Systems, methods and apparatus for generating blood glucose estimations using real-time photoplethysmography data

Abstract

To generate a blood glucose estimation for a subject.SOLUTION: The method includes receiving a real-time measurement of blood glucose via a blood glucose monitoring device attached to the subject and, in response to receiving the real-time measurement of blood glucose, updating one or more parameters of the model in real time to improve blood glucose estimation accuracy of the model. The method may also include detecting whether the generated blood glucose estimation is above or below a threshold and, in response to determining that the generated blood glucose estimation is above or below the threshold, receiving a real-time measurement of blood glucose and then updating the parameters of the model to improve blood glucose estimation accuracy of the model.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates generally to wearable devices, and more specifically to medical, health and fitness devices. Wearable Biometric Sensor Technology for Physiological Monitoring of Fitness Applications Regarding.

[0002] [Related Applications] This application is a continuation of U.S. Provisional Patent Application No. 63 / 132,233, filed December 30, 2020. This application claims the benefit and priority of that provisional patent application. The disclosure of this application is incorporated herein by reference as if set forth in its entirety. It shall be so. [Background technology]

[0003] The "holy grail" of diabetes management would be complete painlessness for the end user. Our portfolio includes truly non-invasive, continuous blood glucose monitoring solutions that are virtually invisible. There have been many attempts to provide such a commercial solution, but none have been successful. not present.

[0004] For continuous glucose monitoring solutions, check out the Dexcom G6 CG today. M System (Dexcom, Inc., San Diego, California) and others are in the market. These conventional monitoring systems use microneedles or periodically collect samples of bodily fluids (e.g., interstitial fluid) using other minimally invasive modalities. However, these systems are inherently invasive and require extensive testing. The effects are minimal and usually cause disruption to the underlying skin. The glucose concentration in the blood usually changes several minutes later than the glucose concentration in the blood. Urgent feedback to end users may be delayed. Just as importantly, The form factor of a conventional glucose monitoring system is usually a patch. This can make it difficult for many end users to use.

[0005] As reported in John Smith's non-patent literature 1, many researchers have investigated insulin or A non-invasive method for measuring blood glucose levels accurately enough to allow glucose to be administered However, the results are not suitable for commercial use. Therefore, new methods are needed to achieve this. [Prior art documents] [Non-patent literature]

[0006] [Non-Patent Document 1] “The Pursuit of Noninvasive Glucose: Hunting the Deceitful Turkey” Summary of the Invention

[0007] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. It should be understood that this summary is provided for illustrative purposes only and is not intended to be limiting. These are not intended to identify key features or essential features of the disclosure, but are intended to be illustrative of the present invention. It is not intended to limit the scope of the present invention.

[0008] According to some embodiments of the present invention, a method for generating an estimated blood glucose level for a subject includes at least a step of detecting a P attached to the subject as a step executed by at least one processor; receiving real-time PPG data from a PG sensor; and using the data to generate an estimated blood glucose level for the subject using an adaptive predictive model. Exemplary adaptive predictive models include regression models, machine learning models, and classifier models. However, the present invention is not limited to the above. The generating of the estimated blood glucose value may include generating a future estimated blood glucose value. In some embodiments, the current estimated blood glucose level and The historical estimated blood glucose levels can be processed to predict future estimated blood glucose levels.

[0009] The method includes receiving a measured blood glucose level from a blood glucose monitoring device; The adaptive prediction model is adapted to improve the blood glucose level estimation accuracy according to the received value. and updating one or more parameters of the rule in real time. In some embodiments, the measured blood glucose level is a real-time measurement.

[0010] The method includes receiving activity information of a subject from a motion sensor attached to the subject. and real-time measurement from a blood glucose monitoring device according to the subject's activity information. The adaptive prediction model is based on the acceptance of blood glucose values ​​and the acceptance of real-time blood glucose measurements. One or more parameters of the adaptive predictive model are adjusted in real time to improve the accuracy of blood glucose estimation in the patient. and updating the time.

[0011] The method includes detecting whether the generated estimated blood glucose level is above or below a threshold. The method may further include determining whether the generated estimated blood glucose value is above a threshold value. Depending on whether the blood glucose level measured by the blood glucose monitoring device is above or below the threshold, The adaptive prediction model is adapted to improve the accuracy of blood glucose estimation. and updating one or more parameters of the model in real time.

[0012] In some embodiments, the method further comprises determining whether the generated estimated blood glucose level is above a threshold. The method further includes sending an alert to a remote device that the threshold has been reached or is below the threshold.

[0013] In some embodiments, the at least one processor is attached to the subject. The wearable device is provided on the subject's ear. Configured to be worn on a limb, as a patch attached to the subject, or on the subject's finger It is possible.

[0014] In some embodiments, the wearable device comprises a PPG sensor.

[0015] In some embodiments, the at least one processor is attached to the subject. The wearable device is provided with a PPG sensor and a blood glucose monitor. The device has a tagging device.

[0016] In some embodiments, the PPG sensor is an imaging sensor.

[0017] According to another embodiment of the present invention, a wearable device includes a PPG sensor and a wearable The estimated blood glucose level of the subject wearing the wireless device is measured in real time from the PPG sensor. and at least one processor for generating an adaptive prediction model using the PPG data. The wearable device is attached to the subject's ear, the subject's limb, or the subject's The device may be configured to be worn as a patch attached to the subject's finger or on the subject's finger. The adaptive predictive models can include regression models, machine learning models, and classifier models. However, the present invention is not limited to these.

[0018] The at least one processor receives the measured blood glucose value from the blood glucose monitoring device. The adaptive prediction model estimates blood glucose levels based on the acceptance of real-time blood glucose measurements. Updating one or more parameters of an adaptive predictive model in real time to improve accuracy The at least one processor may be configured to The device accepts measured blood glucose values ​​from the monitoring device and generates estimated blood glucose values ​​that exceed a threshold. Depending on whether the blood glucose level is above or below the threshold, the accuracy of the adaptive prediction model is improved. updating one or more parameters of the adaptive predictive model in real time to improve The at least one processor may be configured to: It can send alerts to remote devices when sugar levels are above or below thresholds. It can be configured as follows.

[0019] In some embodiments, the PPG sensor is an imaging sensor.

[0020] According to another embodiment of the present invention, an adaptive predictive model (e.g., a regression model, machine learning The method to improve the accuracy of blood glucose estimation (e.g., model, classifier model, etc.) is to use at least one The steps performed by the processor include: a) detecting a PPG signal from a PPG sensor attached to the subject; The real-time PPG data from the study and blood glucose measurements from blood glucose monitors were compared during the registration period. b) generating features from the received PPG data. c) storing the feature amount and the measured blood glucose level; and d) storing the stored feature amount. and processing the measured blood glucose level to determine one or more parameters of the adaptive predictive model. and updating the updated one or more parameters in real time. Improve the accuracy of blood glucose level estimation in the prediction model. Features and blood glucose level measurements are sorted in a FIFO (first in, first out) order. The data can be stored in a data buffer, such as a first-in-first-out buffer. However, other types of data buffers can be used. This can be repeated over one or more time periods to refine the model's estimation accuracy.

[0021] The method includes generating an estimated blood glucose level for the subject using an adaptive predictive model; determining whether the estimated blood glucose level is above or below a threshold value. Depending on whether the generated estimated blood glucose level is determined to be above or below a threshold, The blood glucose monitoring device accepts another blood glucose measurement and uses the adaptive predictive model. One or more parameters of the rule are updated in real time.

[0022] In some embodiments, generating features from the received PPG data. is a feature generation interval that is generated by a sliding time window during a reception period. In some embodiments, generating one or more of the adaptive predictive models. Updating the parameters includes updating the stored measured blood glucose values ​​and the previously stored measured blood glucose values. and generating an interpolation between the stored measured blood glucose value and a previously stored measured blood glucose value. and processing the stored measured blood glucose values ​​and the previously stored measured blood glucose values. The step of measuring the blood glucose level may include processing a plurality of previously stored measured blood glucose levels. Processing the stored measured blood glucose values ​​and the previously stored measured blood glucose values ​​is performed to obtain the expected measured blood glucose values. It may include generating an interpolation of the blood glucose level.

[0023] In some embodiments, the PPG sensor is an imaging sensor.

[0024] In some embodiments, the stored features are processed together with the stored measured blood glucose levels. The processing includes processing a function of the at least one stored feature.

[0025] In some embodiments, the stored features are processed together with the stored measured blood glucose levels. The processing includes computing statistics about the time series of at least one stored feature. In some embodiments, the stored features are stored in the blood glucose measurement. The processing with the values ​​includes processing the multiple time series of at least one stored feature. This includes calculating statistics on the

[0026] In some embodiments, the stored features are processed together with the stored measured blood glucose levels. The processing includes weighting the multiple time series of at least one stored feature. This includes calculating statistics.

[0027] According to another embodiment of the present invention, an adaptive predictive model (e.g., a regression model, machine learning The system for improving the accuracy of blood glucose level estimation (e.g., blood glucose level estimation model, classifier model, etc.) is Real-time PPG data from the attached PPG sensor and blood glucose monitoring device The blood glucose measurement data from the chair must be received within the reception period, and the PPG data received must be generating a feature from the measured blood glucose level; storing the feature and the measured blood glucose level; and and processing the amount of blood glucose measured with the stored measured blood glucose levels to generate one or more of the adaptive predictive models. and updating the parameters in real time; The updated one or more parameters improve the accuracy of the adaptive predictive model's blood glucose estimation.

[0028] In some embodiments, the at least one processor is adapted to adaptively predictive models. generating an estimated blood glucose value based on the estimated blood glucose value; and determining whether the generated estimated blood glucose value is above or below a threshold value. and determining whether the generated estimated blood glucose level is above or below a threshold. update one or more parameters of the adaptive predictive model in real time in response to a determination that the The at least one processor further determines whether the generated estimated blood glucose value is greater than or equal to a threshold value. Configure it to send alerts to a remote device when the It is possible.

[0029] Aspects of the invention described with respect to one embodiment may be incorporated into another embodiment. Although not specifically described in this connection, it is possible. may combine features of any embodiment in any manner and / or in any combination. The applicant may make any claim as originally filed dependent on any other claim and and / or to incorporate any feature of any other claim as originally claimed. The right to change any claim as originally filed, including the right to amend the claim even if it is not The present invention relates to a method for producing a computer-readable recording medium and a method for recording such recording medium. and / or aspects, as well as other objects and / or aspects, are described in detail below.

[0030] The accompanying drawings, which form a part hereof, illustrate various embodiments of the present invention. The aspects and the description together serve to fully explain embodiments of the present invention. [Brief description of the drawings]

[0031] [Figure 1] FIG. 1 illustrates a computing system for generating biometric estimates according to some embodiments of the present invention. [Diagram 2] 1 is a flowchart of a method for generating a biometric estimate according to some embodiments of the present invention. [Diagram 3] 1 is a flowchart of a method for generating a biometric estimate according to some embodiments of the present invention. [Figure 4] 1 is a flowchart of a method for generating a biometric estimate according to some embodiments of the present invention. [Diagram 5] FIG. 1 illustrates a non-limiting exemplary wearable device that may be utilized in accordance with embodiments of the present invention. [Figure 6] FIG. 2 illustrates a time sliding window that may be utilized for accepting PPG data and biometric data according to some embodiments of the present invention. [Figure 7] FIG. 2 is a block diagram illustrating operations for updating one or more parameters of an adaptive predictive model according to some embodiments of the present invention. [Figure 8] FIG. 1 illustrates an adaptive prediction model according to some embodiments of the present invention. [Figure 9]FIG. 1 illustrates a computing system for generating biometric estimates according to some embodiments of the present invention. [Figure 10] 1 is a flowchart of a method for generating a biometric estimate according to some embodiments of the present invention. [Figure 11] FIG. 2 is a block diagram illustrating operations for updating one or more parameters of an adaptive predictive model according to some embodiments of the present invention. [Figure 12] 1 is a data plot showing real-time BP measurement data and real-time PPG-BP estimation collected from a subject wearing a blood pressure cuff and PPG sensor, according to some embodiments of the present invention. [Figure 13] 1 is a graphical output of a subject's estimated and actual blood pressure over a period of time, showing improvement in blood pressure estimation over time with enhancement. [Figure 14] FIG. 13 shows a table comparing volume-clamped BP estimation with PPG-BP estimation according to an embodiment of the invention in terms of accuracy with respect to actual BP measurements. [Figure 15] FIG. 2 is a block diagram showing details of an exemplary processor and memory that can be used in accordance with various embodiments of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0032] The term "subject" as used herein generally refers to a human being in the context of the present specification. However, in the context of the present invention, a subject may also be a non-human organism.

[0033] The term "biometric" generally refers to the processing of physiological (i.e., biological) information of a subject. This refers to a metric of the subject that is generated by processing the Typical examples are heart rate (HR) and heart rate variability (HRV). ), RR interval, respiratory rate, weight, height, sex, physiological condition, overall health condition, medical condition, injury injury status, blood pressure, arterial stiffness, cardiovascular fitness, VO 2max , Gas exchange analysis These may include blood analyte levels, body fluid metabolite levels, and the like.

[0034] As used herein, the terms "biometric" and "physiological metric" mean They are interchangeable.

[0035] The term "real-time" refers to a process that requires a period of time that appears substantially real-time to an individual human. Hence the term "real-time" is used herein to describe the are used interchangeably to mean "near real-time" or "pseudo real-time" That is, a "real-time" process can refer to an "instantaneous process", It produces output within a short enough processing time (in the context of the particular use case) that it is an instantaneous process. It can also refer to a process that is (substantially) as useful as a process of The process of generating blood pressure metrics can take several seconds or minutes, making this use case the With the subject's sitting posture, small changes in pressure may be insignificant and averaged out. Therefore, even though blood pressure may change every second, the blood pressure as used herein is It can be considered a real-time process.

[0036] As used herein, the terms "respiratory rate" and "respiratory rate" are interchangeable.

[0037] As used herein, the terms "heart rate" and "pulse rate" are interchangeable.

[0038] As used herein, the term "system" refers to a system that is unified by a common function. Refers to a collection of physical and / or computational elements that can

[0039] As used herein, the term "motion sensor" refers to a sensor that detects the motion of (e.g., a subject). A non-limiting example of a motion sensor is a single-axis inertial sensor. Or multi-axis inertial sensors (accelerometers, gyroscopes, MEMS motion sensors, etc.), optical Scatter sensors, closed channel sensors, etc. may be included.

[0040] As used herein, the term "photoplethysmography" (PPG) Photoplethysmography (PPG) is a method for obtaining physiological information from PPG waveforms collected by a PPG sensor. This refers to the method of generating information.

[0041] As used herein, the term "PPG waveform" refers to the time course of photon flux through a physiological substance. This refers to physiological waveform data derived from dynamic modulation of the

[0042] As used herein, the term "PPG sensor" refers to a device that detects photons and collects PPG waveform data. A typical PPG sensor generates a wide range of light in the light spectrum (i.e., about 10n Electromagnetic wavelength range from 1000 to 103 μm, or electromagnetic frequency range from approximately 300 GHz to 3000 THz Non-limiting examples of optical sensors include: Inorganic and / or organic photodetectors (photoconductors, photodiodes, phototransistors, etc.) phototransducers, etc.), reverse-biased light-emitting diodes (LEDs These may include a photodiode (e.g., a photoemitter), an imaging sensor, a photodetector array, or the like. In addition, a typical PPG sensor detects photons (photons) that generate a photon flux through the physiological pathway. However, in some cases, the device may also have an ambient photon or another photon emitter. Photons from a light source (not part of the PPG sensor) can be used to generate the photons. Typical PPG sensors use inorganic and / or organic light emitting diodes (LEDs). D), laser diode (LD), microplasma source, etc. The PPG sensor may include a photon emitter. and / or noise criteria for attenuating motion artifacts in the PPG waveform data. A motion sensor may also be provided for providing such purposes.

[0043] As used herein, the terms "sensor," "sensing element," and "sensor module" are used interchangeably. " are interchangeable and refer to information from the subject's body (e.g., physiological information, body motion, etc.) and / or environmental information in the vicinity of the subject. Sensors / sensing elements / sensor modules are: That is, the detector element, the emitter element, the processing element, the optical components or optomechanics, the sensor The device may include one or more of the following: shark mechanics, mechanical support, support circuitry, etc. Both single sensor elements and clusters of sensor elements are referred to as sensors, sensing elements, or sensor elements. The sensor / sensing element / sensor module detects information. The goal is to both learn about the network and process that information to derive one or more metrics. It can be configured as follows.

[0044] As used herein, the term "processor" may be localized and / or distributed. It broadly refers to a signal processing circuit or computing system, or a calculation method that can perform the calculation. For example, localized signal processing circuits are essential for the design of wearable biometric monitoring devices. It may include one or more signal processing circuits or methods that are substantially localized at the location. Examples of such devices include earpieces, headpieces, finger clips, toe clips, Hand and foot bands (arm bands or leg bands, etc.), ankle bands, wrist bands, finger ( For example, finger or toe bands, nose bands, sensor patches, jewelry, patches, accessories Distributed processing circuits Examples are the "cloud," the Internet, remote databases, remote processors, a computer, a plurality of remote processing circuits or computers in communication with each other, or The processing method includes a distributed processing circuit and a localized processing circuit. The difference between distributed processing circuits is that distributed processing circuits can include non-localized elements, whereas localized processing circuits can include non-localized elements. This means that the circuit can operate independently of the distributed processing system. The processor, microcontroller, or digital signal processing circuitry may be a local system and / or 1A-1C represent a few non-limiting examples of signal processing circuitry that may be found in a distributed system.

[0045] As used herein, the terms "mobile application" and "mobile app" and "app" are used interchangeably and are used on mobile phones, digital computers, smartphones, Computers such as computers, databases, cloud servers, processors, and wearable devices This refers to a software program that can run on a computing device.

[0046] As used herein, the term "health" refers to the physiological state of an organism or the physiology of an organism. This is broadly interpreted as relating to any biological factor or physiological process. Vascular health can refer to the overall state of the cardiovascular system, and cardiovascular health assessment includes blood pressure, V O 2max , cardiac efficiency, heart rate recovery, arterial blockage, arrhythmia, atrial fibrillation, etc. "Fitness" evaluation is a method to evaluate the health of a person by measuring the physical activity of the person. It is a subset of health assessment when it refers to how a condition affects performance. For example, VO 2max The test may be used to assess a person's mortality health or to assess a person's ability to utilize oxygen during exercise. It can be used to provide a fitness assessment.

[0047] As used herein, the term "blood pressure" includes diastolic blood pressure, systolic blood pressure, mean arterial pressure, Blood pressure refers to the measurement or estimate of pressure related to blood flow in a person, such as pulse pressure. Any part of the body where the stimuli are present (i.e., upper arm, chest, subclavian artery, femoral, tibial, flexor) The term "blood pressure" is used throughout this specification. It is abbreviated as "BP."

[0048] As used herein, any device or system may be used in conjunction with another device. remote to another device or system, unless there is a physical connection between For clarity, the term "remote" refers to a remote device. The device is a wireless device or a remote device is communicating with the remote device. This does not necessarily mean that the device is located over a long distance. For example, in some cases, two devices may be connected in a single physical In this case, the term "remote device" may be used. A "port" is a device or system that is separate from another device or system. or have no substantial dependency on another device or system for its core functions. For example, a computer connected by a wire to a wearable device. The data is provided to support the assumption that these two devices are separate and / or compatible in terms of core functionality. Since it is not substantially dependent on the network, it can be considered a remote device.

[0049] As used herein, the terms "sampling frequency", "signal analysis frequency", and "Signal sampling rate" is interchangeable and refers to the rate at which the signal is obtained from a continuous sensor or sensing element. The number of samples taken per second (or other time unit) (e.g., the number of thermocouples in an eardrum temperature sensor) Sampling rate of pile output or sampling rate of PPG signal from PPG sensor (This refers to

[0050] It should be noted that there are references in this specification to "algorithms" and "circuits." An algorithm is a set of instructions that has sequential steps and logic that can be stored. A circuit refers to a set of computational instructions, such as a program, whereas a circuit refers to a set of instructions that can be implemented in the digital domain, analog domain, and / or or physical components capable of implementing such logic operations in the quantum domain. and / or traces (or paths). These circuits may typically include electrical circuits. Alternatively, it may include elements that are photonic, electromagnetic, magnetic, acoustic, quantum, etc. do.

[0051] To address these limitations, the method and apparatus according to the present invention provides a method for detecting blood pressure estimates (and / or These include EEG estimates, respiratory metric estimates, core body temperature estimates, estimated blood glucose levels, etc. and other biometric estimates) through real-time adaptive prediction models. These methods and apparatus provide for the continuous generation of PPG measurements of a subject. The measured value is combined with at least one BP (or other biometric) measurement of the subject. The system uses the data to update the predictive model for that subject in real time. The estimated BP (or other biometric) of the subject is more accurate (than before the update). The method of the present invention is based on PPG data and BP (or other biometric) data. and a computing system configured to process the data to improve the accuracy of the estimation. That is, the model assumes that the BP estimates are a function of the PPG features. The method is configured to generate a BP estimate for a given set of PPG input features such that The model parameters can be adjusted for repeated Updated over time as resulting BP measurements are processed to allow for model error improvement The PPG sensor can be wearable and therefore can be used by the subject. The PPG sensor may be integrated into a device or component that is in close proximity to the skin of the patient. The sensor may be an imaging sensor (e.g., a camera), a remote sensor, etc., as described later in this specification. It can also be a stand-off sensor such as a can sensor (e.g., radar, Doppler, etc.). do.

[0052] In some cases, the computing system may be configured to receive and transmit the signal from an ear-worn device (e.g., Hearables / Hearing Aids) 10, limb-worn (e.g., wrist, arm, leg) devices It can be worn as a band 12, a patch 14, or a finger clip 16. Alternatively, other form factors such as finger-worn devices (e.g., on fingers or toes), clothing, etc. The computer may include a computing system.

[0053] A key advantage of biometric estimation (over biometric measurement) is that estimation is continuous. Biometric measurements are discrete and can be automated and painless, whereas Blood pressure measurement using a cuff-based BP monitor or finger prick The problem is that some tests (such as blood glucose measurement using a blood sample) can be cumbersome. Therefore, the measurement sensitivity of the biometric estimation is inferior to that of the actual biometric measurement. Even if there are cases where the measurement is not accurate, the ability to provide a "good enough" estimate (between the actual measurements) is Strength can outweigh the downside of potentially lower acuity.

[0054] These wearable PPG devices 12 to 16 are connected to a blood pressure cuff 18 (see FIG. 5). Blood pressure monitoring, such as that shown on the arm of a subject wearing a PG earpiece 12 It can communicate (e.g., electrically, optically, or wirelessly) with the device; or Alternatively, the blood pressure monitoring device may be a separate device. One of these is a stand-off device, such as an electromagnetic wavelength Doppler-based detection system. or imaging system (i.e., camera). Many known to those skilled in the art (ultrasound, Other blood pressure monitoring devices can be used because the device is equipped with a venous line, arterial line, etc. In another embodiment, the PPG and BP measurements are The measurement is received from the same device that measures both the measurement and the readout. One such device Specific examples include cuff-based BP monitors with integrated PPG sensors.

[0055] In some embodiments of the present invention, referred to as the adaptation process, Multiple BP measurements from a cardiogram or other BP monitoring device and PPG measurements were taken. Both are processed to improve the accuracy of BP estimation. A computing system (e.g., 100, FIG. 1) processes the BP and PPG measurements. Once the model is autonomously optimized and updated for the subject (Figure 6), the blood pressure measurement device 18 (e.g., cuff-based BP monitors) may no longer be necessary, and continuous PP G-based BP estimates can now be generated in real time with updated models. In such cases, this adaptation period may last for a day, a week, a month, or even longer. is recalibrated with each new BP measurement at several times of the year (as shown in Figure 12). In principle, human PPG data can be used as a long-term calibration. and their BP relationships do not change, and PPG data are obtained from the same location on the body in the same way. collected and for that person indefinitely, as long as the BP estimation accuracy remains sufficient for the desired use case. A single calibration may be sufficient for this purpose.

[0056] Alternatively, BP measurements can be accepted and processed periodically. This is called the augmentation process. Adaptive prediction models based on BP over time (e.g., obtained from a cuff-based BP monitor) In this enhancement, the adaptive prediction model according to the embodiment of the present invention can be used. The update of the data is repeated continuously several times per hour with new BP measurement updates each time. can be done.

[0057] Triggers for model updates, either adaptive or enhanced (Figure 6), can be provided through a variety of methods. Examples of autonomous triggering paradigms are: 1) set timing protocols; 2) triggers based on motion detection or activity monitoring; and ) triggers based on removal, reattachment, or movement of the device from the body; and 4) triggers based on physiological conditions. and 5) triggers based on detection of an error, but The following are not intended to be limiting. Trigger processing for wearables based on these paradigms An example of the theory has already been presented in U.S. Pat. No. 9,538,921. The entire contents of these autonomous trials are incorporated herein by reference. The data may be generated internally in the computing system 100 (or may be generated by an external device or the like) or may be generated by a third party such as a third party. The trigger can also be generated externally (such as by external command data as shown in Figure 1). The main advantage is that the biometric measurement (calibration) can be performed while still maintaining the biometric estimation accuracy. Significant power savings can be realized if the number of measurements (such as field-based BP measurements) can be minimized. Moreover, some biometric measurements, such as cuff-based BP measurements, This can be burdensome for the user and therefore may cause problems in maintaining the accuracy of the biometric estimation. Reducing the frequency of measurements while doing so can be very beneficial. The computational power of the wearable computer determines and executes such autonomous triggers. It has been found that this is quite sufficient.

[0058] Triggers based on predefined timing may be based on fixed or user adjustable parameters. This paradigm is used when subjects are (e.g., in a hospital bed) Especially in stationary hospital use cases and similar non-mobile use cases (in hospitals) It can be helpful.

[0059] Motion-based triggers can be implemented using motion sensors (e.g., accelerometers, imaging systems, or other motion sensing device or component) detects activity above or below a threshold. This type of trigger can be achieved by detecting the movement of the This can be particularly useful for ambulatory monitoring of the body's motion status. If evidence indicates an increased frequency of occurrence or excessive activity, the model may be The computing system 100 can be triggered to perform updates and more frequently. The biometric measurement device can be triggered to take a measurement. On the other hand, the motion conditions can help ensure that the accuracy of the measurement is maintained. If the individual demonstrates that they are resting or that excessive activity is infrequent enough, the model is updated. The computing system 100 can be triggered to update less frequently and the measurement The biometric measurement device can be triggered less frequently. ,This trigger can be activity-dependent, as opposed to a motion threshold. For example, The autonomously determined activity states of “running” or “walking” require more frequent model updates. It is possible to trigger new and biometric measurements, whereas the decision is made autonomously. The activity states of “rest” or “sitting” require less frequent model updates and biometric measurements. The method of determining the activity state by accelerometer data or imaging data can be used to trigger the The method is well known to those skilled in the art, such as in U.S. Pat. No. 10,610,158. The US patent is incorporated herein by reference in its entirety.

[0060] A wearable device according to an embodiment of the present invention is removed from a subject and then reattached. When the device is attached or moved on the subject, the biometric Trigger the measurement device and the computing system, respectively, to take another biometric measurement. , the biometric estimation model can be updated. This autonomous trigger is If the wireless device is temporarily disturbed or removed from the body, This can help to regain a constant accuracy. Autonomy of removing, moving, or reattaching the device The practical decision is to process the PPG data to determine the signal quality or to use a wearable PPG. Other biometry that may change when the G device is placed differently along the body This can be done by determining the wearable device’s wearable parameters. The method of autonomously determining how the device is being worn by detecting PPG and motion is, for example, For example, U.S. Patent No. 9,794,653, U.S. Patent No. 10,003,882, U.S. Patent No. 10,512,403, and U.S. Pat. No. 10,893,835. The contents of these U.S. patents are incorporated herein by reference in their entirety. This shall be done.

[0061] Changes in physiological state can also be detected autonomously and tracked using other biometric measurements. It can be used to update triggers and biometric estimation models. processing sensor data (or other biometric data) to determine stress status, cardiac status, This update of physiological status can be used to generate an assessment of respiratory status, etc., and can be used to measure another BP. It can be used as an autonomous trigger for updating rows and model parameters. As an example, heart rate variability data from a PPG sensor (or other suitable biometric sensor) The data is processed to detect changes in the stress state of the person (e.g., stress has increased significantly or This can provide an autonomous trigger. As a specific example, data from a PPG sensor (or other suitable biometric sensor) The data is processed to obtain the subject's respiratory information (respiratory rate, respiratory volume, or respiratory regularity (periodicity) / irregularity). It is possible to generate a wide variety of signals (e.g. non-periodic), which can provide an autonomous trigger. For example, large changes in respiratory rate, volume, or regularity can provide autonomous triggers. This is because the transfer function between PPG information and the subject's blood pressure depends on the respiratory dynamics. This may be particularly important for the accuracy of the present invention in outdoor environments. A method for autonomously determining a change in a physiological state by using the above-mentioned method is disclosed, for example, in U.S. Pat. No. 8,157, 730, U.S. Patent No. 8,929,966, U.S. Patent No. 9,427,191, U.S. Patent U.S. Patent No. 10,413,250, U.S. Patent No. 10,893,835, and U.S. Patent No. 11 No. 5,058,304, the contents of which are incorporated herein by reference in their entirety. The entire disclosure is incorporated herein by reference.

[0062] Similarly, if an error (such as an operational error code) is detected by the computing system, Automatic triggering of biometric measurements and model updates can be initiated. Accurate monitoring helps ensure that the system is robust against operational glitches. In one particular example, a computing system may include a BP model providing a PPG-BP model. If you receive an error code that the monitoring auto cuff device has stopped inflating In practice, this is equivalent to a system reset followed by another BP measurement and another PPG-BP monitor. Dell updates and triggers.

[0063] Figure 12 shows real data collected from a human subject in a biometrics testing lab. 1 shows an example of an embodiment of the present invention utilizing an automated BP cuff (upper arm motion). The device is fitted with an ear PPG sensor, an arm PPG sensor, and a wrist PPG sensor. A head PPG sensor was also attached (however, for simplicity, only ear PPG data is shown in Figure 12). (Not shown). To compare the present invention with the volume clamp method, subjects The arm clamp device was also placed on the index finger of the arm where the BP cuff was located. The sequence includes a period of subject rest followed by a period of subject activity. That is, to increase the subject's BP, the patient was asked to perform a BP measurement while the BP and PPG measurements were in progress. Subjects were asked to push against a stationary barrier with their legs for several seconds (isometric Trick leg press).

[0064] Subjects were then asked to complete an isometric leg press to lower BP. The subjects were asked to relax by using a cuff-based BP monitor. It is shown as a vertical line L1, the uppermost point of line L1 represents the systolic BP of the subject, and the The lower dot represents the subject's diastolic BP, but (depending on the calculation system) every 60 to 90 seconds During the initial calibration phase of approximately 300 seconds, the cuff-based BP monitor Multiple values ​​from the monitor are processed along with multiple PPG readings to produce multiple PPG estimates ( The reading is shown in the same format as the cuff-based BP monitor reading, indicated as a thin vertical line L2. ) were generated. However, these estimates are based on the assumption that the parameters of the adaptive prediction model are The model is updated to improve accuracy during the calibration phase, and this model accuracy is By the end of the study, the model accuracy was set to be equivalent to that of cuff-based BP monitors, so users Not reported.

[0065] Following the calibration phase, continuous BP estimates are calculated by adjusting the model parameters with each new BP measurement. More precisely, the remaining cuff-based BP monitor measurements were generated without updating the The results show excellent tracking between PPG model estimates and cuff-based BP monitor measurements. The PPG estimates shown in Figure 12 are ear PP Although this is from the G sensor only, the same performance can be achieved with the PPG sensor for the wrist and the PPG sensor for the arm. It should be noted that it has been found that the PG sensor can provide For the hand PPG sensor and arm PPG sensor, when the blood pressure cuff is inflated and deflated , (when the wrist sensor and / or arm sensor are worn on the same arm as the cuff) Meaningful PPG-BP estimation is no longer feasible during cuff-based BP monitor measurements There is a period during which blockage can affect blood flow.

[0066] The test sequence in Figure 12 was repeated for several subjects. PPG-BP estimation (also called BP measurement estimate, or PPG-eBP) compared with the The performance of the volume clamp device is shown in the table in FIG. 14. The mean absolute difference in PPG-eBP was 0.01 during the isometric leg press period and during the rest period. In both cases, the mean absolute difference is universally lower (better) than that of the volume clamp. For each subject, both the 5-minute and 10-minute calibration periods were examined (obtained from FIG. 14). Small improvements in the PPG-BP model were observed over a longer calibration period. It should be noted that

[0067] FIG. 13 shows an adaptive prediction of a subject wearing a PPG sensor according to an embodiment of the present invention. The model's BP estimates over time are shown in plot 30. The actual blood pressure measurements (readings) are shown by data points 40. The accuracy of the BP estimation is The adaptive prediction model improves over time as it is updated with BP measurements. This is shown in FIG. 13. The difference between plot 30 and data point 40 over time In FIG. 13, the PPG-BP estimation plot 30 shows the estimated frequency in seconds as However, the estimated frequency does not need to be fixed in the present invention, and the distribution of this BP estimation is Resolution depends on the use case (e.g., the BP estimation accuracy or resolution requirements of the use case) and the benchmark. The accuracy and resolution of the BP measurement device and the features selected for the adaptive prediction model The frequency of generation (Figure 6) can be increased or decreased depending on the frequency of the measurement. For example, if the BP measurement device is It has sufficient accuracy and resolution to adequately train an adaptive predictive PPG-BP model. As far as possible, the estimated BP pulse trace (i.e., with a resolution much smaller than 1 second) The present invention generates complete "beat-to-beat" BP waveforms with can be done.

[0068] The present invention is not limited to PPG-based BP estimation, and may be used in other PPG-based binarizations. It should be emphasized that the method can also be applied to metric estimation. ,Other measurement modalities other than BP measurements are also benchmarks and ,which update the adaptive prediction model. Such biometric measurements and their respective Non-limiting examples of metric estimations include respiratory rate, heart rate, cognitive load, and intent (e.g., For example, the intention to perform a mental or physical action), cardiac output, cardiopulmonary function, cardiac condition or medical condition ( Cardiac arrhythmia, premature contraction, cardiac disorders, cardiac disease, plaque accumulation, etc.), gas exchange dynamics, blood tests Body composition (e.g., blood glucose level, blood urea level, bilirubin level, cholesterol level) Additional examples of other measurement and estimation modalities include: Can monitor ECG, EEG, EMG, EOG, blood flow, chest impedance, auscultation monitoring, arterial line data, blood test data, etc. Specific examples include: can monitor the subjects' EEG readings while they sleep and examine their brainwave patterns. However, EEGs are notoriously uncomfortable to wear, especially while sleeping. Therefore, it is more desirable to monitor a person's EEG using more comfortable techniques such as PPG. Therefore, a set of EEG electrodes is used to generate the EEG signal that is fed into the adaptive prediction model. It can provide G measurement data and simultaneously monitor other sensor modalities such as PPG. The PPG data is then processed and fed into an adaptive predictive model (without the need for EEG data). In other words, the PPG model can be used to estimate the EEG. When adapted to the data, the subject's real-time EEG estimates are based on the real-time PPG data. In one embodiment, the EEG sensing is performed using only the MRI machine.

[0069] Method for generating a biometric estimate of a subject using an adaptive predictive model FIG. 2 illustrates the present invention using a real-time adaptive predictive model implemented by a computing system. 1 illustrates a method for generating a biometric estimate of a subject according to some embodiments of the present invention. The method includes receiving real-time PPG data from a PPG sensor that measures the subject's PPG information. and a step of receiving the blood pressure monitoring device within the specified period to measure the subject's blood pressure. and receiving the real-time blood pressure measurement value from the user during the reception period (block 2). 00). Features are generated from the received PPG data (block 202). The updated feature values ​​and blood pressure measurements are stored in memory. If an update is ready ( Block 204) processes the stored features and the stored blood pressure measurements to generate an adaptive prediction model. Adaptively optimizes the model by generating updated model parameters that reduce the estimation error of the model. The predictive model may be updated in real time (block 206). The adaptive predictive model generates a biometric estimate of the subject (block 20). 8).

[0070] FIG. 3 illustrates a method of generating a biometric estimate of a subject according to another embodiment of the present invention. A computing system (e.g., 100, FIG. 1) may be configured to measure the PPG level of a subject using a PPG sensor attached to the subject. (e.g., 12 to 16, Figure 5) to accept real-time PPG data (Block 2 10) The computational system uses PPG data to derive the subject's biometric information from an adaptive predictive model. A biometric estimate is generated (block 212). An example of a biometric estimate is the subject's Although the main focus is on blood pressure estimation, various other biometrics can be estimated, as described below. The monitoring device attached to the subject (e.g., blood pressure cuff 18, Figure 5) The real-time biometric measurement is accepted by the computing system (block 21 4), the computing system updates one or more parameters of the adaptive predictive model (block 2 16) For example, real-time blood pressure readings may be obtained from a blood pressure monitoring device. The readings are taken from the subjects and are then predicted by an adaptive prediction model using the PPG data. Adjust the adaptive prediction model so that the resulting blood pressure estimates are closer to the actual blood pressure readings Used for:

[0071] It is understood that the steps illustrated in FIG. 3 do not have to be performed in the order illustrated. For example, collecting real-time biometric measurements (block 214) may include This may occur prior to or simultaneously with the PPG data acquisition (block 210).

[0072] FIG. 4 illustrates a method of generating a biometric estimate of a subject according to another embodiment of the present invention. A computing system (e.g., 100, FIG. 1) may be configured to measure the PPG level of a subject using a PPG sensor attached to the subject. (e.g., 12 to 16, Figure 5) to accept real-time PPG data (Block 2 20) The computational system uses PPG data to derive the subject's biometric information from an adaptive prediction model. A biometric estimate is generated (block 222). An example of a biometric estimate is the subject's Although the main focus is on blood pressure estimation, various other biometrics can be estimated, as described below. There is no determination as to whether the biometric estimate is above or below the threshold. (Block 224). For example, a healthy blood pressure range is typically 120 mm systolic pressure or 130 mm systolic pressure. The mean diastolic blood pressure is less than 80 mmHg. If the examiner's systolic blood pressure falls below 90 mmHg and / or the diastolic blood pressure falls below 60 mmHg, If systolic blood pressure falls below 1.5 mmHg, medical intervention may be required. If your blood pressure exceeds 130mmHg and / or your diastolic blood pressure exceeds 90mmHg may also require medical intervention.

[0073] If the biometric estimate is above or below the threshold (block 224) records real-time biometric measurements on a biometric device attached to the subject. 1. Calculation system from blood pressure monitoring device (e.g., blood pressure cuff, Figure 5) Upon receipt (block 226), the computing system may then generate one or more parameters for the adaptive predictive model. Update the meter (block 228). For example, real-time blood pressure readings are updated in the blood pressure monitor. The readings are taken from the subject by a monitoring device and are calculated using PPG data. This allows blood pressure estimates made by the adaptive predictive model to be closer to the actual blood pressure readings. The computing system is used to adjust an adaptive predictive model that estimates the biometric information. Remote alerts when the load estimate is above or below a threshold. The received signal is then sent to the remote device (block 230).

[0074] It should be noted that a calibration cuff reading is required to improve the accuracy of the estimate. The BP estimate does not have to be outside a certain range to be used to BP may be within normal range and subsequent cuff readings will refine accuracy. The adaptive prediction model is different from the comparison of BP values ​​with the threshold. Regardless, it can simply be updated based on the timed cuff-based reading. Similarly, adaptive predictive models can be adapted to detect changes in detected activity levels (e.g., accelerometer measurements). due to changes in body motion (detecting changes in body motion more accurately) or due to other sensor readings In some cases, the information is updated.

[0075] The remote devices can be a healthcare provider's smartphone, a nurse's station within a healthcare facility, or any other device capable of alerting medical personnel about the subject's condition. The alert may be generated by a blood pressure monitoring device (e.g., a blood pressure cuff). In addition, alerts can be sent to the blood pressure monitoring device. It can also be created by

[0076] Although block 224 is shown as a "threshold" decision, block 224 is a model update Note that this can be replaced with conditional logic that determines what should happen. For example, block 224 may include threshold logic based on one biometric estimate. Instead, the thresholds of multiple biometric estimates (e.g., already stored in memory) The logic may include logic for determining the presence of a processing pattern. This pattern is characterized by a series of consecutively high (above normal) or consecutively low (above normal) BP estimates. (below BP estimates), and determining this pattern triggers a model update. In another non-limiting example, the pattern can be above normal or The performance may include an average of multiple biometric estimates that are determined to be below the If a turn is determined to exist, an update of the model can be triggered.

[0077] The method shown in FIGS. 2 to 4 can be performed by a computing system 100 shown as an example in FIG. The computing system 100 is 1) PPG data from a PPG sensor that measures the subject's PPG information (and, if necessary, This is for example the auxiliary sensors such as motion sensors, environmental sensors, etc. The motion sensor data or the environmental sensor data obtained by the sensor are the motion sensor data or the environmental sensor data obtained by the sensor. and accepting blood pressure data from a blood pressure monitoring device that measures the blood pressure of at least One data bus 102; 2) Accept PPG data from the PPG sensor during the acceptance period, and (automatically) Receive blood pressure measurements from a blood pressure monitoring device (such as a blood pressure cuff or arterial line monitor) Attach, generate feature quantities from the received PPG data, and store these feature quantities in memory , store the blood pressure measurement value in memory, and process the stored feature quantity and the stored blood pressure measurement value , generate model parameters updated so that the estimation error of the adaptive prediction model is reduced to update the current parameters of the adaptive prediction model, and generate a biometric estimation of the subject by executing the updated adaptive prediction model A calculation circuit and instruction 104 and can be provided.

[0078] <Receiving of PPG data, receiving of BP measurement values> Refer to Figure 2 again. The update of the adaptive prediction model (block 206) requires at least two inputs: the PPG feature quantity and at least one BP measurement value. These data can be received over a "receiving period". The receiving period is the period during which the computing system 100 of FIG. 1 receives at least one PPG waveform and at least one BP measurement value that is temporally related . The received PPG data can be received as digitized data. At this time, prior digital processing steps may be required to sample the PPG data by digital processing (e.g., at frequency "f ") before it is received by the computing system 100 of FIG. 1. Similarly, BP data can be received digitally. At this time, prior digital processing steps may also be required s . On the other hand, due to the discrete nature of cuff-based BP measurement values, discrete BP values are received instead of streaming continuous BP values . ​​​It can be attached. The PPG data and BP measurement values are related in time (the times are close enough to each other), but these measurements do not need to exactly coincide in time (be truly simultaneous). This is because BP does not change dramatically within a few seconds in many situations, and several PPG waveforms can be received during these few seconds. Moreover, PPG data can be continuously collected, while cuff-based BP measurement values may require more time than 60 to 90 seconds between measurements. Therefore, it may not be practical to perfectly align each PPG waveform with the BP waveform (or BP measurement value) that occurs simultaneously. For a normal ambulatory resting state, the temporal relationship between PPG data and BP measurement values within about 30 seconds has been shown to be sufficient for continuous tracking. This timing may be longer or shorter depending on other factors that can affect the subject's activity state, the dynamics of the subject's cardiac output, or the rate of change of the subject's BP or other physiological changes. In this way, the temporally related PPG data and BP measurement data can be stored in a memory (such as a memory buffer) by a computing system. to each other), but these measurements do not need to exactly coincide in time (be truly simultaneous). This is because BP does not change dramatically within a few seconds in many situations, and several PPG waveforms can be received during these few seconds. Therefore, it may not be practical to perfectly align each PPG waveform with the BP waveform (or BP measurement value) that occurs simultaneously. For a normal ambulatory resting state, the temporal relationship between PPG data and BP measurement values within about 30 seconds has been shown to be sufficient for continuous tracking. Moreover, PPG data can be continuously collected, while cuff-based BP measurement values may require more time than 60 to 90 seconds between measurements. This timing may be longer or shorter depending on other factors that can affect the subject's activity state, the dynamics of the subject's cardiac output, or the rate of change of the subject's BP or other physiological changes. In this way, the temporally related PPG data and BP measurement data can be stored in a memory (such as a memory buffer) by a computing system. to each other), but these measurements do not need to exactly coincide in time (be truly simultaneous). This is because BP does not change dramatically within a few seconds in many situations, and several PPG waveforms can be received during these few seconds. Moreover, PPG data can be continuously collected, while cuff-based BP measurement values may require more time than 60 to 90 seconds between measurements. Therefore, it may not be practical to perfectly align each PPG waveform with the BP waveform (or BP measurement value) that occurs simultaneously. For a normal ambulatory resting state, the temporal relationship between PPG data and BP measurement values within about 30 seconds has been shown to be sufficient for continuous tracking. This timing may be longer or shorter depending on other factors that can affect the subject's activity state, the dynamics of the subject's cardiac output, or the rate of change of the subject's BP or other physiological changes. In this way, the temporally related PPG data and BP measurement data can be stored in a memory (such as a memory buffer) by a computing system. to each other), but these measurements do not need to exactly coincide in time (be truly simultaneous). This is because BP does not change dramatically within a few seconds in many situations, and several PPG waveforms can be received during these few seconds. Moreover, PPG data can be continuously collected, while cuff-based BP measurement values may require more time than 60 to 90 seconds between measurements.

[0079] <Generation of PPG Feature Quantities> The received PPG data is processed to generate a plurality of real-time PPG feature quantities (block 202, Figure 2). These feature quantities are "n" different characteristics from each other. Examples of feature quantities include, but are not limited to, feature quantities based on time domain or transformation. Non-limiting examples of time domain feature quantities are the amplitude of the PPG, the upper envelope of the PPG. Examples of feature quantities include, but are not limited to, feature quantities based on time domain or transformation. Non-limiting examples of time domain feature quantities are the amplitude of the PPG, the upper envelope of the PPG. Examples of feature quantities include, but are not limited to, feature quantities based on time domain or transformation. Non-limiting examples of time domain feature quantities are the amplitude of the PPG, the upper envelope of the PPG. and / or lower envelope, systolic peak separation and diastolic peak separation and / or relative amplitude, convergence Systolic peak-to-trough separation and dicrotic notch peak-to-trough separation Trough separation, the separation of time periods between important features (peaks or troughs) in the PPG waveform. Similarly, the PPG data may be processed to generate derivatives (e.g., first derivatives, etc. It is possible to generate the derivative waveforms (functions, second derivatives, third derivatives, etc.) or integrals. and / or generate time domain features of the integrated waveform (i.e., amplitude of peaks or troughs, relative Generate features such as amplitude, upper and / or lower envelope, and temporal peak separation The features based on the transformation are spectrum features, wavelet features, and tiger features. Features based on the Kaiser-Kaiser energy (KTE) operator, Plett transform feature, Noiselet transform feature, Spaceogram feature , Shapelet features, derivative features, integral features, principal component analysis (PCA) This can include component analysis features, etc.

[0080] Generating features from the received PPG data is a sliding time window. DoΔt w Within the acceptance period, the feature generation interval (time point) t=k i Generate features in The feature quantity can be calculated at any time by a computing system (Fig. 6). However, meaningful PPG features can be generated using at least one complete PPG waveform. and preferably enough PPG data is stored in memory to process multiple PPG waveforms. For example, a feature must be stored in memory with a length of Δt w The time period in advance ( That is, the width is Δt w The data collected and buffered over a time window By processing the digitalized PPG data, we generate features on the feature generation window at t=k. i To This feature generation window is backed by a FIFO (first in, first out) buffer. In a FIFO buffer, the PP G data is continuously acquiring new samples of data and deleting the oldest sample of data. The feature generation process is performed in the time domain. This buffered PP is then converted into a PP2 buffer in the time domain or by conversion of the stored time domain data. As noted above, various different time domain or transforms may be used to process G. The based processing method can be used to generate PPG features. Non-limiting examples of transforms that generate Isa Energy Operator, Chirplet Transform, Noiselet Transform, Spaceogram, Non-limiting examples of time domain processing include: PPG amplitude, upper and / or lower envelope of PPG, systolic peak separation and diastolic Generates peak separation, systolic peak-to-trough separation and dicrotic peak-to-trough separation. Non-limiting examples of transforms and time domain processing that can be used include: No. 10,856,813 and PCT Application No. US20 / 49229. These patents and applications are hereby incorporated by reference in their entireties. This shall be done.

[0081] Before generating a BP estimate (or any other biometric estimate), PPG features (characteristic The features of the BP measurement (or other biometric features) are actively normalized (e.g., weighted) and Smoothing of PPG-based BP estimation (or other biometric estimation) using the It should also be noted that this can help ensure accurate temporal tracking. The normalization method uses the integrator of stored features (e.g., previous PPG features stored in memory). The goal of normalization is to process the total number of features and normalize them by these statistics. generating statistics on the historical data by processing the historical data over time; This can be done by normalizing these statistics. This normalization process The time when each feature is generated (e.g., t=k in Figs. 6 and 8) i ) can be used to update Alternatively, the regularization can be done at each model update (e.g., t=u in Fig. 6). j ) can be used. There are numerous normalization methods known to those skilled in the art, some examples are z-normalization, These include z-norming, min-max normalization, mean normalization, etc. A common normalization method is to use the Cochran equation for pooled statistics. To use the model equation with each model update, we use the j-1 at t=u j (in By processing (pooling) with the statistics of Therefore, the probability of the data generated by the Cochran equation can be normalized (weighted). The mean and standard deviation can be used as a basis for normalizing the characteristic features. As a specific, non-limiting example, using z-norming, the value of a characteristic feature can be It can be normalized by the mean and standard deviation produced by the Cochran equation, For example, the mean and standard deviation are calculated based on the past updates (e.g., t=u j-1 in Calculate the mean and standard deviation of the features following the past update (e.g., t=u j ) feature average It is generated by weighting with the mean and standard deviation.

[0082] The aforementioned feature statistics themselves are used as features to the adaptive predictive model in accordance with an embodiment of the present invention. This allows for a smoother (e.g., BP estimate vs. measured) This can help provide smooth tracking.

[0083] As part of (or prior to) feature generation, the received sensor information (e.g., PP G-sensor data) and / or accepted biometric measurement data (e.g., BP It should be noted that pre-processing of the measurement data may be required. Modifying accepted data to prevent "invalid" data; generating a reliability score for the data identifying "good" data or classifying data for further processing PPG data (including associated motion sensor data) Various pretreatment methods for the above have been published and are well known to those skilled in the art. The processing methods are described in U.S. Patent No. 10,834,483, U.S. Patent No. 10,798,471, U.S. Patent No. 10,631,740, U.S. Patent No. 10,542,893, U.S. Patent No. 1 No. 0,512,403, U.S. Pat. No. 10,448,840, U.S. Pat. No. 9,993,2 No. 04, U.S. Patent No. 10,413,250, and PCT Application No. 20 / 49229. All of these U.S. patents and PCT applications are incorporated herein by reference in their entirety. and are incorporated herein by reference in their entireties. Both passive and active methods of eliminating noise can be used. It should be noted that the optimal preprocessing may depend on the features. For example, Regarding G data, in the case of spectral domain features, "D It may be desirable to remove or attenuate the "(C) component" (e.g., the non-pulsatile component). The DC component may be important for other features (e.g., time domain features), and the DC component may be important for other features (e.g., time domain features). In some cases, the amount of PPG sensor data is a feature in itself. The motion data may include motion data from optical sensor readings. It should also be noted that the 3D motion artifacts can be reduced by using the The motion sensor can be integrated into the PPG sensor or can be used in combination with the PPG sensor. Motion sensor data can also be processed as features. This can be done.

[0084] Pre-processing of the biometric measurement data can also be useful. For example, In the example, the BP measurements from the BP cuff are represented as discrete values ​​for the systolic BP measurement and the diastolic BP measurement. In some use cases, this data may be provided via APIs (application application programming interface (API) They can be made available to computing systems through application-specific interfaces. However, in some use cases, the BP accepted by the computational system in Figure 1 The measurement data must be processed to extract BP measurements before the present invention can be implemented. It can contain data streams that may need to be processed (such as raw data streams). Cut.

[0085] <Update model parameters> As shown in Figure 8, a biometric estimation (BE) is generated. The adaptive predictive model 300 can take the form BE=f(F,S), where: F is the time t=k i A set of n generated features (e.g. normalized features) where S is a set of statistics on F. The function f(F,S) is a biometric function. The method may include a transfer function that links the trick estimates with the features and statistics described above. For each new BP measurement (or biometric measurement) taken, a new update time t = u j To The adaptive predictive model 300 can be updated at any time (as shown in FIG. 7). Updating the rule includes updating one or more parameters of the adaptive predictive model 300.

[0086] Depending on the type of model used, the model parameters can vary. For example, In a regression model, the model parameters include at least one coefficient to the regression model. Non-limiting examples of suitable regression models include linear models, SVMs, random forests, and These models can include strings, neural networks, decision trees, or a combination of these models. Other types of models besides regression models can also be used, non-limiting examples include You can also use a classifier (convolutional neural network (CNN)). Combination of classification and regression (such as can be used in a general neural network) Model updates can be done by using characteristic features (e.g., normalized features). The estimation error of the adaptive prediction model 300 is calculated by processing the stored blood pressure measurements and the feature values ​​(features) of the adaptive prediction model 300. For example, the regression model may include generating updated model parameters to reduce The model can be solved for recent BP measurements (or biometric measurements), After that, the model coefficients can be updated. Alternatively or additionally, a gradient-based optimization procedure can be used. Methods (Traditional Gradient Descent, Adam, Momentum, AdaGrad, RMSProp, AMSgrad, etc.) to generate a model function with each new BP measurement. The number can be updated.

[0087] The adaptive prediction model is updated in real time based on the stored recent blood pressure measurements (hours). point t=u j ) and the stored previous blood pressure measurement (at time t = u j-1 Related to In one embodiment, this can include processing with the stored The most recent blood pressure measurement and the stored previous blood pressure measurement (or multiple stored previous blood pressure measurements) Interpolation of expected blood pressure measurements between blood pressure measurements taken over time, such as interpolation between Specific examples are summarized with respect to FIG. The blood pressure measurement associated with time u2 and the blood pressure measurement associated with time u j (In this particular case, 3) The blood pressure measurement values ​​related to each feature generation interval t=k are stored in memory and processed. i etc. An interpolation of expected blood pressure measurements for multiple feature generation intervals can be generated. In this case, the adaptive model is updated over multiple intervals t=k i The set of features and the interpolation This can include updating the model parameters for each BP measurement taken. Thus, There is more information to optimize the regression model than just two blood pressure measurements, which , the tracking of the BP estimate with the actual BP measurement becomes smoother.

[0088] <Generating Biometric (BP) Estimation> As summarized above, there are many different methods that can be used to generate a biometric estimate. There are many model configurations. One of the features used to generate a biometric estimate is The general representation is shown in Figure 8. For the specific case we have discussed for generating a blood pressure estimate, The process of generating BP estimates may include systolic blood pressure, diastolic blood pressure, pulse pressure, mean arterial pressure, or blood pressure. It may be possible to generate other types of pressures related to the flow. The types of blood pressure that can be measured are brachial, thoracic, subclavian, femoral, tibial, radial, and carotid. Virtually any part of the body, including but not limited to the veins, central nervous system (aorta), and brain. Each of these blood pressure estimates can be from the The process can be used to generate the method of FIG. 2 to FIG. 4, but in the subject Ideally, the location of the BP measurement should coincide with the location of the desired BP estimation. If the biometric estimation of the brachial artery includes systolic and diastolic estimation, the BP measurement A constant-value BP monitoring device (ideally) measures systolic BP from the brachial artery. Both diastolic and diastolic BP values ​​will be measured.

[0089] To implement the methods of FIGS. 2-4, a computing system 100 is utilized as shown in FIG. The adaptive prediction model allows for the estimation of the subject's biometrics (in this particular case The computational system 100 for generating the BP estimate is 1) PPG data from a PPG sensor that measures PPG information from a subject and blood pressure of the subject and at least one blood pressure monitoring device that measures blood pressure. A data bus 102 of 2) computing circuitry and instructions 104 for: a) receiving PPG data from a PPG sensor; and b) receiving blood pressure measurements from a blood pressure monitoring device during said period. c) receiving the PPG data within the deadline, and c) generating features from the received PPG data. d) storing the feature amount in a memory; and e) storing the blood pressure measurement value in a memory. f) processing the stored features and the stored blood pressure measurements to estimate an adaptive predictive model; Adaptive prediction models are developed by generating updated model parameters that reduce the error. and g) updating the current parameters of the predictive model. and generating a biometric estimate of the subject by Order 104 and It can be provided with:

[0090] The computing system 100 may be comprised of discrete components, a fully integrated system, or a combination of As a specific example, the computing system 10 0 is a fully integrated microprocessor that performs the computational instructions for the processing steps of Figs. 2 to 4. FIG. 15 shows a block diagram of a device that can be used in accordance with various embodiments of the present invention. FIG. 1 is a block diagram showing details of an example processor P and memory M that can be used. The processor P communicates with the memory M via an address and data bus B. The processor P may, for example, Can be a commercially available or custom microprocessor Moreover, the processor P may include multiple processors. The memory M may be non-transient. 2 as described herein. -Memory device containing software and data used to implement the method of FIG. The memory M can represent the entire hierarchy of devices of the following types: cache ROM, PROM, EPROM, EEPROM, Flash, Static RAM ( SRAM (Static RAM) and / or DRAM (Dynamic RAM) These can include, but are not limited to:

[0091] The memory M stores computer readable program code PC and / or an operating system. It can hold various categories of software and data, such as operating systems, operating systems, etc. The operating system OS is a processor P, a PPG sensor (e.g., 12 to 16, Figure 5) , controls the operation of a biometric monitoring device (e.g., BP cuff 18, Figure 5) and can coordinate the execution of various programs by the processor P. For example, computer readable program code PC, when executed by a processor P, The processor P executes any of the operations shown in the flowcharts of FIGS. It is possible.

[0092] Alternatively, the computing system 100 may be an analog device that processes steps through an analog process. As another example, the computational instructions may be included in a computing system (e.g., a processor As another example, the implementation may be implemented as a software library executed by a Thus, the computing system may include neural circuitry or quantum computing. It can use classical, quantum or neural processors. You can also use a combination of each.

[0093] The various components that enable the system 100 of FIG. The present inventors developed a commercial smart home device that communicates with wearable devices 10 to 16, as shown in Figure 5. It uses calculation instructions (algorithms) executed by software on the Phone 20. Laboratory tests have demonstrated that it is possible to achieve real-time performance that is Therefore, the computational overhead required to execute the method of FIGS. 2-4 by a microprocessor is The source is practical for wearable or portable systems.

[0094] The system accepts data from and sends data to external systems. It has input / output lines (i.e., ports or buses) for communicating with other systems. For example, the input / output lines may be connected to at least one external processor, an external computing system, In one particular embodiment, the generated The biometric estimate can be digitized and externally transmitted via a digital bus 106. In another embodiment, the input / output lines can be It can communicate with one or more transceivers that communicate wirelessly with external systems. Communications arrangements are well known to those skilled in the art.

[0095] If BP estimates are generated by the computational system of Figure 1, the availability of them for use by external systems is To do this, the external system modifies the computation process (i.e., modifies the algorithm). ) information to a computing system. For example, in one embodiment, The BP measurement is recorded by the computer system shown in FIG. If the device can be equipped with a BP cuff that feeds the stem, the BP estimate generated will be an upper arm BP estimate. The BP cuff can include a PP measurement generated by the computing system between BP measurements. A visual screen may also be provided to view the estimated G-BP reading. The PPG processing rate (sampling rate, feature generation interval, update interval, etc.) can be adjusted depending on the interface. ), and this information is then used to 1 as "external instruction data" for use in the computing system of FIG. Additionally, the computational system is aware that sensor estimates may be inaccurate due to motion noise. Provide an external system (i.e., BP cuff) with a warning that a blood pressure monitor is in use or other useful information. Similarly, the system may have feedback from either the computational system or an external system. Also, information about when BP measurement and / or BP estimation begins (e.g., when BP measurement and / or frequency of BP estimates).

[0096] One form of external system data may include subject metadata, which may include The data may be useful in processing biometric estimation in accordance with an embodiment of the present invention. It should be noted that the computing system 100 of FIG. Accepts metadata (i.e. height, weight, age, sex, medication use, etc.) and converts this data This metadata can be stored in memory. Alternatively or additionally, the stored metadata can be used as a feature. The data can be used to create a profile of the subject. Adaptive, optimized for the individual (i.e. across several biometric measures) The key advantage of a user profile is that it can contain model parameters. The profile is "cold start" (i.e., the subject is starting a new estimation and measurement session). This means that it is possible to prevent model adaptation delays caused by In other words, a finite number of parameters (as shown in Figure 12) are used to adjust (calibrate) the subject. A period of time may be required for this calibration phase, and this calibration phase is performed to ensure that the subject's previous model parameters are If stored in memory it can be shortened.

[0097] <Other biometric estimation> As previously mentioned, the system of FIG. 1 and the corresponding methods of FIGS. 2-4 (as well as FIGS. 6, 7, and the supplementary example in Figure 8) can be applied more broadly than the continuous estimation of BP. Continuous physiological estimation can be achieved by the embodiments of the present invention. A metric estimate is generated and a biometric measurement is accepted. In addition, the other core elements of the present invention shown in FIG. 1 and FIG. 2 to FIG. 4 are still valid. This can be done.

[0098] Since PPG information contains a wealth of information about blood flow, blood pressure can be measured by embodiments of the present invention. It is just one of many real-time hemodynamic parameters that can be extracted. i.e., arterial pressure, mean arterial pressure, systolic pressure variation, pulse pressure variation, stroke variation, right arterial pressure , right ventricular pressure, pulmonary artery pressure, mean pulmonary artery pressure, pulmonary artery wedge pressure, left atrial pressure, cardiac output, cardiac index, 1 stroke volume, stroke volume coefficient, systemic vascular resistance, systemic vascular resistance coefficient, pulmonary vascular resistance, pulmonary vascular resistance Various hemodynamic parameters such as (but not limited to) stroke work index, ejection fraction, etc. The parameters can be estimated by embodiments of the present invention. The ability to estimate parameters of depends on suitable measurement devices. For example, To accurately estimate real-time ejection fraction based on morphology, echocardiographic monitoring devices are required. It is necessary to collect time-correlated measurement data from accurate benchmark devices. It is said.

[0099] As just one example, the measurement data may include EEG measurements, and the relationships generated may include: The associated biometric estimation can be an estimation of an EEG evaluation. Non-exhaustive examples include alertness, dominance patterns (e.g., alpha, beta, theta, or delta ), including the subject's intention, identification of abnormality, identification of normality, brain disorders, drowsiness, and wakefulness In this case, among the characteristic features generated from processing the PPG sensor, At least some of the parameters are weighted or weighted significantly higher than those used in estimating blood pressure. is given a small weight because the physiological relationship between EEG and PPG is similar to that between BP and PP For example, the PPG peak location ( Heart rate variability, time location of systolic and diastolic peaks, and time location of dicrotic notch PPG features related to time domain variations in neural networks (e.g., neural network applications) are more closely related to EEG features. can be related to.

[0100] As another example, if the measurement data includes gas exchange (respiratory) analysis measurements, Related biometric estimates may include estimates of gas exchange analysis measurements. Non-limiting examples of exchange analysis measurements include carbon dioxide, oxygen, arteriovenous oxygen gradient, exercise rhythmicity, and the like. Respiratory variation, percentage of carbon dioxide or oxygen in exhaled air, exhaled air volume, metabolic equivalent, maximum ventilation, oxygen intake Intake gradient, end-tidal carbon dioxide or end-tidal oxygen partial pressure, carbon dioxide output, respiratory exchange rate, minute ventilation, dead space ventilation, ventilatory threshold, ventilatory equivalent of oxygen or carbon dioxide, oxygen intake, etc. may include.

[0101] In a similar example, if the measurement data includes arterial blood gas measurements, the associated The biometric estimation may include estimation of arterial blood gas measurements. Non-limiting examples of constants include H+ level or pH level, total CO2 content, total O2 content, acid Partial pressure of hydrogen or carbon dioxide, HCO3 - (Bicarbonate), SBC e , base excess, arterial oxygenation, Includes venous oxygenation, oxygen extraction, etc. When generating gas exchange analysis estimates or arterial blood gas estimates To generate a set of unique signatures of multiple wavelengths of light, simultaneous multi-wavelength (MWL) :multiwavelength) data (streaming PPG data from MWL PPG sensor It may be particularly important to accept PPG data that includes PPG features (e.g., The distribution of the amount of radiation being monitored may depend on the wavelength of light used. Differentially modulated in time depending on the respiratory or blood gas parameters being measured. As just one example, the relative amplitudes of PPG signals for different wavelengths of light may be The width modulates in a unique manner at low versus high levels of blood oxygenation.

[0102] In another example, when the measurement data of FIG. 1 and FIG. 2 to FIG. 4 includes a deep body temperature measurement value, Related biometric estimates that are made can include an estimate of core body temperature. It is known that changes in body temperature, especially heart rate, are related to core body temperature (see, for example, U.S. Pat. No. 6,333,631). See U.S. Pat. No. 5,206,627, which is incorporated herein by reference in its entirety. (which is considered to be part of the PPG data), and therefore the PPG data of the subject is matched with the temperature measurements. There are specific PPG features that can be used to measure core body temperature throughout the course of daily living activities. Since it is difficult to measure continuously, the invention herein is directed to core body temperature measurements. Therefore, the data collected from the PPG device is based on the previously updated adaptive model. Processing allows for a useful method of portable core body temperature estimation.

[0103] In another example, when the measurement data of FIG. 1 and FIG. 2 to FIG. 4 includes a blood glucose level, The relevant biometric estimations can include estimation of blood glucose levels. The subject wears a blood glucose meter (continuous glucose monitor, CGM, etc.) or measures blood glucose levels on his / her own finger. A blood glucose meter can be used to periodically prick the blood sugar level to generate a series of glucose readings. According to an embodiment of the present invention, a computing system 1 00 (Figure 1), and these combined data are processed to generate the PPG A glucose estimate can be generated based on As summarized in PCT Application No. 20 / 49229, which constitutes the present invention, the blood glucose trend There are PPG features related to the orientation of the brain, particularly the respiration action represented in multi-wavelength PPG data. There are features related to changes in the arterial compliance and changes in the arterial pressure. The quantity can provide the basis for the feature quantities in Figure 8, and the statistics for updating the model parameters are , these respiratory function features or arterial compliance features (or other The calibration phase following multiple glucose measurements is completed. Then, once the adaptive model stabilizes for continuous PPG-based glucose estimation (i.e. i.e., no model updates are required), invasive (or minimally invasive) over long durations This allows the patient to be free from invasive (blood glucose monitoring) if additional glucose measurements are required. Long duration (hours, days, or weeks) before being fed into adaptive predictive models Therefore, the accuracy of glucose estimation by PPG during this long duration was comparable to that of CGM or finger glucose. It may be lower than that provided by glucose prick testing. However, the benefits of painless PPG-based glucose estimation are well-suited for many use cases with reduced accuracy. For example, PPG estimation techniques (when calibrated for glucose measurements) , collect glucose measurements to more precisely ascertain glycemic status (i.e., blood sampling) The subject will be informed that a blood sample or other body fluid sample will be taken. It can be particularly useful in predicting or estimating sudden spikes or drops in sugar levels.

[0104] <Imaging purposes> As mentioned above, the present invention can be used in the systems and methods of FIGS. 1 and 2-4. One type of PPG sensor that can be used includes an imaging sensor. Double and stand-off devices (those that are separated from the subject or not worn by the subject) ), and / or worn by the subject (see U.S. Pat. No. 10,623,849). This U.S. patent is hereby incorporated by reference in its entirety for use in a digital camera. Various imaging sensors are known in the art. It is well known to those who are familiar with CCD imagers, CMOS imagers, NMOS imagers, Photodetector arrays and the like. Modern mobile electrification The ubiquity of cameras in Ronix also allows people to wear blood pressure monitors that generate blood pressure readings. This results in various setups for imaging subjects in a sitting position.

[0105] Important benefits can be gained by using an image sensor as a PPG sensor There are multiple biometric points from different body locations (i.e., biometrics across the entire body). It is also possible to simultaneously obtain data on electromagnetic energy Simultaneous data at multiple wavelengths (whether photons are in the visible range or outside) For example, the implementation of the invention described herein may be The morphology, when utilized with an imaging sensor as a PPG sensor, provides at least one blood The measurement data provided by the pressure monitor is used to determine the subject's blood pressure at multiple body locations. In one embodiment, the subject may have a blood pressure cuff attached to their arm (e.g., with an upper arm blood pressure cuff) and this The measurement data is then provided to the systems and methods of FIGS. 1 and 2-4. When processing the BP measurement data, the adaptive prediction model (e.g., 300, FIG. 8) It can be configured to estimate continuous BP at multiple body locations simultaneously. In an exemplary embodiment, PP measurements collected from the brachial artery region (the region from which BP measurements are collected) are The G data can be processed to generate an estimate of BP in that region of the body. The relationship is based on the imaging sensor to allow continuous estimation of blood pressure over various regions of the body. The biomarkers can then be expanded to other body regions within the camera's field of view. The metric estimates can be further processed to generate a comprehensive hemodynamic assessment of the subject. For example, irregular or non-uniform blood pressure can result in insufficient blood flow (i.e., vascular blockage, bleeding, etc.). In addition, blood pressure in the heart relative to blood pressure in the extremities can indicate Processes relative blood pressure (central blood pressure) to indicate peripheral resistance to blood flow or other cardiovascular problems In addition, the assessment of the time dynamics between central and peripheral blood pressure can be used to determine the Pulse wave transit time (PTT) and pulse wave velocity (PWV) ity) can be evaluated.

[0106] It should also be noted that the blood pressure measurement device itself can be used as the remote imaging sensor in the present invention. And in such a case, the advantage of the wearable PPG sensor is the ability to passively evaluate blood pressure according to the BP measurement values obtained from the video (collected by the imaging sensor) for blood pressure estimation. Conversely, the blood pressure measurement device itself in the present invention can be a wearable PPG sensor (i.e., optimized for measuring or estimating BP) or a blood pressure cuff, and the biometric sensor can be an imaging sensor. In such a case, an important advantage of this approach is that new data from blood pressure measurements continue to improve the adaptive prediction model, so that the remote imaging system can continuously and passively improve the BP estimation of the subject (in the background without requiring changes in the subject's behavior). <Alternative forms of PPG> The present invention is also applicable when the biometric sensor data is not PPG sensor data but rather sensor data of different sensor modalities such as electromagnetic, auscultation, electrical, magnetic, mechanical, thermal, etc. (or a combination of these modalities in a sensor fusion approach). In such a case, a more general invention is shown in FIGS. 9, 10, and 11. For the present invention to function properly with these other sensor modalities, those sensor modalities enable a continuous stream of waveform data with respect to accurately and continuously estimating the subject's BP, and the rate of statistical change of the feature quantities is related to the PPG waveform data.

[0107] ​​​​​​​​​​It is important to compare the rate of change of the PPG waveform with that of the auscultatory test. These particular modalities (as described above) have the following characteristics: The most suitable method for estimating blood pressure according to the present invention (compared to the list of It can be said that it is.

[0108] <Definition of forecast> In some cases, the estimates generated by the adaptive predictive model are not truly real-time. can be an estimate of what the model predicts to be the case, although in some cases The estimates generated by the adaptive forecasting model are based on the assumption that the estimates are for the near future. One way to achieve this is to use current biomarkers to estimate what the In contrast to metric estimation, we take a set of real-time data as a given (for a future point in time). By adaptive predictive models that are adjusted to generate future biometric estimates (for Another way to achieve this is to use the method outlined above, but After that, the trends of past and present biometric estimation are further processed to predict future biometric estimation. By applying additional layers to the model, predictions of the current biomedical The method may generate a trick estimate. Other techniques may also be used.

[0109] A key advantage of predicting future biometric estimates is that The present invention is also applicable to subjects wearing an embodiment of the present invention who are at increased risk of future undesirable bias (as opposed to bias estimation). These desirability factors can be taken into account so that preventative measures can be taken to prevent o-metric This allows the subject to be informed that a negative result may be imminent. For example, subjects with controlled diabetes may experience sudden increases and decreases in their blood glucose levels. Knowing that you have a high risk of developing this condition, you can take preventative doses of insulin and glucose to prevent this negative outcome. Similarly, subjects with controlled hypertension may benefit from Know that your blood pressure can rise and fall suddenly and treat it accordingly with medicine. Benefits can be gained by enabling

[0110] Example embodiments are described herein with reference to block diagrams and flow diagrams. The blocks of these block diagrams and flow diagrams, and The combination of blocks in the diagram represents electrical components having analog and / or digital elements. to computer program instructions that are executed by one or more computer circuits, such as It will be appreciated that these computer program instructions may be implemented in a general purpose manner. computer circuits, special purpose computer circuits, and / or other programmable data processing The circuitry may be provided with a processor circuit for processing computer and / or other programmable data. The instructions executed by the processor of the processing unit are stored in transistors, memory locations, and Transforming and controlling the values ​​stored in the memory and other hardware components within such circuitry; Implement the functions / operations specified in the block diagrams and flow diagrams, thereby Means (functions) and / or structures for performing the functions / operations specified in the diagrams and flow charts It is possible to create a machine that produces

[0111] A function that instructs a computer or other programmable data processing device to function in a particular way. These computer program instructions, which may be represented by the following symbols, may be stored on a computer readable medium. The instructions stored in the memory implement the functions / acts specified in the block diagrams and / or flow diagrams. The method may also be stored on a tangible computer readable medium to generate an article of manufacture that includes instructions for Cut.

[0112] A tangible, non-transitory computer readable medium may be electronic, magnetic, optical, electromagnetic, or semiconductor. The computer-readable medium may include any data storage system, apparatus, or device. More specific examples of the object include the following: portable computer diskettes , random access memory (RAM) circuits, read-only memory (ROM: read-only memory) circuit, erasable programmable read-only memory (EP ROM or flash memory) circuits, portable compact disk read-only memory (CD-ROM: compact disc read-only memory), and portable digital video discs There is a disk read-only memory (DVD / BlueRay).

[0113] Instructions executed on a computer or other programmable device may be illustrated in block diagrams and / or diagrams. The flowchart illustrates how the computer is programmed to provide steps for performing the functions / operations specified in the flowchart. The computer program instructions are then executed on a computer and / or other programmable data processing device. and load a sequence of operational steps onto a computer and / or other programmable device. The present invention can also be implemented by performing the above steps. The embodiments of the present invention may be referred to as "logic," "circuitry," "modules," "engines," or the like. Hardware and / or digital signal processors, etc., which may collectively be referred to as variations of the above. The software (firmware, resident software, microcontroller, etc.) that runs on the processor of This can be embodied in the form of a code, etc.

[0114] The functionality of a given block in the block and flow diagrams can be separated into multiple blocks. and / or at least partially omit the functions of two or more blocks in the block diagrams and flow diagrams. It should also be noted that the blocks shown in the figure can be integrated in a similar manner. You can add / insert other blocks into the diagram. Moreover, some of the diagrams are Although the communication paths include arrows showing the primary direction, communication may occur in the opposite direction to that shown. It should be understood that it is possible.

[0115] The foregoing is illustrative of the present invention and is not to be construed as limiting. Although exemplary embodiments have been described, it is to be understood that certain modifications and variations may be made without materially departing from the teachings and advantages of the present invention. Those skilled in the art will readily appreciate that many variations are possible in the exemplary embodiments. Accordingly, all such modifications are intended to be included within the scope of the present invention as defined in the following claims. The present invention is defined by the following claims, including their equivalents. It can be seen.

Claims

1. 1. A method for generating an estimated blood glucose level for a subject, the method comprising: The steps to be taken are: A step of receiving real-time PPG data from a PPG sensor attached to the subject. Top and The real-time PPG data is used to estimate the subject's estimated blood glucose level using an adaptive prediction model. generating a value; The method includes:

2. receiving a blood glucose level measured by a blood glucose monitoring device; When the measured blood glucose level is accepted, the blood glucose level estimation accuracy of the adaptive prediction model is improved. updating one or more parameters of the adaptive predictive model in real time to and The method of claim 1 further comprising:

3. The measured blood glucose level is a real-time measurement, and the blood glucose monitoring device The method of claim 2 , which is worn by a subject.

4. A step of receiving activity information of the subject from a motion sensor attached to the subject. Top and receiving a measured blood glucose level from a blood glucose monitoring device in response to the activity information of the subject; and When the measured blood glucose level is accepted, the blood glucose level estimation accuracy of the adaptive prediction model is improved. updating one or more parameters of the adaptive predictive model in real time to and The method of claim 1 further comprising:

5. A step of detecting whether the generated estimated blood glucose level is above or below a threshold. Top and The generated estimated blood glucose level is not determined to be above or below the threshold. accepting a blood glucose level measured by the blood glucose monitoring device; One or more of the adaptive prediction models are adjusted to improve the blood glucose level estimation accuracy of the adaptive prediction model. Steps to update the above parameters in real time The method of claim 1 further comprising:

6. The generated estimated blood glucose level is alerted when the estimated blood glucose level is above or below a threshold value. The method of claim 1 , further comprising the step of sending the information to a host device.

7. The at least one processor is configured to include a wearable device attached to the subject. The method of claim 1 , wherein

8. The wearable device is attached to the subject's ear, the subject's limb, or the subject's 8. The method of claim 7, wherein the subject is provided with a patch attached to their finger or worn on their finger.

9. The method of claim 7 , wherein the wearable device comprises the PPG sensor.

10. The at least one processor is configured to include a wearable device attached to the subject. the wearable device includes the PPG sensor and the blood glucose monitoring device. The method of claim 2 comprising a device.

11. The method of claim 1 , wherein the PPG sensor comprises an imaging sensor.

12. The adaptive predictive model is one of a regression model, a machine learning model, and a classifier model.

2. The method of claim 1, comprising:

13. The method of claim 1 , wherein generating the estimated blood glucose value comprises generating a current estimated blood glucose value.

14. The method of claim 1 , wherein generating the estimated blood glucose value comprises generating a future estimated blood glucose value.

15. A process for predicting a future estimated blood glucose level by processing the current estimated blood glucose level and the past estimated blood glucose levels. The method of claim 13 further comprising the step of:

16. A wearable device, A PPG sensor; The estimated blood glucose level of the subject wearing the wearable device is measured from the PPG sensor. At least one predictive model is generated using real-time PPG data. Processor and A wearable device comprising:

17. The at least one processor receiving a blood glucose measurement from a blood glucose monitoring device; When the measured blood glucose level is accepted, the blood glucose level estimation accuracy of the adaptive prediction model is improved. updating one or more parameters of the adaptive predictive model in real time to and The wearable device of claim 16 , further comprising:

18. The at least one processor receiving a blood glucose measurement from a blood glucose monitoring device; A determination is made that the generated estimated blood glucose level is above or below a threshold. and one of the adaptive prediction models so as to improve the blood glucose level estimation accuracy of the adaptive prediction model. The above parameters are updated in real time. The wearable device of claim 16 , further comprising:

19. The at least one processor detects whether the generated estimated blood glucose value is above a threshold.

2. The method of claim 1, further comprising the step of: sending an alert to a remote device that the threshold is exceeded or below the threshold.

7. A wearable device as described in claim 6.

20. The wearable device may be attached to the subject's ear, the subject's limb, or the subject's 17. The method of claim 16, wherein the device is attached as a patch or worn on the subject's finger. Wearable device.

21. The wearable device of claim 16 , wherein the PPG sensor comprises an imaging sensor.

22. The adaptive predictive model is one of a regression model, a machine learning model, and a classifier model. The wearable device of claim 16, comprising:

23. A method for improving the accuracy of blood glucose estimation in an adaptive predictive model, comprising: The steps performed by the processor are: a) Real-time PPG data from a PPG sensor attached to the subject and blood glucose monitor receiving a measured blood glucose level from the monitoring device within a reception period; b) generating features from the received PPG data; c) storing the feature amount and the measured blood glucose level; d) processing the stored features together with the stored measured blood glucose levels to obtain the updating one or more parameters of the adaptive predictive model in real time. the updated one or more parameters improve the accuracy of the adaptive predictive model's blood glucose level estimation. Steps and The method includes:

24. The method further comprises repeating steps a)-d) for one or more subsequent time periods. The method according to claim 23.

25. generating an estimated blood glucose level for the subject using the adaptive prediction model; Determining whether the generated estimated blood glucose level is above or below a threshold value. Steps and The generated estimated blood glucose level is not determined to be above or below the threshold. accepting another blood glucose measurement by the blood glucose monitoring device. 、 updating the one or more parameters of the adaptive predictive model in real time. and 24. The method of claim 23 further comprising:

26. The step of generating a feature from the received PPG data includes: generating a feature quantity at a feature quantity generation interval within the reception period by using an interval window; 24. The method of claim 23, comprising:

27. Updating the one or more parameters of the adaptive predictive model further comprises: processing the stored measured blood glucose values ​​and previously stored measured blood glucose values; Generate an interpolation between the stored measured blood glucose value and the previously stored measured blood glucose value. Steps and 24. The method of claim 23, comprising:

28. Processing the stored measured blood glucose values ​​and the previously stored measured blood glucose values ​​includes 28. The method of claim 27, further comprising the step of processing a plurality of previously stored blood glucose measurement values. How to.

29. Processing the stored measured blood glucose values ​​and the previously stored measured blood glucose values ​​includes 30. The method of claim 28, further comprising generating an interpolation of the expected measured blood glucose value. 。

30. The method of claim 23 , wherein the PPG sensor comprises an imaging sensor.

31. The adaptive predictive model is one of a regression model, a machine learning model, and a classifier model.

24. The method of claim 23, comprising:

32. The method of claim 23, wherein the characteristic quantities and the measured blood glucose levels are stored in a data buffer. Law.

33. 33. The method of claim 32, wherein the data buffer comprises a FIFO (First In, First Out) buffer. How to.

34. Processing the stored features together with the stored measured blood glucose levels includes:

24. The method of claim 23, further comprising processing at least one function of the determined features. method.

35. Processing the stored features together with the stored measured blood glucose levels includes: and calculating statistics about the time series of at least one of the features obtained by the method according to the present invention.

24. The method according to item 23.

36. Processing the stored features together with the stored measured blood glucose levels includes: and computing statistics for the plurality of time series of at least one of the features determined. The method of claim 23.

37. Processing the stored features together with the stored measured blood glucose levels includes: Calculating weighted statistics for the multiple time series of at least one of the features selected.

24. The method of claim 23, comprising:

38. A system for improving blood glucose estimation accuracy of an adaptive predictive model, the system comprising: The system comprises at least one processor, The at least one processor Real-time PPG data from a PPG sensor attached to the subject and blood glucose monitoring receiving a measured blood glucose level from the device during a reception period; generating features from the received PPG data; storing the feature amount and the measured blood glucose level; The stored feature amount is processed together with the stored measured blood glucose level to obtain the suitable blood glucose level. updating one or more parameters of the adaptive predictive model in real time, The updated one or more parameters improve the accuracy of the blood glucose estimation of the adaptive predictive model. Steps and To carry out system.

39. The at least one processor further comprises: generating an estimated blood glucose level using the adaptive prediction model; Determining whether the generated estimated blood glucose level is above or below a threshold value. Steps and The generated estimated blood glucose level is not determined to be above or below the threshold. and updating the one or more parameters of the adaptive predictive model in real time. Tep and 40. The system of claim 38, further comprising:

40. The at least one processor further comprises: determining whether the generated estimated blood glucose value is above the threshold value; sending an alert to a remote device that the billing level is above or below the threshold; 40. The system described in item 39.

41. The adaptive predictive model is one of a regression model, a machine learning model, and a classifier model.

40. The system of claim 38, comprising:

Images