pressure

Abstract

A wearable mask for monitoring blood pressure values and providing positive pressure airflow to a patient is provided. The mask includes a first sensor (24a, 24b) configured to measure a time-dependent optical waveform from a first region below the first portion of the wearable mask, where the time-dependent optical waveform includes a first pulse. A second sensor (30a, 30b, 32a, 32b) measures a time-dependent impedance waveform from a second region adjacent the mask, wherein the time-dependent impedance waveform includes a second pulse. A microprocessor (14) is attached to the wearable mask and is configured to 1) receive a digital representation of both the first and second pulses, 2) process the digital representation to determine a time difference between the first and second pulses or a parameter calculated therefrom, and 3) process the time difference between the first and second pulses or a parameter calculated therefrom to determine a blood pressure value.

Description

Cross-reference to related applications

[0001] This patent application claims priority to U.S. Provisional Application No. 63 / 528,220, filed July 21, 2023, pursuant to section 119(e) of the U.S. Patent Act, the contents of which are incorporated herein by reference. Technical Field

[0002] This invention relates to the field of devices that provide positive airway pressure (“PAP” herein) using a wearable mask, and to the field of patient physiological monitoring. Background Technology

[0003] Obstructive sleep apnea (“OSA” in this article) is a sleep disorder characterized by repetitive partial or complete obstruction of the upper airway during sleep, resulting in pauses in breathing. These pauses, known as breathing apneas, can occur multiple times throughout the night and last from seconds to minutes. OSA is often accompanied by disrupted sleep patterns characterized by loud snoring, wheezing, and / or choking. It affects people of all ages and is most common in overweight individuals and those over 40 years of age. In 2021, approximately 22 million Americans had OSA.

[0004] From a physiological perspective, during an OSA patient's upper airway is obstructed and breathing is interrupted, resulting in a reduction in the amount of oxygen entering their lungs. This leads to a gradual decrease in oxygen saturation (SpO2) levels. During this period, the patient's blood pressure (BP) may temporarily rise due to the body's response to hypoxia and the resulting activation of the sympathetic nervous system. This response triggers vasoconstriction (narrowing of blood vessels) and an increase in heart rate (HR), resulting in a temporary increase in BP. Simultaneously, the patient's stroke volume (SV) begins to increase, which represents the amount of blood flowing into the aorta with each contraction of the left ventricle. The sympathetic response leads to increased cardiac contractility, resulting in a temporary increase in SV.

[0005] As the apnea event progresses and continues, the patient's blood carbon dioxide ("CO2") level increases, triggering a compensatory response. The body's chemoreceptors sense the elevated CO2 level and signal the brain to increase the respiratory rate ("RR"). As a result, vasodilation (widening of blood vessels) occurs in response to the increased CO2, followed by a decrease in blood pressure and an increase in respiratory velocity (SV).

[0006] Once the apnea event ends and normal breathing resumes, patients typically experience a sudden fluctuation in blood pressure (BP) due to the release of stored stress hormones (such as adrenaline) as the body recovers from the hypoxia experienced during the event. This, in turn, can lead to vasoconstriction, an increase in heart rate (HR), and a decrease in cardiac output (SV). Cardiac output (CO, SV) is also a factor. The product of HR remains relatively constant during this period because an increase in HR suppresses a decrease in SV.

[0007] During sleep apnea, particularly during BP and SV, such fluctuations in physiological parameters can stress a patient’s cardiovascular system and increase the risk of disease-state hypertension (Hypertension), congestive heart failure (CHF in this article), heart attack, stroke, and other cardiovascular diseases.

[0008] Known PAP treatments include continuous positive airway pressure (“CPAP”) and variable airway pressure. In CPAP, a constant positive pressure is provided to the patient’s airway to open it with a splint. In variable airway pressure, the pressure provided to the patient’s airway varies with the patient’s respiratory cycle. This therapy is typically administered to the patient at night while they sleep. PAP treatment involves the patient wearing a special mask designed to deliver pressurized air during sleep. The PAP device is connected to the mask via a flexible hose and delivers a constant flow of pressurized air to keep the patient’s airway open. The pressurized air prevents soft tissue collapse in the throat, allowing uninterrupted breathing throughout the night. PAP devices typically have pressure settings indicated by the severity of the patient’s OSA level; patients usually adjust the pressure settings as their PAP therapy is improved, typically under the guidance of a healthcare professional. PAP devices are also used for purposes other than managing OSA. For example, PAP devices are used to supplement or replace ventilation, manage CHF, stroke, or Cheyne-Stokes breathing.

[0009] PAP therapy offers numerous benefits to patients with OSA. The positive air pressure it provides keeps the patient's airway open during sleep, effectively reducing or eliminating snoring and sleep apnea events. Ultimately, this leads to improved sleep quality and daytime alertness. PAP therapy helps alleviate symptoms such as excessive daytime sleepiness, morning headaches, and cognitive impairment, as well as improve the aforementioned harmful cardiovascular conditions associated with untreated OSA.

[0010] While PAP therapy is highly effective, it does require consistency and adjustment to ensure optimal results. Some people may find wearing the mask initially uncomfortable or experience dryness in their nose or throat. However, modern PAP devices are designed to be quieter, smaller, and more user-friendly than their predecessors. Various mask styles and sizes are available to accommodate different preferences and facial structures. Regular follow-up with your healthcare provider is important for monitoring progress, addressing any issues, and making necessary adjustments to the PAP setup or device to ensure optimal treatment benefit.

[0011] Typical PAP devices use internal sensors to measure a patient's relative risk (RR) by monitoring airflow from the patient to the machine. However, these devices lack sensors for measuring other vital signs (such as heart rate and SpO2) and more complex "hemodynamic" parameters (such as serovolemia and coagulation).

[0012] Beyond positive airway pressure devices, there are many types of patient monitors that operate in hospitals and at home. For example, such monitors use electrodes worn on the torso to measure electrocardiogram (ECG) and impedance breathography (IPG) waveforms, from which HR, HR variability (HRV), and RR are calculated. Most conventional monitors also utilize sensors typically clipped to the patient's finger or earlobe to measure light signals, known as photoplethysmography (PPG) waveforms. Algorithms associated with such sensors can calculate SpO2 and pulse rate (PR) from the PPG waveform. More advanced monitors can also measure BP, specifically systolic blood pressure (SYS), diastolic blood pressure (DIA), and mean (MAP) BP. This measurement is typically performed using cuff-based techniques called oscillometric or auscultatory techniques, or pressure-sensitive catheters inserted into the patient's arterial system, known as arterial lines. Digital stethoscopes, which can be portable and wearable devices, can measure phonocardiogram (PCG) waveforms that indicate heart sounds and murmurs.

[0013] Some patient monitors are worn entirely on the body. These typically take the shape of a patch that measures ECG, HR, HRV, and in some cases, RR. Such patches may also include accelerometers that measure motion waveforms along the x, y, and z axes (“ACC” in this context). Algorithms can process the ACC waveforms to determine the patient’s posture, degree of motion, falls, and other motion-related parameters. Patients typically wear these types of patches in hospitals or alternatively in outpatient and home settings. Patients are usually worn for relatively short periods (e.g., from a few days to a few weeks) and are primarily used as cardiac event monitors to detect life-threatening arrhythmias. They typically include a wireless transceiver based on technologies such as Bluetooth® or Wi-Fi to transmit information over short distances to a secondary “gateway” device. The gateway device typically includes cellular and / or Wi-Fi radios to transmit information to a cloud-based system.

[0014] More sophisticated patient monitors use invasive sensors called Swan-Ganz or pulmonary artery catheters to measure hemodynamic parameters such as SV, CO, and cardiac wedge pressure. To perform these measurements, these sensors are placed in the left side of the patient's heart, where they are "wedged" into small pulmonary vessels using a balloon catheter. As an alternative to this highly invasive measurement, patient monitors can use non-invasive techniques such as bioimpedance and bioresponse to measure similar parameters. These methods deploy networked wearable electrodes (typically deployed in the patient's chest, legs, and / or neck) to measure IPG and / or bioresponse (“BR” in this paper) waveforms. Analysis of the IPG and BR waveforms yields SV, CO, and pleural impedance, which represent the fluid in the patient's chest (“FLUID” in this paper). Notably, IPG and BR waveforms typically have similar shapes and are sensed using similar measurement techniques, and are therefore used interchangeably throughout this paper.

[0015] Remote patient monitoring (“RPM” in this article) refers to the use of digital technologies to collect patient health data and transmit it from patients in their homes or other non-clinical settings to healthcare providers for monitoring and management. Typical RPM patients include those with chronic diseases that frequently lead to hospital readmissions, such as CHF, chronic obstructive pulmonary disease (“COPD” in this article), hypertension, and diabetes.

[0016] Unsurprisingly, there is significant overlap between patients with OSA and those with these chronic diseases. For example, in 2021, approximately 6.2 million Americans had CHF; estimates suggest that about 50-70% of these patients may also have OSA. The link between these conditions is bidirectional, as OSA can contribute to the development and progression of CHF, while CHF can exacerbate the severity of OSA. Similar situations exist for COPD (approximately 10-20% of OSA patients also have COPD), hypertension (30-70%), and diabetes (40-70%). Often, the coexistence of chronic diseases and OSA is of significant importance for patient management. OSA in these individuals can worsen symptoms, increase the risk of cardiovascular or pulmonary events, and affect treatment outcomes. Therefore, proper assessment and management of OSA in patients with chronic diseases is crucial for optimizing their overall health and quality of life.

[0017] In recent years, RPM has gained significant attention and recognition due to its potential to improve healthcare outcomes and reduce costs. The Centers for Medicare & Medicaid Services (“CMS”) – the federal agency responsible for managing U.S. healthcare programs – has recognized the value of RPM and implemented reimbursement policies to support its use.

[0018] When certain criteria are met, CMS reimburses remote patient monitoring services under Medicare plans. In 2019, CMS introduced new codes and guidelines that allow healthcare providers to bill RPM services, making it easier for providers to receive reimbursements for remote patient monitoring. CMS reimbursement typically includes initial setup and patient education, as well as ongoing patient data monitoring and interpretation. It includes activities such as using connected devices to collect physiological data (e.g., BP, HR) and the time spent by healthcare professionals reviewing and analyzing the transmitted data.

[0019] CMS-funded RPM offers several benefits. It encourages healthcare providers to adopt and implement RPM techniques, facilitating remote management of chronic disease and post-acute care, and reducing the need for in-person visits. This not only improves patient access to care, particularly in rural or underserved areas, but also enhances patient engagement and empowers individuals to take an active role in managing their own health. By enabling earlier detection of health problems, remote patient monitoring can help prevent complications, reduce hospital readmissions, and ultimately lead to better patient outcomes. Summary of the Invention

[0020] Based on the above, it would be beneficial to combine the therapeutic benefits of positive airway pressure (“PAP”) therapy (e.g., CPAP), variable pressure therapy (e.g., CFlex or BiFlex), or bilevel positive airway pressure therapy with monitors that measure vital signs and hemodynamic parameters, such as during RPM. For example, an ideal system would feature a positive airway pressure device and a mask that together deliver treatment to OSA patients while simultaneously measuring vital signs, hemodynamic parameters, and time-dependent physiological waveforms; this information can characterize other chronic conditions, such as CHF, COPD, hypertension, and diabetes, individually or in combination. Furthermore, providing these types of data can further engage patients in their PAP treatment, thereby improving adherence. This, in turn, enhances the efficacy of PAP treatment. Additionally, incorporating patient monitoring capabilities directly within the PAP mask allows for the identification of parameters such as awakenings during sleep, apnea events, inappropriate mask donning, mask deterioration, and other physiological and mask-related parameters. It can also indicate whether the mask is unsuitable for the patient, is leaking, or needs replacement. Understanding these issues allows clinicians to adjust the patient's PAP treatment, thereby increasing its efficacy and improving patient health.

[0021] This invention effectively achieves these objectives using an ensemble of embedded electronic sensors coupled to a control system, each of which is directly integrated within the PAP mask. The control system provides power and includes analog and digital electronics that, during use, process the signals measured by the sensors to determine vital signs, hemodynamic parameters, and time-dependent waveforms. It also includes a data transmission system that sends raw and processed information to an external patient management system. Ultimately, this facilitates connecting OSA patients to their clinicians. Taken together, these systems provide PAP treatment to patients while allowing clinicians to remotely monitor patients, for example, according to RPM (Recovery Per Moment).

[0022] More specifically, the present invention provides a “smart mask” (“SMK” herein) comprising one or more of the following sensors, each directly embedded in the mask and configured to directly measure signals from (or near) the patient’s face: 1) an optical sensor for measuring PPG waveforms and thereby measuring SpO2 and PR; 2) an impedance sensor for measuring IPG / BR waveforms and thereby measuring SV, CO, and FLUID; 3) an EMG sensor for measuring muscle activity and EMG waveforms from the patient’s face and thereby measuring arousal-related muscle activity; 4) a flow sensor for measuring respiratory rate, temperature, humidity, and volatile organic compounds (“VOC” herein) and volatile sulfur compounds (“VSC” herein) from the patient’s breathing and thereby measuring RR; 5) a digital microphone for measuring sounds emitted by the patient, such as snoring, wheezing, and coughing; and 6) an accelerometer for measuring ACC waveforms along the x, y, and z axes and determining the patient’s posture and degree of movement from these.

[0023] Additionally, algorithms operating on the microprocessor within the SMK can further analyze PPG, ECG, and IPG / BR waveforms to determine SYS, MAP, or DIABP, as described in more detail below. In an embodiment, a battery-powered control system located directly on the SMK controls each of the aforementioned sensors to perform its corresponding measurements. Wired and wireless transmitters in the control system send information to an external gateway, which then forwards the information to the cloud. There, ML and AI-based algorithms process the SMK measurements to determine parameters such as wake-up, apnea events, physiological information for RPM, the SMK's cooperation with the patient (if the SMK needs to be replaced), and other parameters. By providing this information, the SMK delivers PAP treatment while monitoring the patient. It helps inform remote clinicians about the patient's physiological status and the progress of PAP treatment. Ultimately, this information can help connect and stimulate dialogue between clinicians and patients, thereby improving patient adherence to PAP treatment.

[0024] In one aspect, the present invention provides a system for monitoring blood pressure (BP) values ​​from a patient. The system is characterized by a wearable face mask (e.g., SMK) coupled to a PAP machine that delivers air to the patient at positive pressure. The wearable face mask includes a first sensor directly connected thereto, such as an optical sensor configured to measure a time-dependent optical waveform (e.g., a PPG waveform) from a first region below a first portion of the wearable face mask, the time-dependent optical waveform including a first pulse. The face mask also includes a second sensor directly connected thereto, such as an impedance sensor, which measures a time-dependent impedance waveform (e.g., an IPG and / or BR waveform) from a second region adjacent to the wearable face mask, the time-dependent impedance waveform including a second pulse. A processing system characterized by a microprocessor is attached to the wearable face mask and configured to: 1) receive digital representations of both the first and second pulses; 2) process the digital representations to determine a time difference between the first and second pulses, or parameters calculated therefrom; and 3) process the time difference between the first and second pulses, or parameters calculated therefrom, to determine the BP value.

[0025] In this embodiment, the optical sensor includes first and second light sources and a photodiode, wherein the first light source may be a first light-emitting diode (“LED”) that emits light radiation near λ=660nm, and the second light source may be an LED that emits light radiation near λ=940nm. “Near” means that the LED emits radiation within approximately ±20nm of the target emission wavelength.

[0026] In an embodiment, the impedance sensor includes at least one sensing electrode and at least one driving electrode. Both electrodes are typically characterized by conductive materials selected from rubber, polymers, fabrics, metals, wires, meshes, and hydrogels; other conductive materials may also be used. The driving electrode is configured to inject a first current into a second region, wherein the current is modulated at a frequency in the range of 5-500 kHz and has an amplitude in the range of 0.01-5 mA. In an embodiment, the first sensing electrode and the driving electrode are connected to a first side of the wearable mask. Typically, on the opposite side, the SMK includes a second sensing and driving electrode connected to a second side. Here, the second driving electrode injects a modulated current having an amplitude similar to the current injected by the first driving electrode, the second current being modulated at a frequency approximately 90° different from the frequency corresponding to the first current.

[0027] In an embodiment, to determine the BP value, the processing system determines a "foot" for the first pulse and a foot for the second pulse, where the foot is located near the minimum amplitude of the pulse, just before it begins to rapidly increase in amplitude. Typically, to determine the time difference between the first and second pulses, the processing system calculates the time interval between at least one of the following characteristics (also referred to as a "reference") of the first and second pulses: foot, maximum slope of the pulse's rising edge, base, peak value, and rising edge. The processing system then calculates the reciprocal of the time difference value and a calibration value to determine the BP value. In an embodiment, the algorithm may process digitized pulses to aid in identifying the aforementioned reference. Such an algorithm includes digital filtering, Fourier analysis (or associated frequency domain transformation), smoothing, and / or taking one or more derivatives of the pulse. The derivatives may be single-point or multi-point derivatives.

[0028] In other embodiments, the wearable mask includes a control system, which is typically integrated directly into the wearable mask, and in some embodiments, the control system is located near the top of the wearable mask. The control system includes a processing system. A microprocessor is located within the processing system and runs computer code that determines blood pressure values. The computer code can be written in one or more languages, such as C, C++, assembly, Java, JavaScript, and Python. A rechargeable battery powers the system, which is electrically connected to first and second sensors via a flexible cable extending alongside the mask.

[0029] On the other hand, the present invention provides another type of mask that performs the functions described above. For example, in addition to providing CPAP therapy, the mask can also be used with a variable pressure device (e.g., a BiPAP device manufactured and sold by Philips Respiratory Systems, Inc., Mullisville, Pennsylvania). ® Bi-Flex ® Or C-Flex ™ (Equipment) can be used together. Alternatively, the mask can be used with a respirator, oxygen mask, fighter pilot mask, goggles, face covering, scuba mask, decorative mask (such as a mask used in Hollywood), or other types of mask worn near the face.

[0030] On the other hand, the present invention provides a system for monitoring blood pressure (BP) values ​​from a patient, characterized by a wearable face mask (e.g., SMK) comprising an optical sensor and an impedance sensor, both sensors being in contact with the patient's face. The optical sensor measures a first time-dependent optical waveform (e.g., PPG waveform) including a first pulse, and the impedance sensor measures a second time-dependent impedance waveform (e.g., IPG waveform) including a second pulse. A processing system processes the time difference between the first and second pulses, or parameters calculated therefrom, to determine the BP value.

[0031] These and other advantages of the invention should be apparent from the following detailed description and claims. Attached Figure Description

[0032] Figure 1A This is a schematic diagram showing the front view of the SMK according to the present invention;

[0033] Figure 1B yes Figure 1A SMK's photos;

[0034] Figure 2 It is worn by the patient Figure 1A and Figure 1B The diagram illustrates how SMK transmits patient-measured information (such as vital signs, hemodynamic parameters, and time-dependent waveforms) to a cloud-based software system and from the cloud-based software system to a third-party software system via gateway devices and / or mobile devices.

[0035] Figure 3A , Figure 3B and Figure 3C They are combined in Figure 1A and Figure 1B A schematic diagram of a reflective optical sensor, a 4-electrode impedance sensor, and a pressure sensor used in the SMK to measure signals from the patient;

[0036] Figure 3D This is a schematic diagram of a transmission multi-wavelength spectrometer incorporated into the tube connecting the PAP device to the SMK and configured to measure the spectrum of the patient's breath within the tube.

[0037] Figure 3E yes Figure 3A , Figure 3B and Figure 3C Photos of the sensor;

[0038] Figure 4A and Figure 4B These refer to the front and rear surfaces of a printed circuit board (hereinafter referred to as "PCB") and its constituent parts. Figure 1A and Figure 1B A top view of the electrical components of the control system within the SMK;

[0039] Figure 5A and Figure 5B They are shown respectively Figure 1A and Figure 1B The SMK control system connects the SMK to the rotary joint of the PAP device, as shown in the top and side mechanical views.

[0040] Figure 5C and Figure 5D They are respectively Figure 4A and Figure 4BPhotos of the PCB inside the housing integrated into the SMK control system and the PCB outside the housing;

[0041] Figures 6A-6F It is a graph of time-dependent waveforms measured from the patient using SMK, characterized by ECG waveforms ( Figure 6A ), PPG waveform ( Figure 6B ), IPG waveform ( Figure 6C ), pressure waveform ( Figure 6D ), EMG waveform ( Figure 6E ) and ACC waveform ( Figure 6F );

[0042] Figure 6G It shows the measurement Figures 6E-6F The diagram shows the placement of the SMK in the time-dependent waveform.

[0043] Figure 7A and Figure 7B These are graphs of time-dependent IPG waveforms measured from the face of patients using SMK and from the chest of patients using a close-fitting patch, respectively.

[0044] Figure 7C and Figure 7D These are shown separately for measurement. Figure 7A SMK of IPG waveform and used for measurement Figure 7B A schematic diagram of the arrangement of wearable patches with IPG waveforms;

[0045] Figure 8A and Figure 8C The graphs show the time-dependent PPG waveforms measured from the patient's wrist using an optical sensor within a standard "smartwatch" during normal breathing, wheezing, coughing, and apnea, and the time-dependent pressure waveforms measured simultaneously from the patient's face using a pressure sensor within an SMK.

[0046] Figure 8B and Figure 8D These are shown separately for measurement. Figure 7A and Figure 7C A schematic diagram showing the placement of a time-dependent waveform optical sensor and a pressure sensor;

[0047] Figure 9 This is a table showing the accuracy and area under the curve (AUC) values ​​corresponding to different machine learning (ML) models used to process data similar to those generated by SMK and used to determine awakening in OSA patients;

[0048] Figure 10 This is a subject operating characteristic (ROC) plot, which plots the results for... Figure 9The table shown illustrates the relationship between the true positive rate and the false positive rate of the data processed using Model 8.

[0049] Figure 11A and Figure 11B These are mechanical views of the bedside hub, which is connected to the SMK to charge its internal battery and download data stored on its internal flash memory for display and analysis.

[0050] Figure 11C It is connected to the SMK and PAP machines by Figure 11A and Figure 11B A photograph of a bedside hub shown in the mechanical diagram;

[0051] Figures 12A-12F It operates on the bedside hub and displays the standard clock separately. Figure 12A Part 1 of the survey of OSA patients wearing SMK ( Figure 12B ), time-dependent ECG waveforms measured by SMK ( Figure 12C The second part of the investigation (12D), the third part of the investigation, and the instruction (12E) containing the physiological data have been successfully uploaded to the cloud, as well as the time-dependent IPG waveform measured by SMK. Figure 12F Screenshots of the graphical user interface ("GUI" in this article);

[0052] Figure 13A This is a schematic diagram illustrating the electrode positions on a patient wearing an SMK for measuring ECG waveforms according to an alternative embodiment of the present invention;

[0053] Figure 13B It is based on Figure 13A The electrode positions shown are measured from the patient and then wirelessly transmitted to a time-dependent ECG waveform, which is then displayed on a mobile device.

[0054] Figure 14A This is a schematic diagram of an alternative embodiment of the present invention, characterized in that a transmission multi-wavelength optical spectrometer is incorporated into a tubing that connects the PAP device to the SMK and measures the spectrum of the patient’s respiration within the tubing;

[0055] Figure 14B Is Figure 14A A photograph of the chip-scale multiwavelength spectrometer (ASMAS7341) used in an alternative embodiment of the present invention is shown.

[0056] Figure 14C It is a graph of a series of frequency-correlation measurement bands, each band being controlled by... Figure 14B The different software register activations shown in the chip-scale multiwavelength spectrometer; and

[0057] Figure 15 This illustrates the use of a conventional optical spectrometer (Thorlabs, shown on the vertical axis on the left side of the figure) and Figure 14B A graph of frequency-correlated absorption spectra measured from a sample using a chip-scale multiwavelength optical spectrometer (AS7341, shown on the right side of the figure). Detailed Implementation

[0058] As used herein, the singular forms “a,” “an,” and “the” include plural references unless the context clearly specifies otherwise. As used herein, the expression “coupled” means that, wherever a link occurs, these parts are directly or indirectly (i.e., through one or more intermediate parts or components) connected or operate together. As used herein, “directly coupled” means that two elements are in direct contact with each other. As used herein, “fixedly coupled” or “fixed” means that two components are coupled so as to move as a single component while maintaining a constant orientation relative to each other.

[0059] As used herein, the term “monolithic” means that a component is manufactured as a single piece or unit. That is, a component that includes parts manufactured separately and then coupled together as units is not a “monolithic” component or body. As used herein, the expression “joined” between two or more parts or components means that the parts exert force on each other directly or through one or more intermediate parts or components. As used herein, the term “number” means an integer of one or more (i.e., a plurality).

[0060] Directional terms used herein, such as, but not limited to, top, bottom, left, right, up, down, front, back, and their derivatives, refer to the orientation of the elements shown in the accompanying drawings and do not limit the claims unless expressly stated herein. 1. System Overview

[0061] Figure 1A and Figure 1B Mechanical drawings and photographs of the SMK 10 according to the present invention are shown respectively. The SMK 10 uses sensor sets 12a and 12b to measure vital signs, hemodynamic parameters, and time-dependent waveforms from a patient. The first portion of the embedded sensor set 12a is located on the left-hand side of the SMK 10; the second portion 12b is on the right-hand side. Located near the top 37 of the SMK 10 and primarily composed of… Figure 4A and Figure 4BThe control system 14, consisting of multiple PCBs, is encapsulated in a mechanical housing 13, which wraps around a rotary connector 18. A rechargeable lithium-ion (“Li:ion”) battery 16 powers the SMK 10. The rotary connector 18 includes an opening 19 for connecting the SMK 10 to a flexible hose, which in turn is connected to a remote PAP device (e.g., in…). Figure 11C The pressurized air generated (as shown in the diagram) is supplied to the patient via SMK 10.

[0062] In summary, the sensors within the SMK 10 measure time-dependent waveforms from the patient's face and cheeks, for example... Figures 6A-6F The waveform is shown. For this purpose, flexible ribbon cables 34a and 34b, which contain a group of wires encapsulated in a silicone coating, are connected to the side of the control system 14. Figure 3E , Figure 5C and Figure 5D The photographs of the control system and sensors shown depict ribbon cables in more detail. These cables span the length of the left tube 39a and right tube 39b of the SMK 10 and are electrically connected to sensors arranged along these tubes. The left tube 39a and right tube 39b of the SMK are typically flat, flexible, and hollow silicone structures; they connect to a “pad” or “mouthpiece” component 11 that is attached to the patient’s mouth and nose during sleep to actively supply pressurized air.

[0063] A pair of reflective optical sensors 24a and 24b, positioned on the left tube 39a and right tube 39b, measure time-dependent PPG waveforms from the patient's cheeks. PPG waveforms are particularly suitable for this measurement because they represent a region characterized by a dense capillary bed. The first optical sensor 24a is connected to a wire in a ribbon cable 34a on the left side of the mask; the second optical sensor 24b is connected to a wire in a ribbon cable 34b on the right side. (Reference) Figure 1A and Figure 3A Each optical sensor 24a, 24b includes a light source 60, which is typically characterized by one or more LEDs or diode lasers. The light source 60 typically includes a first LED operating in the red spectral region (e.g., λ = 660 nm) and a second LED operating in the infrared spectral region (e.g., λ = 940 nm); such wavelengths are ideal for SpO2 measurement, as is known in the art.

[0064] During the measurement, individual LEDs within the light source sequentially emit beams of radiation (schematically indicated by arrow 64) into the patient's muscles, where they are partially absorbed by the blood-filled capillaries within. The capillaries expand and contract with the pulsating blood flow driven by each heartbeat, thus modulating the absorptivity of the incident red and infrared radiation according to Beer's Law (which depends on the path of the incident radiation and is therefore related to the vessel diameter). After passing through the capillaries, a portion of the modulated radiation (schematically indicated by arrow 66) illuminates a photodetector 62, where it is optically absorbed to generate a photocurrent. This photocurrent is further processed (e.g., filtered and then amplified) to produce PPG waveforms independently corresponding to the red and infrared spectral regions.

[0065] Typically, to collect these waveforms, the control system 14 includes a chip-level analog front-end (“AFE” herein, such as the analog device MAX86176) connected to the light source 60 and photodetector 62, controlling them to sequentially emit red and infrared radiation, and then measuring the corresponding PPG waveforms. Notably, the MAX86176 includes a separate AFE for measuring both PPG and ECG waveforms. An algorithm operating on a microprocessor (“CPU” herein) within the control system 14 receives and processes digital versions of the waveforms, and uses digital signal processing to generate representations of high-frequency pulsating PPG elements (typically referred to as “RED”). AC "and IR" AC ") and low-frequency static PPG elements (commonly referred to as "RED") DC "and IR" DC The signal component is ">". This process ultimately produces SpO2 and PR values, as known in the art.

[0066] Optical sensors 24a and 24b positioned on each side of the SMK 10 provide redundancy and increase the likelihood that the SMK 10 can measure effective PPG waveforms, as well as SpO2 and PR values, from the patient. For example, optical sensors 24a and 24b will likely have good contact with the patient's face and cheeks, thus producing good measurements when the patient is lying supine with their head upright and experiencing minimal movement. In an embodiment, SpO2 and PR measurements obtained from the left and right cheeks can be averaged together to increase measurement accuracy. However, for patients lying on their side or moving, one optical sensor (e.g., a sensor between their cheek and the pillow) may have good contact with the patient's cheek and produce good measurements, while the opposing sensor (e.g., an upright sensor) may produce poor contact and produce subparallel measurements. In this configuration, it is ideal to use the optical sensor that produces good measurements to characterize the patient. Such an embodiment requires the CPU to run an algorithm that evaluates the PPG waveform to determine its signal quality.

[0067] In other embodiments, the heartbeat-induced pulses in PPG waveforms measured from the left and right cheeks can characterize the temporal differences between their pulsatile features. Typically, the time difference is measured from the base of each pulsatile feature and referred to as pulse arrival time (“PAT”), pulse transit time (“PTT”), or vessel transit time (“VTT”). Previous studies have shown that PAT, PTT, and VTT are indirectly related to blood pressure, such as SYS, MAP, and DIA. Figures 6A-6C These time components, also known as “contraction time intervals,” are shown in more detail.

[0068] The relationship between PAT, PTT, VTT, and BP is influenced by various factors, including arterial compliance, the distance between measurement sites, and the stiffness of the patient's arterial wall. A linear equation describes this relationship. in It is a patient-specific slope describing the relationship between the reciprocals of PAT, PTT, and / or VTT and BP. These are the initial calibration values. The equation is the same for SYS, DIA, and MAP, and the constants of the equation ( and It depends on the specific BP value being measured.

[0069] In the relevant embodiments, the presence of two optical sensors 24a, 24b means that the SMK can simultaneously measure PPG waveforms from both sides of the patient's face. The time difference of the heartbeat-sensing pulses within these waveforms can indicate certain aspects of the patient's physiology. This measurement is equivalent to that described in U.S. Patents 7,803,120 and 9,622,710 to Banet et al., both of which describe the measurement of "bilateral pulse transit time" ("BPTT" herein) and are incorporated herein by reference. In these documents, BPTT measurements are performed using optical sensors positioned on the patient's finger. They describe how, in BPTT measurements, the asymmetrical position of the heart, coupled with the assumption that blood pressure follows equal lengths along the left and right vascular pathways in the body, means that the PTT of the right pathway may be slightly longer than that of the left pathway. When used with calibrated measurements, this time difference (BPTT, similar to the systolic time interval of PAT, PTT, and VTT) can be used to estimate BP, such as SYS, MAP, and DIA.

[0070] Typically, the time difference measured by BPTT is similar to that of PAT, PTT, and VTT, for example, between approximately 10–200 ms. Very long BPTTs, such as >250 ms, or examples where the shape of the pulses caused by heartbeats measured with different optical sensors differs significantly, can indicate that the arteries on one side of a patient's neck and / or face (e.g., the carotid artery) differ from those on the other side. For example, this difference might arise if one artery has significant plaque accumulation compared to another. In this way, SMK can be used as a screening tool to estimate whether a patient has significant plaque accumulation in one of their carotid arteries.

[0071] Reference Figure 1A and Figure 3B Electrodes 30a, 30b, 32a, and 32b are typically constructed of flexible conductive materials (such as conductive rubber, fabric, mesh, or textiles) and connected to wires within ribbon cables 34a and 34b, measuring weak bioelectrical signals during sleep. As the heart contracts and relaxes, the body's impedance changes over time; this, in turn, modulates the bioelectrical signals. More specifically, during cardiac systole (when the heart contracts), the IPG waveform shows an initial increase in impedance due to the decrease in blood volume in the thoracic cavity. This increase then rapidly decreases as blood is ejected into the systemic circulation. During cardiac diastole (when the heart relaxes), as blood returns to the thoracic cavity, the waveform shows a gradual increase in impedance.

[0072] To measure the IPG waveform from patient 15, the SMK 10 includes a pair of "sensing" electrodes 30a, 30b and a pair of "driving" electrodes 32a, 32b, which together perform bioimpedance and bioresponse measurements. The driving electrodes 32a, 32b inject a high-frequency (e.g., 5–500 kHz, and typically about 70 kHz), low-ampere (e.g., 0.1–4 mA, and typically about 1 mA) current into the patient's cheek; this... Figure 3B The arrows 85a and 85b represent this. As shown in the figure, the current injected by each drive electrode typically features a sinusoidal (or alternatively, a square wave) profile, and is 90°. o The sensors 30a and 30b detect weak bioelectrical signals from each cheek; this is indicated by arrows 83a and 83b in the diagram. Bioelectrical impedance is measured by determining the resistance (i.e., the resistance to the flow of injected current) and reactance (the ability to store and release electrical energy generated by the injected current) in the tissue. Analyzing the voltage changes that occur when current passes through the cheek generates bioimpedance and bioresponse values ​​that appear as time-dependent waveforms (e.g., IPG and BR waveforms). As with the optical measurements described above, bioimpedance and bioresponse measurements are typically managed using a chip-level AFE (e.g., an analog device MAX30009) arranged on the control system 14.

[0073] In one embodiment, the electrodes are conductive fabrics or textiles directly integrated (e.g., sewn into) the tubes of a mask that supplies pressurized air to the patient.

[0074] A BR waveform is a type of IPG waveform with a similar morphology, but it represents a time-dependent phase change caused by physiological events in the tissue beneath the electrodes (e.g., blood flow caused by a heartbeat). This change is due to the capacitance and inductance in the tissue and is generally weaker than the changes in thoracic volume that produce the IPG waveform. BR waveforms can have several improvements over IPG waveforms, primarily in their less sensitivity to noise and external sources, and are characterized by a higher percentage of AC signal components relative to DC signal components, as described in more detail below.

[0075] Impedance measurement systems similar to the MAX30009 measure time-dependent impedance waveforms, which are very similar to the PPG waveform described above, characterized by an AC component (i.e., the pulsating component) and a DC component (i.e., the baseline). Algorithms can process the AC and DC components together to determine SV and CO, and process the DC component separately to estimate FLUID in the patient. SV, CO, and FLUID represent hemodynamic parameters and can be good predictors of chronic diseases, particularly CHF. Examples of such algorithms are the Sramek-Bernstein and Kubicek equations, both of which estimate SV based on the relationship between changes in thoracic impedance and the rate of change in blood volume during systole. These two equations are described in detail in U.S. Patent 11,129,537 to Banet et al., the contents of which are incorporated herein by reference.

[0076] In most cases, ECG waveforms cannot be measured with a good signal-to-noise ratio from a location above the patient's neck (e.g., the face). Therefore, sensing electrodes 30a and 30b typically do not produce ECG waveforms sufficient for calculating HR, RR, etc. This is because the biopotential signals measured with electrodes on opposite cheeks are extremely similar, and therefore, when processed with a differential amplifier present in conventional ECG circuits (e.g., the differential amplifier within the ECG AFE in the MAX86176), the resulting signal has virtually no amplitude. However, a differential amplifier similar to the one used in ECG circuits can process the signal measured by sensing electrodes 30a and 30b to produce a time-dependent waveform similar to ECG, referred to as an electromyography (EMG) waveform indicating electrical activity generated by skeletal muscle ("EMG" in this context). Figure 6EAn example of such an EMG waveform measured using the MAX86176 is shown. This activity itself manifests as time-dependent pulses in the EMG waveform, which typically indicate the movement of muscles in a patient's face during sleep (e.g., a patient clenching their jaw), and can be incorporated into numerical algorithms for estimating arousal. Here, analog filters embedded with an AFE for impedance, ECG (and by means of EMG) measurements are designed to process the bioelectrical signals measured by sensing electrodes 30a, 30b to simultaneously generate IPG, EMG, and ECG waveforms.

[0077] refer to Figure 1A and Figure 3C To measure the time-dependent pressure within the SMK 10, and more specifically, the time-dependent pressure within the left-hand tube 39a and right-hand tube 39b of the mask, a flexible ribbon cable 34a on the left-hand side of the SMK 10 is connected to a pressure sensor 20 (e.g., a Bosch BME688) located near the padding component 11. Typically, the pressure sensor 20 includes a small opening for detecting the exhalation expelled by the patient 15, such as... Figure 3D Arrow 75 and the "cloud" graphic 76 are schematically indicated. A porous membrane (not shown) can be used to cover the opening to prevent it from contaminating the airway. Based on respiration, pressure sensor 20 detects time-dependent pressure changes in the mask regulated by the patient's breathing pattern. Analysis of these changes yields the patient's respiratory rate (RR). Furthermore, the BME688 sensor can also detect other parameters that may indicate patient compensatory dysfunction; these include humidity, respiratory temperature, VOCs, and VSCs.

[0078] Figure 3E A photograph of the sensor within the aforementioned SMK is shown. The top, middle, and bottom of the photograph show... Figure 3A optical sensors, Figure 3C Digital microphones and pressure sensors.

[0079] In relevant applications, pressure sensor 20 can be coupled to an optical sensor (e.g., an infrared optical sensor) that measures the concentration of carbon dioxide (“CO2” in a patient’s breath. For example, to perform such a measurement, Figure 3D The pressure sensor 20 shown can be used with Figure 3AThe reflective optical sensor 24 shown is coupled; here, an LED (or laser diode) within the optical sensor emits light radiation at wavelengths strongly absorbed by CO2, such as 2, 2.9, and 4.3 micrometers. The concentration of CO2 in the patient's breath modulates the optical absorption at these wavelengths, and when combined with calibration (e.g., calibration performed during the manufacture of the SMK), an absolute value for the gas can be produced. When used in this application, the combined system including pressure and optical measurements effectively functions as a carbon dioxide plethysmography sensor, providing the concentration or partial pressure of CO2 in the breath gas emitted by the patient. In this way, the SMK 10 functions similarly to a sensor that measures tidal volume and end-expiratory CO2 (“et-CO2” herein), measurements of which are typically maintained for hospitalized patients in intensive care units (“ICU” herein). Using the SMK, this measurement can be performed at home.

[0080] Recent studies have shown that such measurements of et-CO2, particularly their variants, can be good predictors of brain natriuretic peptide (“BNP” in this article), a known reliable marker for compensatory dysfunction in patients with CHF (see, for example, Koyama et al.). Technology Applications of Capnography Waveform Analytics for Evaluation of Heart Failure Severity “, J Cardiovasc Transl Res. Dec 2020; 13(6):1044-1054. doi: 10.1007 / s12265-020-10032-5. ePublished May 28, 2020. PMID:32462611). BNP is a hormone produced by the ventricles in response to the stretching of cardiomyocytes and plays a key role in regulating blood pressure and fluid homeostasis. BNP levels typically increase when there is increased pressure on the heart, such as during CHF. Therefore, BNP has been extensively studied as a predictor and diagnostic tool for this chronic condition.

[0081] In a preferred embodiment, as described above, the pressure sensor is a Bosch BME688 sensor or other equivalent. This sensor is configured to measure the following parameters from the patient's respiration: pressure, temperature, humidity, VOC, and VSC. Sensors, alone or coupled with complementary sensors, can also measure other parameters, such as alcohol content, ketones (e.g., acetone and acetylacetone), glucose levels, and other chemicals in the patient's respiration.

[0082] The SMK 10, located near the pad 11, may also include a digital microphone 22 connected to flexible ribbon cables 34a, 34b and measuring sounds emitted by the patient during sleep, such as snoring, coughing, wheezing, and apnea events. When sampled at high rates (e.g., several kHz), the digital microphone 22 can measure full-resolution sounds. Lower sampling rates (e.g., 100–500 Hz) produce downsampled signals, which are used as approximations of these sounds; lower sampling rates are generally desirable for reducing the amount of memory required in the SMK 10. Similar to pressure sensors, the microphone may be covered with a thin film or porous membrane to prevent contamination of the patient's airway.

[0083] Although Figure 1A and Figure 1B Sensors integrated into specific types of PAP masks are shown, but in theory, they can be integrated with any PAP mask, regardless of size or shape. This may require specialized, mask-specific adapters that connect the various sensors described above to flexible components within the mask. For example, sensors can be integrated with the following PAP masks manufactured by Philips / Respironics: Dreamware, DreamWisp, Wisp, ComfortGel, Amara, Pico, Nuance, and ComfortGel. Such masks can be full-face masks or dedicated nasal or oral PAP systems. It should also be noted that sensors can be integrated into any other mask that provides similar functionality to a PAP mask, such as a mask used to provide ventilation to a patient using, for example, a non-invasive ventilator.

[0084] Figure 2 The diagram illustrates how the SMK 10 and its collection of sensors 12a, 12b can monitor a patient 15 suffering from OSA, for example, during sleep. A PAP device 42 is connected to the SMK 10 via a hose 40, as indicated by arrow 56, and provides positive pressure to the patient through the SMK 10 to improve the effectiveness of OSA treatment. It communicates bidirectionally with a cloud 46, as indicated by arrow 54, to send parameters it measures (e.g., RR using internal sensors, pressure values ​​describing the air delivered to the patient, and other patient information) and to receive information related to PAP treatment, such as pressure value settings and patient information. Simultaneously, Figure 1A and Figures 3A-3DThe SMK 10 and various sensors shown and described above measure time-dependent waveforms, vital signs, and hemodynamic parameters from the patient. A Bluetooth® transmitter within the SMK 10 wirelessly transmits a digital version of these signals, or derivatives calculated from them, to an external mobile device 44 (or alternatively, a bedside hub), as indicated by arrow 50. The mobile device 44 can be a mobile phone, tablet, laptop, other computer, server, or wearable device; the bedside hub is typically a custom-designed device, such as those shown in 10A-C.

[0085] Once the mobile device 44 (or hub) receives information from the SMK 10, it sends it to the cloud 46, as indicated by arrow 52. The cloud 46 (e.g., Amazon Web Services, "AWS" in this document) comprises a collection of servers and software systems capable of processing waveforms, vital signs, and hemodynamic parameters measured by the SMK to characterize the patient 15. AWS-like cloud-based systems include sophisticated algorithms and computational libraries, such as ML and AI-based algorithms and libraries, to process the data generated by the SMK 10, thereby producing reports and analyses, such as... Figure 9 and Figure 10 The report and analysis are shown. Such an algorithm can be used to characterize a patient by RPM or to detect physiological events such as awakening and apnea (e.g., characterized by low SpO2 and high HR), which can be guaranteed by changes in settings on PAP device 42. Additionally, mobile device 44 may include a user interface that presents survey-type questions to patient 15, such as... Figure 12A , Figure 12B , Figure 12D , Figure 12E The issues illustrated help engage patients in their PAP treatment and provide information to ML and AI algorithms to help improve patient representation. In embodiments, for example, the user interface can conduct surveys or questionnaires to help determine a patient's cognitive level after using the PAP device; this information can then be relayed back to the PAP device and used to improve the treatment it provides.

[0086] Third-party software systems 43, such as electronic medical records (“EMR” systems in this text), can also communicate bidirectionally with the cloud 46, as indicated by arrow 55. Here, the EMR can provide further information that can inform ML and AI algorithms, such as the patient’s medical history, the medications they are taking, and the results of blood tests, laboratory work, and other medical tests.

[0087] In an embodiment, the PAP device 42 measurements can supplement the parameters measured by the SMK. In an embodiment, information measured by the PAP device 42 and sent to the cloud 46 (e.g., RR, tidal volume, and the flow rate and pressure of the air delivered to the patient), as indicated by arrow 54, can be combined and processed together with the data collected by the SMK. For example, parameters such as tidal volume can be incorporated into ML and AI models (such as those concerning...). Figure 9 and Figure 10 (Those described) to better estimate wake-up. Alternatively, the RR measured by PAP device 42 can be compared with the RR measured by SMK to confirm the accuracy of that particular measurement.

[0088] In other embodiments, Figure 2 The system shown can operate in a closed-loop manner, wherein the SMK measures sleep-related parameters from the patient (e.g., number of awakenings; physiological information (e.g., HR, SpO2, BP, RR, SV, and CO) and sends signals to the PAP device in response, and the PAP device then adjusts parameters related to the pressurized air delivered to the patient (e.g., its pressure and / or flow rate). This closed-loop system can improve the patient's sleep and physiological outcomes.

[0089] In relevant embodiments, the SMK 10 can be integrated with additional sensors that are positioned remotely from the actual face mask but still measure complementary parameters. For example, refer to... Figure 3D A multi-wavelength spectrometer 71, characterized by a broadband light source 72 and a dedicated photodetector 70, can be coupled to a PAP tubing 40 and measures patient exhalation (as indicated by arrow 73) and gaseous compounds 74 (e.g., CO2) that propagate within the tubing 40 during PAP treatment. The tubing is an ideal location for the multi-wavelength optical spectrometer 71 because, when connected to this component, it can operate in a transmission mode geometry; this (compared to a reflection mode geometry) generally improves the signal-to-noise ratio of the optical absorption spectra it measures. In addition to characterizing CO2 in patient respiration (as described in the aforementioned references, a parameter that can indicate BNP and thus the compensatory dysfunction in CHF patients), the multi-wavelength optical spectrometer 71 can measure optical spectra from patient exhalation that can indicate its composition.

[0090] The spectral signals can then be processed using algorithms (e.g., ML- and / or AI-based algorithms) to characterize the rich molecular information contained in the breath sample, thereby providing insights into a patient's health. For example, exhaled human breath contains a wide range of VOCs, VSCs, and trace gases, which can indicate various physiological and pathological processes occurring in the body. The presence of certain VOCs can indicate mask aging used in SMKs, while VSCs can indicate the presence of bacteria. Spectroscopic techniques that can be performed using a multi-wavelength spectrometer 71 include optical absorption spectroscopy, infrared spectroscopy, Raman spectroscopy, laser spectroscopy, ultrafast laser spectroscopy, and optical comb filtering. The light source 72 used for the spectrometer can be an LED, a broadband light source (e.g., a tungsten light source), or a laser (e.g., a continuous wave or pulsed laser, such as an ultrafast laser). Applications in health monitoring include disease diagnosis, as certain diseases and conditions can alter the composition of breath, resulting in the presence of specific biomarkers that can be identified spectroscopically. Specific examples include lung diseases, metabolic disorders, and gastrointestinal diseases. Additionally, by monitoring time-dependent changes in respiratory composition, spectroscopy can provide insights into disease progression and treatment efficacy, and can help assess patient response to therapy and guide personalized treatment approaches.

[0091] For example, Figure 14B and Figure 14C A multiwavelength spectrometer 71 featuring an AMS AS7341 optical sensor is shown, which is used as a dedicated photodetector 70. This component is a small-scale, chip-level system, which is suitable for... Figure 3D The embodiments shown (and again in) Figure 14A (As shown in the diagram) It works particularly well, which requires direct coupling of the spectrometer to the PAP hose 40, and is therefore small and lightweight. The broadband light source 72 used in the spectrometer 71 is a white LED that emits light radiation ranging from infrared to visible frequencies and into the ultraviolet. AS7341 is a dedicated sensor 70 characterized by a broadband photodetector covered by a set of filters from a computer-controlled microelectromechanical system (“MEMS” herein), which is controlled in software by setting certain programmable registers. Figure 14C The transmission spectra of the passbands of the filters are shown, labeled in the figure as F1-F8 (corresponding to different narrowband filters with a passband of approximately Δλ=50 nm in the visible spectrum), FXL (corresponding to relatively broadband filters ranging from λ=400-700 nm), VIS (λ=350-800 nm), and NIR (λ=800-900 nm). Other detectors similar to the AS7341 can be used for spectroscopic measurements of respiration in other optical regions, such as infrared.

[0092] During use, computer code running on the CPU within the SMK (reference) Figure 4A(More detailed description) Specific registers are set within the AS7341, which in turn activate specific MEMS filters. The emission characteristics of these filters lie in the radiation within a specific passband, such as... Figure 14C As shown, a broadband photodetector positioned after the MEMS filter detects the transmitted radiation and generates a photocurrent. A 20-bit analog-to-digital converter coupled to the photodetector digitizes the voltage corresponding to the photocurrent measured across a known resistor. The digitized voltage represents a spectral data point corresponding to the center frequency of the optical passband. A single measurement typically takes several milliseconds, and the process of sequentially setting specific registers and then measuring the corresponding signals is repeated until the complete spectrum is measured. Taken together, Figure 14A The small multi-wavelength spectrometer 71 shown in Figure C represents an ideal system for measuring different optical properties of the breath exhaled by the patient 15 during PAP treatment. The signal measured by this system, along with physiological information measured by the SMK 10, is transmitted to the cloud, where it is analyzed jointly using ML and AI models to characterize the patient's health and improve their PAP treatment, as described in more detail below.

[0093] In related embodiments, instead of the chip-level AS7341, the multi-wavelength spectrometer 71 may be based on more conventional techniques, such as a broadband tungsten light source for illuminating the sample, characterized by a detection system for diffraction gratings or prisms to disperse the light radiation after it has passed through the sample, a multi-pixel charge-coupled device (“CCD”) camera for detecting the dispersed light frequencies, and an external analog-to-digital converter and / or data acquisition system for digitizing the signal detected by each pixel in the CCD camera.

[0094] Figure 15 The frequency-dependent spectrum of a human blood sample is shown, measured simultaneously at discrete frequencies using an AS7341 and a white LED (square), and quasi-continuously measured using a conventional spectrometer with a broadband tungsten light source and a spectrometer with a diffraction grating and a CCD camera (continuous dark lines) manufactured by Thorlabs. The Thorlabs spectrometer provides quasi-continuous data points (e.g., one per nm) in the range of approximately 200–1000 nm, but is relatively large, heavy, and expensive compared to the AS7341; therefore, it is unsuitable for wearable applications like those used in SMKs. Conversely, the spectrum measured by the AS7341 is relatively limited in information, but each data point measured by the system in the range of λ = 400–700 is in good agreement with the data points measured by the Thorlabs system. Importantly, as mentioned above, this chip-level system is well-suited for integration into SMKs due to its size, weight, and cost. 2. Hardware system used in SMK

[0095] To effectively measure physiological information from the patient, the SMK includes an assembly of sensing electronics positioned directly on the mask, such that they contact parts of the patient's cheeks and face, like... Figure 1A , Figure 3A , Figure 3B , Figure 3D As shown and described in the attached text. These areas are uniquely suited for measuring physiological signals, primarily because they include dense capillary beds (ideal for measuring SpO2 and PR), proximity to the mouth (ideal for measuring RR, sounds produced during sleep, and compounds in respiration such as VOCs, VSCs, CO2, etc.), and proximity to large vessels in the chest and neck (ideal for measuring hemodynamic parameters such as SV, CO, and FLUID). The sensing electronics are typically independent digital systems that measure specific analog signals from the patient, digitizing the signals on-board using an analog-to-digital converter, and then transmitting the signals via a wired serial bus, as described in more detail below. This approach has the advantage of transmitting weak analog signals over long, “lossy” cables and then digitizing them. Each independent digital system requires power (typically 1.8–5.0V) supplied by a single-board computing platform within the control system and its associated Li:ion battery. Typically, the Li:ion battery produces a voltage ranging from 4.2V (fully charged) to 3.6V (depleted); a voltage regulator on the single-board computing platform converts this to the voltage required by the sensing electronics.

[0096] The single-board computing platform within the SMK control system controls the sensing electronics by operating them via a series of digital commands and processing the signals they generate. It features: a CPU for controlling the system and running algorithms to process sensor measurement data; flash and RAM memory; an AFE (Automatic External Component) set for controlling sensors and processing their generated signals; and one or more wireless transmitters (e.g., Bluetooth®, Wi-Fi, cellular) for transmitting digitized vital signs, hemodynamic parameters, and time-dependent waveforms to remote mobile devices and / or hubs. To control the sensing electronics distributed around the SMK, the single-board computing platform connects to a flexible ribbon cable, which is inserted into its left and right sides using separate multi-pin connectors. Together, the multi-pin connectors and the flexible ribbon cable supply power and ground to the sensing electronics and communicate with them via a serial bus operating digital protocols such as Inter-Integrated Circuit Protocol (“I2C”), Inter-Integrated Circuit Protocol (“I2S”), Serial Peripheral Interface (“SPI”), Universal Asynchronous Receiver / Transmitter (“UART”), or similar protocols. Various protocols—I2C, I2S, SPI, UART—typically require a serial line (i.e., a conductive cable in a flexible ribbon cable, such as...). Figure 3E(As shown) are used for clock, data, and chip selection. Because the communication protocols operate on the bus, they can communicate simultaneously with multiple sensing electronics using chip select lines, which identify the specific sensor used for communication. Typically, the CPU includes multiple buses for communication. In a preferred embodiment, the CPU is an STM32u545 / 575 / 585 component manufactured by STMicroelectronics.

[0097] The single-board computing platform within the SMK's control system can also be directly integrated with third-party systems, such as websites on the Internet (e.g., social media platforms or AI-based systems like ChatGPT), Amazon's Alexa, or a "smart" network in the patient's home. In embodiments, for example, the single-board computing platform may include a two-way speech-to-text translation system that allows the patient to verbally send commands to the SMK to adjust specific settings, and the SMK audibly transmits parameters measured during the previous night's sleep, such as wakefulness and / or physiological information, to the patient.

[0098] Reference Figure 4A and Figure 4B In this embodiment, the control system in the SMK includes a single-board computing platform 14, which includes the following components (note that not every component on the PCB is described below):

[0099] Figure 5A and Figure 5B A single-board computing platform 14 is shown integrated into a mechanical housing 13 located at the top 37 of the SMK 10. The mechanical housing 13 includes openings for multi-pin connectors 102a, 102b disposed on the left and right sides of the single-board computing platform 14; these connect to flexible ribbon cables (not shown in the figure, but...). Figure 3E , Figure 5C and Figure 5D As shown in the diagram, a flexible ribbon cable is attached to a tube within the PAP mask, supporting and powering the patient contact sensing electronics, and providing a bidirectional serial interface between the sensing electronics and the CPU within the single-board computing platform 14. The housing 13 houses the Li:ion battery 16 that powers the system and also includes an opening 29 for a USB-C cable. Inserting the cable into the opening 29 charges the Li:ion battery 16 and also downloads data collected during patient sleep and stored in the internal flash memory of the single-board computing platform to external devices such as hubs and / or mobile devices.

[0100] Figure 5C and Figure 5DPhotographs are shown of the PCB inside the housing and the PCB outside the housing, respectively. In both cases, a flat, flexible ribbon cable connects to multi-pin connectors on both sides of the PCB and connects the PCB to various sensors distributed throughout the SMK, such as... Figure 3E As shown in the photo.

[0101] The top 37 of the housing includes a "rotary joint" 17 to which an elbow connector (not shown) on the distal end of the PAP hose connects. The hose connects to the PAP device and supplies positive pressure air to the SMK 10. An important design consideration for the housing 13 is not to impede the rotation of the elbow connector within the rotary joint 17, as this rotation is important for maintaining a comfortable connection of the SMK 10 to the patient, even when they are tossed and turned and generally moved during sleep.

[0102] In this embodiment, the dimensions of the housing are as follows: 55mm × 30mm × 10mm. The housing, together with the PCB and Li:ion battery housed within it, weighs between 15-20 grams. The sensor, which is connected to the housing and then to the PAP mask via a ribbon cable, weighs 5-10 grams. 3. Clinical research

[0103] SMK utilizes its sensing electronics and single-board computing platform to collect time-dependent waveforms from a patient's face and cheeks during sleep. The CPU within the platform then uses algorithms to process these waveforms, such as digitally filtering the waveforms, transforming them (e.g., using Fourier transform, Laplace transform, or similar transforms), and / or performing heartbeat pickup operations to detect reference markers characterizing, for example, pulses and respiratory pulses caused by heartbeats. These could be, for example, lower limits of pulses, pulse amplitude, pulse area, pulse-to-pulse intervals, variations in intervals, etc. The algorithm running on the CPU can then process these reference markers to determine vital signs and hemodynamic parameters, such as, corresponding to the patient's HR, HRV, PR, RR, SpO2, SV, CO, and fluid. Such algorithms are described in the prior art. For example, many of them are described in U.S. Patent 11,357,453 to Banet et al., which is incorporated herein by reference.

[0104] In an alternative embodiment, the SMK and its associated system simply collect time-dependent waveforms from the patient and then forward these waveforms (using wired or wireless devices) to an external system such as a bedside hub, which then processes them as described above. In other words, in this embodiment, the SMK is merely a data collector / router that provides raw data to an external system for more complex analysis. This configuration has certain advantages because it offloads much of the signal processing, reduces the computation cycles of the SMK's CPU, thereby saving battery life and reducing the general requirements of the CPU (e.g., its size and cost). Additionally, this can mean a smaller (and lighter) Li:ion battery powering the system, thus increasing patient comfort.

[0105] More specifically, in the embodiments, the aforementioned SMK determines vital signs (e.g., HR, RR, SpO2) and hemodynamic parameters (e.g., SV, CO, FLUIDS) by jointly measuring and processing time-dependent ECG, PPG, IPG, pressure, EMG, and ACC waveforms. Figures 6A-6F As shown (Note: BR and IPG waveforms have similar shapes, therefore for simplicity, in...), Figure 6C Only the IPG waveform is shown in the figure. The analog-to-digital converter within the SMK digitizes the waveform shown, which was originally measured in analog form at 250Hz. (The waveform is shown separately in...) Figure 6A , Figure 6B and Figure 6C The ECG, PPG, and IPG waveforms shown typically include “pulses” caused by the heartbeat; these are represented in the figure by dashed lines 140a, 140b (ECG waveform), 141a (PPG waveform), and 141b (IPG waveform). The time intervals of the pulses in the ECG waveform, as shown by dashed lines 140a and 140b, are inversely correlated with HR, as... Figure 6A As shown; this value is typically 30-200 beats / minute (“bpm” in this text). The SMK can also measure HR from pulse-to-pulse intervals in both IPG and PPG waveforms. For example, dashed lines 141a and 141b represent the lower limit of the pulse within these waveforms; the interval between adjacent pulses indicates the HR. The lower limit is typically chosen as a reference because it indicates when a pulsating blood clot reaches the capillary bed below the optical sensor (in the case of a PPG waveform) or approaches a relatively large artery near the sensing and driving electrodes (in the case of an IPG waveform).

[0106] Figure 6A It shows when in Figure 13AThe configuration shown uses ECG waveforms measured by SMK, which will be described in more detail below. ECG waveforms include the “QRS complex” caused by the heartbeat, i.e., sharp, time-dependent spikes that informally mark the beginning of each cardiac cycle. Compared to other physiological waveforms, ECG waveforms generally have a relatively good signal-to-noise ratio and are easy to analyze using heartbeat detection algorithms; therefore, they are commonly used to measure HR, and the QRS complex is used as a benchmark generator for analyzing some of the more complex waveforms described below. Figure 6B The PPG waveform, measured by one or two optical sensors deployed in the SMK, is shown, indicating the volume changes in the underlying capillaries caused by heart-induced blood flow. As is known in the art, the AC and DC components of the PPG waveform, measured with red (λ~660 nm) and infrared (λ~940 nm) light radiation, can be processed together to determine the SpO2 value.

[0107] Figure 6C The IPG waveform shown also includes both AC and DC components: the DC component indicates the amount of fluid (i.e., FLUIDS) in the face and neck region by measuring baseline impedance. Figure 6C The AC component shown tracks blood flow in the face and neck and represents the pulsatile component of the IPG waveform. The time-dependent derivative of the AC component includes a well-defined peak value, which indicates the maximum acceleration of blood flow in the thoracic vascular system. Both the AC and DC components can be processed with a parameter referred to as the left ventricular ejection time (“LVET” in this paper) and the equations mentioned above (e.g., the Sramek-Bernstein or Kubicek equations, or equivalents thereof) to determine the SV. LVET, the parameter included in the Sramek-Bernstein and Kubicek equations, represents the time interval between the opening and closing of the aortic valve.

[0108] More specifically, during IPG measurement, the sensing electrode measures a time-dependent voltage (V) that varies with the resistance (R) encountered by the injected current (I). This relationship is based on Ohm's law, as follows:

[0109] An impedance circuit (e.g., an impedance AFE) measures the voltage as described above and digitizes it using an internal analog-to-digital converter. A microprocessor receives the digitized voltage along with other parameters described below and processes it using computer code with equations (e.g., the Sramek-Bernstein equations or the Kubicek equations, or mathematical variations thereof) to calculate SV. These equations, primarily based on a model assuming volume expansion, are shown below: Sramek-Bernstein equations Kubicek equations Where Z(t) represents the IPG waveform (i.e., the AC component of the waveform), and δ represents the change in body mass index (BMI). The compensation is given by Z0, which is the base impedance (i.e., the DC component of the waveform), L is estimated from the distance between the separating sensing and driving electrodes, ρ is the static resistance of the blood (135 Ωcm), and LVET is the time between the opening and closing of the aortic valve. LVET can be determined directly from the IPG waveform by analyzing a feature known as the "diphasic notch," which is typically present in each heartbeat elicited in the waveform, or from the HR value using an equation known as "Weissler regression," as follows:

[0110] When used in Weissler regression, HR can be determined based on multiple different signals, such as ECG, PPG, or IPG waveforms.

[0111] The equation and several mathematical derivatives are described in detail in the following references, the contents of which are incorporated herein by reference: Bernstein et al. Impedance Cardiography, Pulsatile blood flow and the biophysical and electrodynamic basis for the stroke volume equations 'Journal of Electrobiological Impedance, Vol. 1, pp. 2-17, 2010. Both the Sramek-Bernstein equation and the Kubicek equation assume (dZ(t) / dt) 最大值 / Z0 represents the radial velocity of blood due to the expansion of the aorta (in Ω / s).

[0112] In the above equation, the parameter Z0 will vary with the fluid level. Generally, high resistance (e.g., above about 30 Ω) indicates a dry, dehydrated state. Here, the lack of conductive pleural fluid increases the resistivity of the patient's chest. Conversely, low resistance (e.g., below about 19 Ω) indicates that the patient has more pleural effusion and may be overhydrated. In this case, a large amount of conductive pleural fluid decreases the resistivity of the patient's chest. The impedance circuit used for measurement and the specific electrodes can affect these values. Therefore, these values ​​can be further refined by conducting clinical studies (e.g., with a large number of subjects, preferably those with high variability in their fluid state) and then empirically determining “high” and “low” resistance values.

[0113] Figure 6DThe output of the pressure sensor represents a direct measurement of RR, as described in detail above. Besides providing an accurate RR value, this waveform can be used to filter out respiratory artifacts from other waveforms, such as PPG and IPG. For example, the RR measured from the pressure waveform can be incorporated into a digital filter called an adaptive filter. The adaptive filter then sets its passband to exclude physiological events that specifically occur at the RR frequency. In this embodiment, the filtering described by the passband is performed using an infinite impulse response (“IIR”) digital filter known in the art.

[0114] Figure 6E The EMG waveform characterized by three distinct pulses is shown, indicated by dashed lines 147a, 147b, and 147c, with each pulse indicating electrical activity caused by muscle movements in the patient's face (in this case, simulated by jaw clenching for approximately 170, 177, and 181 seconds). The pulses correspond to natural movements experienced by the patient during sleep, such as experiencing awakening. Measurements were used to determine this waveform. Figure 6C The sensing electrodes of the IPG waveform simultaneously sense bioelectrical signals generated by electrical activity associated with muscle movement.

[0115] The sensing electrodes are connected to an IPG AFE (MAX30009) and an ECG AFE (MAX86176), the latter representing the EMG circuitry. The IPG AFE filters out the bioelectrical signal, but the bioelectrical signal passes through a filter within the ECG AFE; then, an internal differential amplifier electronically calculates their difference and amplifies the resulting value to generate... Figure 6E The waveforms are shown. It is noteworthy that the bioelectrical signals generated by muscle noise differ significantly from one side of the face to the other, thus the pulses represented by dashed lines 147a, 147b, and 147c have relatively good signal-to-noise ratios. Conversely, the bioelectrical signals generated by the heartbeat, measured by sensing electrodes on different sides of the face, show very small differences, therefore ECG waveforms are generally not measurable from this area. However, as... Figure 13A and Figure 13B As shown, ECG AFE can use an electrode within the SMK that contacts the patient's face and a second "auxiliary" electrode that is attached to the patient's chest to measure ECG waveforms with a relatively good signal-to-noise ratio.

[0116] Figure 6F The ACC waveform measured by the accelerometer within the SMK is shown. Figure 4A As shown and as described above, the accelerometer is typically located directly on the PCB, which in turn is positioned near the top of the patient's head. The accelerometer measures along three unique axes (x, y, and z) with... Figure 6FThe ACC waveform shown is similar to other ACC waveforms and typically includes a gyroscope to measure the patient's angular motion; algorithms can process each of these to estimate motion, which in turn can indicate awakenings during sleep. Other algorithms determine the vector amplitude of the patient's motion by squaring the values ​​of the ACC waveform measured along the x, y, and z axes and then taking the square root of the resulting values. Once measured, the vector amplitude can indicate the degree of patient motion. Algorithms known in the art can process the individual ACC waveforms measured along the x, y, and z axes to determine the patient's posture. This value can be important during sleep because it can indicate a sleep posture more favorable to apnea events, such as when the patient sleeps on their back, and a sleep posture less favorable to apnea events, such as when the patient sleeps on their side or stomach.

[0117] In relevant embodiments, the microprocessor within the PCB can employ signal processing techniques other than IIR digital filters to improve signal processing. Figures 6A-6F The waveforms and signal-to-noise ratios of the heartbeat-sensing pulses are shown. These include smoothing, averaging, heartbeat stacking, and other types of digital filtering.

[0118] Parameters related to SYS, DIA, and MAP BP can be determined by analyzing the time differences between pulsation characteristics in different waveforms. For example, an algorithm operating in the firmware on SMK can calculate the time interval between the QRS group and a reference marker on each of the other waveforms. One such interval is the separation of the IPG waveform ( Figure 6C The lower limit of the pulse and the PPG waveform in ) Figure 6B The lower limit of the time. This is, for example, indicated by the first dashed line 141a in the PPG waveform (indicating the lower limit of the pulse). Figure 6B ) and the second dashed line 141b indicating the lower limit of the pulse in the IPG waveform ( Figure 6C The time interval is shown in the figure. This time interval typically represents PTT and / or VTT, and is indicated in the figure; it is usually between 10-100 ms and is inversely proportional to BP. Additionally, dashed line 144a indicates the QRS complex of the pulse in the ECG waveform. Figure 6A The time interval between this reference and the lower limit of the pulse in the PPG waveform indicated by dashed line 141a is PAT. Alternatively, PAT can be determined based on the lower limit of the pulse in the ECG QRS indicated by dashed line 144a and the IPG waveform indicated by dashed line 141b. Typically, the PAT value is slightly longer than the PTT / VTT value, usually in the range of 50-200 ms.

[0119] In an embodiment, the reference corresponding to the lower limit of the pulse can be compared with the peak value of the pulse (e.g., the IPG waveform). Figure 6C ) or PPG waveform ( Figure 6BThe reference values ​​for the peak values ​​of the pulses in the ECG are interchanged. Typically, PTT and VTT can be determined using any set of time-dependent references derived from waveforms other than ECG. In summary, the PAT, PTT, VTT, and other time-dependent contraction intervals extracted from the pulses in the four physiological waveforms described above are inversely proportional to BP.

[0120] Typically, BP measurements based on contraction intervals indicate changes in BP; they require calibration from cuff-based systems (e.g., manual auscultation or automated oscillometric methods) to determine the absolute value of BP. This calibration method typically provides an initial BP value and a patient-specific relationship between BP and PAT, PTT, and VTT. These values ​​are measured quasi-continuously during cuffless measurements and then combined with the BP value determined during calibration to produce a quasi-continuous BP value. Such calibration typically involves multiple (e.g., 2–4) measurements of the patient using an oscillometric, cuff-based BP monitor, while simultaneously collecting PAT, PTT, and VTT values ​​as described above. Each cuff-based measurement produces a separate BP value. Calibration typically lasts approximately one day before needing to be repeated.

[0121] Accurately determining BP values ​​from PAT, PTT, and VTT may also require identifying patient-specific constants that correlate variations in these transit times with changes in BP. Such patient-specific constants can be determined by measuring transit times and calibrating measurements at different BP values, and then determining them via linear interpolation. Alternatively, the constants can be estimated from metadata and population models, or by analyzing the pulse shape in time-dependent waveforms. In other embodiments, constants are determined from large datasets using ML and AI.

[0122] In related embodiments, one of the cuff-based BP measurements coincides with a “challenge event” that alters the patient’s BP (e.g., squeezing the handle, changing posture, or raising their leg). This imparts a variation to the calibrated measurement, thereby increasing the sensitivity of the calibrated measurement to BP swing. In other embodiments, a “universal calibration” (e.g., a single calibration for all patients) may be used for BP measurements. In other embodiments, BP measurements remain uncalibrated, and only relative measurements of BP are calculated.

[0123] refer to Figures 7A-7D IPG waveforms measured from the face using SMK exhibit several characteristics that make them particularly suitable for measuring transit times (e.g., PAT, PTT, VTT) and physiological parameters (e.g., BP, SV, and CO). For example, somewhat surprisingly, the pulses in these waveforms show very small modulations due to respiration. Figure 7A The IPG waveform in the image was measured from the face of patient 15, with the SMK located on the patient's head, as shown. Figure 7CAs shown in circle 159. Here, the sensing and driving electrodes used in IPG measurements are distributed in the SMK, as shown... Figure 1A and Figure 3B As shown. These electrodes generate time-dependent waveforms with well-defined pulsation characteristics, such as... Figure 7A The dashed lines 143a and 143b are shown in the figure; HR is calculated from these, as shown in the figure. The rise time of these pulsatile graphs is very sharp, indicating that blood flow to the face accelerates rapidly with each heartbeat. Importantly, IPG has very little regulation due to respiration.

[0124] on the contrary, Figure 7B and Figure 7D IPG waveforms measured from the chest of patient 15 are shown, as shown in circle 158. Here, the IPG waveforms show the characteristics caused by heartbeat as shown by dashed lines 146a and 146b, and the characteristics caused by respiration as shown by dashed lines 145a and 145b; the intervals between these pairs of dashed lines represent HR and RR, respectively. Figure 7B The IPG waveforms in the image show clear modulation due to respiration, as respiration alters the capacitance in the patient's chest and thus the impedance, and most significantly, the impedance in the patient's lungs. The modulation of the respiratory-induced IPG waveform dominates the signal, to some extent obscuring the characteristics caused by heartbeats and making them difficult (or in some cases impossible) to extract and analyze from the signal. This physiological change is not present in the patient's face, therefore IPG waveforms measured from this area typically lack a respiratory component, such as... Figure 7A As shown. This makes these signals relatively easier to extract and analyze, ultimately leading to improved accuracy in utilizing their measurement parameters (e.g., HR, BP, SV, CO).

[0125] When with Figure 7B When comparing, in Figure 7AThe relatively rapid rise of the pulse caused by the heartbeat in the IPG waveform is likely due to the rapid influx of blood from the common carotid arteries located on each side of the neck into the face. As these arteries reach the face, they branch into smaller vessels called facial arteries. The facial arteries then further divide into numerous smaller arteries that deliver oxygenated blood to different areas of the face. This vascular system contrasts with the vascular system of the chest, characterized by the large and relatively flexible aorta. The IPG signal originating from the aorta includes the characteristic volumetric expansion of this vessel (which is the largest artery in the human body) and the blood flow-induced cigar-shaped arrangement of red blood cells (called erythrocytes); both of these physiological events increase electrical conduction in the thoracic cavity, thereby reducing the impedance measured therefrom. However, their relative contributions to the IPG signal typically vary on a patient-by-patient basis and can therefore be difficult to derive and incorporate into mathematical models such as the Sramek-Bernstein or Kubicek equations. In addition, pleural effusion is quite common, especially in patients with CHF, and this fluid can artificially increase the baseline value of the IPG waveform; this in turn artificially increases the parameter (Z0) in the above equation, thereby reducing its accuracy.

[0126] Conversely, the blood vessels in the face and neck are less flexible than the aorta and are typically surrounded by muscle. Therefore, contrary to volume expansion, the IPG signals emanating from them are primarily driven by the arrangement of red blood cells, making it relatively easier to develop physiological models describing them. Such models are similar to those used to accurately calculate SV from IPG signals measured by the brachial artery, as described by an impedance-based measurement known as transbrachial electrophoresis (“TBEV” in this document). U.S. Patent 10,278,599 to Banet et al. describes TBEV measurements and is incorporated herein by reference. Additionally, conditions like CHF increase fluid in the chest but not necessarily in the neck, meaning they can lead to inaccuracies in the SV calculation equation that may affect chest-worn sensors but not SMK.

[0127] The sharp rise of pulses in the IPG waveform measured from the face also allows for a relatively accurate determination of the lower limits of these pulses. This, in turn, reduces the error in calculating transit times such as PAT, PTT, and VTT, ultimately determining the accuracy of BP calculation from these signals.

[0128] SMK measurements are particularly effective pressure waveforms for indicating respiratory events; processing them can produce accurate RR values, as well as other respiratory events such as coughing, wheezing, and apnea. Figures 8A-8D This is indicated by comparison with PPG waveforms measured from the wrist and hand using optical sensors, such as those found in conventional pulse oximeters, health trackers, "smart" watches, and "smart" wristbands. Here, as... Figure 8A and Figure 8B As shown, patients 15 wear smartwatches containing optical sensors that contact their wrists, as illustrated in the first circle 156. Meanwhile, as... Figure 8C and Figure 8D As shown, patient 15 wears an SMK with a pressure sensor near their mouth, as indicated by the second circle 159. Over a period of approximately 3 minutes, the optical sensor measures the PPG waveform (using λ=940nm) from the patient's wrist in a reflective mode geometry, and the pressure sensor measures the pressure waveform from the patient's mouth, similar to... Figure 6D The pressure waveform is shown. During the measurement, patient 15 intentionally initiated simulated wheezing breathing patterns for approximately 60 and 120 seconds, as shown in boxes 150a and 150b. Patient 15 also simulated coughing for approximately 80 and 135 seconds, as shown in boxes 152a and 152b, and even simulated apnea while holding their breath for approximately 30 to 150 seconds. Optical and pressure sensors continuously measured PPG and pressure waveforms (at a sampling rate of 250 Hz) during a 3-minute measurement cycle that included the wheezing, coughing, and apnea "challenge."

[0129] Figure 8A and Figure 8C Time-dependent plots of PPG and pressure waveforms are shown separately. The gray dashed lines spanning both plots indicate the annotated respiratory events. First refer to... Figure 8A The PPG waveform shows quasi-periodic, well-defined pulses throughout the measurement period, with each pulse corresponding to a single heartbeat of the patient. Some modulation exists in the “envelope” that defines the peak and zero values ​​of the pulses, but this modulation does not appear to specifically correspond to respiratory events, as shown by the gray dashed lines. The PPG waveform shows some variability during wheezing, coughing, and apnea, as indicated by boxes 150a, 150b, 152a, 152b, and 154 respectively, although this variability does not appear to include specific features indicating these particular challenges; that is, the variability corresponding to wheezing shown in box 150a appears indistinguishable from the variability corresponding to cough shown in box 152a or the variability corresponding to apnea shown in box 154.

[0130] Conversely, refer to Figure 8CThe pressure waveform performs better in indicating respiratory events and challenges caused by wheezing, coughing, and apnea. For example, during each breath indicated by the gray dashed line, the pressure waveform shows a well-defined pulse that precisely corresponds to the respiratory event. During apnea, as shown in box 154, the respiratory-induced pulse disappears completely. The wheezing periods indicated in boxes 150a and 150b show pulses with significantly higher amplitudes than those caused by heartbeats and are more similar in amplitude to respiratory pulses; however, for short “bursts” corresponding to wheezing events, they occur at a relatively high frequency (e.g., >2 pulses / second). Coughing, as shown in boxes 152a and 152b, produces pulses with even higher amplitudes than those corresponding to wheezing. They occur at a frequency corresponding to the cough frequency (approximately one pulse every 1–2 seconds), which is generally lower than the wheezing frequency, and have a unique shape different from the pulses produced by wheezing.

[0131] In respectively Figure 8A and Figure 8C The PPG and pressure waveforms shown in the figure indicate that pressure waveforms measured directly from the patient's mouth are generally superior to PPG waveforms measured near the wrist and fingers for detecting respiratory events such as normal breathing, wheezing, coughing, and apnea. PPG waveforms remain valuable for determining SpO2 and PR, which remain important parameters characterizing the efficacy of OSA and PAP treatment. 4. ML model for processing SMK-generated data

[0132] In this embodiment, a cloud-based ML model can process physiological information generated by SMK (e.g., numerical and time-dependent waveforms of vital signs) to characterize the patient. Such an ML model is first generated using data measured during polysomnography from patients undergoing a lab-based sleep study used to screen for OSA. This data is categorized in the Sleep Heart Health Study Database (“SHHSD” in this context) and is similar to (but not identical to) the data measured by SMK. During the sleep study, clinicians with expertise in sleep disorders annotated the following sleep-related conditions (referred to as “data classes” in SHHSD): arousal, hypopnea, central apnea, obstructive apnea, and SpO2 desaturation events. The clinicians recorded each annotated data class, meaning the event was observed, and documented its severity and timing. Furthermore, data during periods when no data class was present were marked as “clean.”

[0133] To generate the report shown in the figure, a series of ML-based models were first tested on a first group of patients in SHHSD using annotated data classes (specifically arousal) and the relevant physiological information leading to them. Once the models were optimized, they were tested on a second group of patients to predict data classes. Because these events were previously annotated by sleep clinicians and recorded in SHHSD, this method allows for testing various model parameters, such as accuracy, specificity, and selectivity, to predict data classes based on physiological information measured using SMK. Finally, the optimal model is selected.

[0134] For various ML models, the “benchmark” as described above and the calculated parameters in SHHSD were detected from time-dependent waveforms collected using 30-second windows around different data categories. This information was selected due to its similarity to the data measured by SMK and included: 1) HR and related variability measures (e.g., standard deviation, root mean square standard deviation, mean HR variability); 2) RR and related variability measures (e.g., instantaneous RR, respiratory-to-respiratory time interval, and additional statistical measures such as standard deviation, mean, maximum, minimum); 3) SpO2 and related variability indicators (e.g., variability, mean, maximum, minimum); and 4) motion-related information extrapolated from EMG measurements (this was used as a substitute for ACC waveforms because accelerometer data was not available in SHHSD). The data was then split in several ways to evaluate the effectiveness of different classification models (binary and multiclass) and to understand the impact of class imbalance on the dataset. Imbalance was then performed based on: 1) a single data class; 2) grouping of data classes, such as wake-up, clean data and other events; or 3) binary, such as wake-up and clean data; wake-up and other events; and wake-up and everything else.

[0135] Once these parameters are extracted from the dataset, they are formatted and processed using various ML models available from AWS. In one embodiment, other cloud-based platforms, such as Microsoft Azure or any other similar cloud-based software system, can be used instead of AWS. In other embodiments, the cloud-based software system is directly integrated with the regenerative AI model, such as an AI model used with a system like ChatGPT.

[0136] The performance of ML models is based on accuracy (or alternatively, F1 score, defined as follows) and factors such as... Figure 10The area under the curve (AUC) calculated from the receiver operating characteristics (ROC) plots shown in this paper is used for evaluation. The ROC curves show the correlation between the true positive rate (y-axis) and the false positive rate (x-axis). The AUC calculated from them provides an overall performance measure for all classification types, where 1 represents a perfect classifier, values ​​of 0.9–1.0 are considered excellent, 0.8–0.9 are very good, 0.7–0.8 are good, 0.6–0.7 are satisfactory, 0.5–0.6 are unsatisfactory, and 0.5 or lower is a completely random classifier.

[0137] Figure 9 The results of using the aforementioned ML models to detect data classes are summarized. Specifically, these figures describe the different models tested, the data classes used, how the data were processed, the specific ML model used, the obtained accuracy (or F1 score of the overall mean), and AUC. In this context, a data class refers to polysomnography events annotated in SHHSD, including EEG-based arousal, respiratory events, pulse oximeter artifacts, awakening, central apnea, obstructive apnea, and hypopnea. These annotations are based on assessments by trained and certified technicians. A data class is considered "clean" if none of these events occur.

[0138] exist Figure 9 In this context, the term "class balance" refers to the distribution of each data class within a dataset. This is important to ensure that the data is not biased by having too many or too few different data classes. For example, if only 5% of the data consists of annotated awakenings, the ML model will be biased towards false negatives. Similarly, if the dataset includes a disproportionate number of annotated insufficiency (or any other) events, this will also bias the performance of the ML model towards those specific events. Thus, during the analysis, the data classes are "balanced" such that the target data class (in this case, awakening) comprises approximately 50% of the data, and the other data classes are represented in approximately the same proportion expected in the population. In the table, "All" means that the data is evenly divided among the included data classes. "Type" means that the data is evenly divided by data associated with a specific type of data, where "Awake" means awakening, "Clean" means data without any events, and "Event" or "Other Event" means any respiratory event or pulse oximeter artifact. "Target" refers to what the ML model is trying to predict (e.g., awakening). For example, a "Binary Target" is designed to determine whether a data class is awakening or something else. “Num pClasses” represents the number of predicted classes, that is, the total number of data classes that the ML model attempts to predict.

[0139] The term "feature set" refers to the set of features evaluated by the ML model. For 'v1', this includes: 'sao2_mean', 'sao2_max', 'sao2_min', 'hr_mean', 'hr_max', 'hr_min', 'hrv_rmssd', 'hrv_mean_r2r', 'hrv_stdnn', 'hrv_nn50', 'hrv_pnn50', 'hrv_max_r2r', 'hrv_min_r2r', 'rr', 'rr_max', 'rr_min', 'btbi_rmssd', 'btbi_rmssd_min', 'btbi_rmssd_max', 'btbi_var', 'btbi_var_max', 'btbi_var_min', and 'emg_tat'.

[0140] The ML model labeled 'v1+' includes all of these features as well as the "sleep state".

[0141] "ML model" refers to the specific model used in AWS for the computations described in the table. Specifically, the following AWS models were trained and tested: LightGBM, CatBoost, XGBoost, Random Forest, Extra Tree, Linear Models, Neural Networks implemented in PyTorch, Neural Networks implemented using fast.ai, and Multilayer Perceptrons. In many cases, weighted ensembles were used to obtain the best model performance.

[0142] "Accuracy" is the total number of annotations correctly classified by the model divided by the total number of annotations, and is defined using the following equation, where 'TP' represents "true positive", 'TN' represents "true negative", 'FP' represents "false positive", and 'FN' represents "false negative": For the computation of weighted ensembles using ML models, the F1 score is used as a representative of accuracy and is defined as: in as well as

[0143] For details, please refer to the following: Figure 9 The model, using a binary objective and a weighted ensemble of the ML model to process all data classes, was run 8 times using SHHSD to produce the best results. The F1 score and AUC calculated as described above were 0.822 and 0.9, respectively. Figure 10The ROC curve for this specific model is shown. Based on these results, this ML model appears ideal for calculating arousal and other data categories present in patients with OSA.

[0144] During SMK deployment, the aforementioned ML models will typically operate on servers running in the cloud, such as... Figure 2 As shown. When the SMK measures new data from the patient, the bedside hub collects it and sends it to the cloud, where the model processes it to estimate various data categories. The results are then transmitted to... Figure 2 Third-party software systems can be used to analyze and, for example: 1) adjust PAP treatment; 2) characterize patient health and progression to specific chronic diseases (e.g., CHF); 3) replace equipment (e.g., leaky or aging PAP masks); or 4) for other applications aimed at improving patient outcomes. 5. Related and Alternative Examples

[0145] In one embodiment, the SMK is "docked" into a bedside hub, and then its internal Li:ion battery is charged while data is downloaded and displayed simultaneously. Figure 11A and Figure 11B A mechanical model of this bedside hub 42 is shown; Figure 11C A photograph shows the hub connected to the PAP machine and docked with the SMK via a flexible hose. The bedside hub 42 includes a base 208, which is weighted to prevent the hub from tipping over and houses a small computing platform (not shown, but similar to the Raspberry Pi 4 described at https: / / www.raspberrypi.com / products / raspberry-pi-4-model-b / ). The base 208 is connected to a vertical support structure 209 that houses a touch panel display 206. The touch panel display 206 is connected to the computing platform in the base 208 via a video cable (e.g., USB-C or HDMI). The touch panel display can present a user interface, such as... Figures 12A-12F The user interface shown and described below. Hub 42 also includes a neck 205 connected to mounting region 204, which houses the mechanical housing 37 of the SMK during charging. Mounting region 204 features a USB-C (male) connector that inserts into a USB-C (female) connector mounted on PCB 13 and can be accessed through an opening in housing 13 (e.g., see...). Figure 5BComponent 29 is obtained. A USB-C (male) connector is connected via a cable to the small-scale computing platform in the base 208 of the hub 42. Once the SMK is plugged into the hub 42, the USB-C (male) connector in the mounting area 204 provides current to be passed from the small-scale computing platform to the PCB 14 and through the cable, and the current is then used to charge the Li:ion battery. The cable also outputs the data collected from the patient, initially stored in the flash memory on the PCB 14, to the small-scale computing platform for subsequent analysis. Such analysis may include processing time-dependent waveforms collected from the patient as described above, or sending raw and / or processed data to a cloud-based system via wired or wireless devices, such as... Figure 2 As shown.

[0146] Figures 12A-12F The image shows a screenshot of the GUI displayed on the touch panel during operation. A small computing platform runs the computer code that controls the GUI. Figure 12A As shown, in one embodiment, when a patient wears the SMK, the GUI displays the current time like a standard alarm clock. Software running on a small-scale computing platform periodically checks the USB port located in mounting area 204 to determine if the SMK is inserted. When inserted, the GUI displays... Figure 12B The screen shown presents questions for a simple survey, first inquiring about sleep quality, then mask comfort, and finally how many times the patient woke up during the night. Figure 12D As shown in the image. Upon completion of the survey, the user is prompted to click a button (in...). Figure 12B , Figure 12D and Figure 12E The button marked "Click to Upload" collects data stored on the flash memory of the SMK and sends it to the cloud, such as... Figure 2 As shown. When all data is uploaded, the software running in the cloud sends the data packets back to the hub, indicating that the upload process is complete and displaying the number of uploaded files in the GUI, such as... Figure 12E As shown.

[0147] According to the invention, other surveys can be performed. For example, in embodiments, these surveys may include simple games or puzzles to test the patient's mindset. Results from the surveys, particularly when coupled with physiological data collected by the SMK, can be used for ML and AI computations and report generation, such as... Figure 9 and Figure 10 As shown.

[0148] like Figure 12C and Figure 12FAs shown, the GUI can also display real-time measured data from the SMK (e.g., digital and waveform data). The SMK typically uses Bluetooth® to transmit data to the bedside hub. In this mode, the combination of the SMK and the hub can function similarly to a routine vital signs monitor in a hospital. For example, as shown, the GUI can display real-time ECG waveforms (…). Figure 12C ) and IPG waveform ( Figure 12F ), as well as any other physiological parameters measured by SMK.

[0149] Other embodiments are also within the scope of this invention. For example, as mentioned above, ECG waveforms cannot always be measured from the area above a patient's neck with a good signal-to-noise ratio. However, as... Figures 13A-13B As shown, high-quality ECG waveforms can be measured from a first electrode (indicated by circle 30b) within the SMK (Suitable Memory Kit) that contacts the patient 15 near the head or cheek, and a second “satellite” electrode (indicated by circle 30a) that contacts the patient 15 anywhere on the chest. A thin cable 79 connects the electrodes indicated by circles 30a and 30b. Typically, in this embodiment, the first electrode in the SMK is a reusable electrode, such as one made of conductive rubber or fabric; the second electrode is typically an adhesive electrode characterized by conductive hydrogel rivets coated with an Ag:AgCl film, all supported by an adhesive backing. A 3M red dot electrode is an example of this. In this embodiment, the SMK includes ports (e.g., ports similar to stereo jack connectors), and the thin cable 79 is easily inserted into and removed from the ports. This configuration could be used, for example, if other HR monitoring sensors on the SMK (e.g., optical sensors measuring PPG waveforms or impedance sensors measuring IPG waveforms) indicate that the patient may have an arrhythmia. If such a condition is detected, the GUI prompts the patient to unfold... Figure 13A The measurement conditions shown allow for the measurement of ECG waveforms (which are ideal for detecting arrhythmias).

[0150] Figure 13B It shows Figure 12C The image shown depicts a GUI (operating on a mobile tablet in this case) that displays an ECG waveform generated by this configuration.

[0151] In other embodiments, signals from wireless transceivers within the SMK can be analyzed (e.g., triangulated) to determine the patient's location. In this case, a computer operating at a central monitoring station, such as a hub, can perform triangulation to determine the patient's location. In other embodiments, the sensor may include a more conventional positioning system, such as the Global Positioning System (“GPS” herein). For example, in embodiments, the GPS and its associated antenna are typically included in the PCB within the SMK.

[0152] In relevant embodiments, the bedside hub may include a camera (e.g., a video camera) to record the patient during sleep. The output from the camera can be incorporated into the aforementioned ML and AI models to better characterize the patient's sleep.

[0153] In other embodiments, the Li:ion battery within the SMK is charged via other means, such as wireless inductive charging. This avoids USB-C-based charging, which requires an opening in the SMK's casing, making it vulnerable to damage from intruding fluids.

[0154] In other embodiments, the SMK includes additional sensors (e.g., optical and chemical sensors) for enhanced measurements from the patient's respiration. For example, such sensors can be used to enhance results from the aforementioned pressure sensors (e.g., the Bosch BME688 sensor). Such sensors can measure chemicals emitted from a patient's respiration that are correlated with their glucose levels and are generally associated with their diabetes, such as actual glucose in the breath, or alternatively, compounds such as acetone, β-hydroxybutyrate, and acetoacetic acid, all of which can indicate diabetic ketoacidosis, a potentially life-threatening condition in diabetic patients. As a specific example, the MQ138 sensor manufactured by Zhengzhou Weisen Electronic Technology Co., Ltd. has shown efficacy in measuring acetone from human respiration, which in turn has shown a strong correlation with routine glucose measurements using a blood glucose meter (see, for example, Salman et al.'s "..."). Blood Glucose Level Measurement from Breath Analysis "International Journal of Biomedical and Bioengineering, Vol. 12, No. 9, 2018".

[0155] In other embodiments, the sensors described above can also be used to characterize the physical properties of the face mask. For example, a microphone sensor can detect sounds indicating a leak or poor fit in the face mask. A pressure sensor can detect something similar. Impedance electrodes are connected to the actual face mask and can therefore detect the impedance characteristics of its silicone rubber, which relate to mechanical properties such as elastic modulus, flexibility, stiffness, etc. These parameters can in turn indicate whether the face mask is deteriorating. Optical sensors coupled to the face mask and hose, particularly the aforementioned multi-frequency optical sensors, can detect discoloration in these components and, like the impedance sensors, indicate that they may be deteriorating.

[0156] In the claims, any reference numerals placed between parentheses should not be construed as limiting the claims. The words "comprising" or "including" do not exclude the presence of elements or steps other than those listed in the claims. In an apparatus claim listing several means, several of these means may be implemented by the same hardware. The words "a" or "an" preceding an element do not exclude the presence of a plurality of such elements. In any apparatus claim listing several means, several of these means may be implemented by the same hardware. The fact that certain elements are listed in mutually different dependent claims does not mean that these elements cannot be used in combination.

[0157] Although the invention has been described in detail based on embodiments currently considered to be most practical and preferred for illustrative purposes, it should be understood that such details are for that purpose only, and the invention is not limited to the disclosed embodiments, but rather is intended to cover modifications and equivalent arrangements within the spirit and scope of the appended claims. For example, it should be understood that the invention contemplates that, to the extent possible, one or more features of any embodiment may be combined with one or more features of any other embodiment.

Claims

1. A system for monitoring blood pressure values ​​from a patient, comprising: Wearable face mask, suitable for being coupled to a positive airway pressure (PAP) machine and configured to deliver an airflow generated by the PAP machine to a patient's airway; A first sensor, directly connected to the wearable mask, includes an optical sensor configured to measure a time-dependent optical waveform from a first region below a first portion of the wearable mask, the time-dependent optical waveform including a first pulse. A second sensor, directly connected to the wearable mask, includes an impedance sensor configured to measure a time-dependent impedance waveform from a second region adjacent to the wearable mask, the time-dependent impedance waveform including a second pulse; and A processing system, attached to the wearable mask, includes a microprocessor configured to: 1) receive digital representations of both the first pulse and the second pulse; 2) Process the digital representation to determine the time difference between the first pulse and the second pulse, or a parameter calculated from the time difference; And 3) Process the time difference or the parameter calculated from the time difference to determine the blood pressure value.

2. The system according to claim 1, wherein the optical sensor comprises a first light source and a photodiode.

3. The system of claim 1, wherein the impedance sensor comprises at least one sensing electrode and at least one driving electrode.

4. The system of claim 4, wherein both the at least one sensing electrode and the at least one driving electrode comprise a conductive material selected from the group consisting of rubber, polymer, fabric, metal, wire, mesh, and hydrogel.

5. The system of claim 4, wherein the at least one driving electrode is configured to inject a first current into the second region.

6. The system of claim 5, wherein the impedance sensor is further configured to modulate the first current at a frequency in the range of 5 kHz to 500 kHz.

7. The system of claim 6, wherein the first current has an amplitude in the range of 0.01 mA to 5 mA.

8. The system of claim 1, wherein the processing system is further configured to determine a lower limit of the first pulse and a lower limit of the second pulse.

9. The system of claim 8, wherein, in order to determine the time difference between the first pulse and the second pulse, the processing system is further configured to calculate a time interval between at least one of the following characteristics of the first pulse and the second pulse: a lower limit, a maximum slope of the rising edge of the pulse, a base, a peak value, and a rising edge.

10. The system of claim 1, wherein the processing system is further configured to calculate the reciprocal of the time difference between the first pulse and the second pulse.

11. The system of claim 10, wherein the processing system is further configured to process the reciprocal of the time difference between the first pulse and the second pulse using a calibration value to determine the blood pressure value.

12. The system of claim 1, further comprising: A control system directly integrated into the wearable mask, the control system including the processing system.

13. A system for monitoring blood pressure values ​​from a patient, comprising: Wearable face mask; A first sensor, directly connected to the wearable mask, includes an optical sensor configured to measure a time-dependent optical waveform from a first region below a first portion of the wearable mask, the time-dependent optical waveform including a first pulse. A second sensor, directly connected to the wearable mask, includes an impedance sensor configured to measure a time-dependent impedance waveform from a second region adjacent to the wearable mask, the time-dependent impedance waveform including a second pulse; and A processing system, attached to the wearable mask, includes a microprocessor configured to: 1) receive digital representations of both the first pulse and the second pulse; 2) Process the digital representation to determine the time difference between the first pulse and the second pulse, or a parameter calculated from the time difference; And 3) Process the time difference or the parameter calculated from the time difference to determine the blood pressure value.

14. The system of claim 13, wherein the optical sensor comprises a first light source and a photodiode.

15. The system of claim 13, wherein the impedance sensor comprises at least one sensing electrode and at least one driving electrode.

Images