Method, apparatus and computer program product for analyzing a pulse wave signal

Abstract

According to an aspect, there is provided a computer-implemented method for analyzing a pulse wave signal PWS obtained from a subject. The PWS comprises pulse wave measurements of a plurality of cardiac cycles of the subject during a first time period. The method comprises (i) analyzing (40) the PWS to identify a plurality of cardiac cycles and a respective reference point for each identified cardiac cycle; (ii) determining 2PWS as a second derivative of the PWS with respect to time; (iii) determining a normalized 2PWS by normalizing, for each portion of 2PWS corresponding to a respective identified cardiac cycle, the portion of 2PWS relative to an amplitude of 2PWS at the identified reference point of the cardiac cycle; (iv) for a first lag time value, determining (42) an n-th order polynomial fit for a first set of values of the normalized 2PWS, wherein the first set of values of the normalized 2PWS comprises values of the normalized 2PWS occurring at a first lag time value from the reference point of each identified cardiac cycle, wherein n is equal to or greater than 1; (v) performing (44) one or more further iterations of step (iv) for one or more further lag time values to determine respective further n-th order polynomial fits for respective sets of values of the normalized 2PWS, wherein the respective sets of values of the normalized 2PWS comprise values of the normalized 2PWS occurring at the respective further lag time values from the reference point of each identified cardiac cycle; and (vi) forming (46) a first average cardiac cycle waveform for a first time point in the first time period, wherein the first average cardiac cycle waveform is formed from values of the plurality of n-th order polynomial fits at the first time point.

Description

Technical Field

[0001] This disclosure relates to pulse wave signals obtained from an object, the pulse wave signals comprising pulse wave measurements over multiple cardiac cycles of the object. More specifically, this disclosure relates to methods, apparatus, and computer program products for analyzing pulse wave signals. Background Technology

[0002] Blood pressure (BP) is an important indicator of a person's health. In the United States, it is estimated that approximately 30% of adults have hypertension. Hypertension is a common health problem without obvious external symptoms. Blood pressure typically increases with age, and the risk of developing hypertension later in life is considerable. Persistent hypertension is one of the key risk factors for stroke, heart failure, and increased mortality. A person's condition can be improved through lifestyle modifications, choosing a healthy diet, and medication. Especially for high-risk patients, continuous 24-hour blood pressure monitoring via a system that does not interfere with normal daily activities is crucial. Continuous blood pressure monitoring can also be used for patients in healthcare settings such as hospitals (e.g., operating rooms (ORs) or intensive care units (ICUs)). Low blood pressure can lead to poor oxygenation of vital organs and may cause organ damage. Excessively high blood pressure can cause bleeding, especially during and after surgical procedures, particularly in neurosurgery, and bleeding should be prevented.

[0003] In some cases, an absolute measurement of blood pressure can be obtained, while in others, a relative measurement, such as a measurement of blood pressure variability, can be obtained. In particular, blood pressure can vary within short time windows, such as approximately a few minutes, and these variations may be relevant to further medical examinations and potential medical interventions.

[0004] Many different techniques are available for measuring blood pressure and / or changes in blood pressure. Some of these techniques measure blood pressure itself, while others measure other physiological characteristics of the subject and use these characteristics as substitutes for blood pressure, for example, by correlating changes or values ​​of physiological characteristics with changes or values ​​of blood pressure. Some techniques for directly measuring blood pressure require invasive access to the subject's arteries or the use of bulky / inconvenient devices, such as inflatable cuffs. However, simple and / or inconspicuous sensors applied to the subject's body can be used to measure some physiological characteristics that serve as substitutes for blood pressure.

[0005] Tonometers use externally placed force or pressure sensors to measure arterial dilation (i.e., a waveform representing arterial dilation) when pressure is applied to an artery. Alternatively, one or more photoplethysmography (PPG) sensors can be placed on a part of the body to obtain one or more PPG signals representing changes in the volume of blood flow in the body part during multiple cardiac cycles (cardiac cycles). Both techniques obtain pulse wave signals (PWS) covering multiple cardiac cycles of the subject. This pulse waveform / signal can be analyzed to determine one or more physiological characteristics that can be used as an alternative measurement of blood pressure.

[0006] One physiological property that can be used as an alternative to blood pressure measurement is pulse wave velocity (PWV). When the heart beats, blood flowing through the aorta and another arterial system generates a pulse wave. The speed of this pulse wave (called pulse wave velocity) is influenced by blood (fluid) properties and some arterial properties (such as diameter and compliance). These blood and arterial properties are also affected by blood pressure; therefore, changes in PWV may be related to changes in blood pressure.

[0007] Some techniques for measuring pulse wave velocity (PWV) use a two-point or dual-point approach. This requires two sensors (e.g., a PPG sensor) to capture two signals simultaneously. The signal from the first sensor is used to detect the start of the pulse wave at a proximal location (e.g., near the heart). The signal from the second sensor is used to detect the arrival of the pulse wave at a distal location, such as the femoral artery in the patient's finger.

[0008] However, to minimize the inconvenience of measuring instruments on the subject, single-point techniques for measuring PWV are being developed. These techniques utilize pulse wave reflections in the arterial tree. There exists a direct pulse wave traveling from the aorta to, for example, the finger, and there exists an indirect pulse wave traveling first from the aorta to the renal artery branch and then from the renal artery branch to the finger. In this way, the indirect (reflected) pulse wave arrives at the finger position later than the direct pulse wave. When the arrival times of the direct and indirect pulse waves are measured at the finger, the subtraction of these arrival times yields the time required for the pulse to travel from the aortic arch to the renal artery branch and back. Using knowledge (or an approximation) of this additional travel distance of the reflected pulse wave, the pulse wave velocity of the reflected wave can be estimated according to the following equation (1):

[0009]

[0010] Where L hr This is the distance between the aortic arch and the branches of the renal arteries, while PRT is the so-called pulse reflex time, defined as the time between the start of the direct pulse wave (upper side) and the start of the indirect pulse wave (upper side). The time t required to calculate PRT is... c and t aIt can be determined by the so-called "acceleration waveform" measured by PPG, that is, for example, by the double derivative of the PPG waveform with respect to time.

[0011] Typically, it is assumed that the acceleration waveform consists of five "reference points" or "base points", such as Figure 1 As shown. Figure 1 (a) shows an exemplary PPG signal covering a 1-second time period, in which the dicrotic notch is indicated. Figure 1 (b) shows Figure 1 (a) The first derivative of the PPG signal with respect to time. Figure 1 (c) shows Figure 1 (a) The second derivative of the PPG signal with respect to time. The first derivative is denoted as v-PPG, and the second derivative (acceleration waveform) is denoted as a-PPG. Figure 1 Five reference points are shown in the acceleration waveform in (c). Point a marks the beginning of the direct pulse wave, and point b marks the end of the direct pulse wave. Point e marks the end of the systolic phase (aortic valve closure), and for the purposes of this disclosure, it is assumed that point c marks the beginning of the reflected wave, and point d marks the end of the reflected wave.

[0012] Figure 2 An example of single-point PPG measurement and PWV derived according to equation (1) is shown. Figure 2 The graph in (a) shows the mean arterial pressure (MAP) of ICU patients over 5 hours, in mmHg. Figure 2 The graph in (b) shows the PWV calculated from the PPG signal according to equation (1), in units of m / s, where 2L hr The estimated distance (twice the distance from the heart to the renal artery branch) is 75 cm.

[0013] exist Figure 2 As can be seen, there is a strong positive correlation between MAP and pulse wave velocity measured at a single point. However, the robustness of pulse wave velocity measurement may be questionable because there are many outliers in the pulse wave velocity values, especially when compared with pulse wave velocities derived from two-point measurement methods. Summary of the Invention

[0014] One problem with single-point pulse wave velocity measurement is that the reference point for the reflected pulse wave is not always easy to detect. This can be seen from... Figure 1As seen in the a-PPG plot in (c), calculating the second derivative of the original PPG signal results in very poor signal quality due to quantization and noise. Signal smoothing (time-based or over multiple cardiac cycles) can be applied to obtain improved reference point detection. After such smoothing, reference points can be robustly detected. However, to detect points c and d well (which are related to the reflected wave), considerable smoothing is required before robust reference point detection is possible. Excessive time smoothing leads to the loss of high-frequency reference points, while excessive averaging over multiple cardiac cycles can cause problems under time-varying conditions.

[0015] Therefore, there is a need for improvement in the averaging of cardiac cycle waveforms.

[0016] The technique described herein applies averaging over multiple cardiac cycles to reduce or remove noise in the resulting average, thereby enabling improved analysis of reflected pulse waves and / or other features of the cardiac cycle waveform. As mentioned above, analysis of reflected pulse waves can be used as an alternative to blood pressure measurement; however, it should be understood that analysis of the average cardiac cycle waveform can be used to monitor other aspects of a subject's health, such as trends in arterial compliance throughout hospitalization.

[0017] According to a first aspect, a computer-implemented method is provided for analyzing pulse wave signals (PWS) obtained from an object. The PWS includes pulse wave measurements of multiple cardiac cycles of the object during a first time period. The method includes (i) analyzing the PWS to identify multiple cardiac cycles and a corresponding reference point for each identified cardiac cycle; (ii) determining a 2PWS as the second derivative of the PWS with respect to time; (iii) determining a normalized 2PWS by normalizing the portion of the 2PWS corresponding to each identified cardiac cycle with respect to the amplitude of the 2PWS at the identified reference point of the cardiac cycle; and (iv) for a first lag time value, determining an n-order polynomial fit for a first set of values ​​of the normalized 2PWS, wherein the first set of values ​​of the normalized 2PWS includes a first lag time distance from the reference point of each identified cardiac cycle. The normalized 2PWS values ​​appearing at n, where n is equal to or greater than 1; (v) performing one or more additional iterations of step (iv) for one or more additional lag time values ​​to determine corresponding additional n-order polynomial fits for corresponding group values ​​of the normalized 2PWS, wherein the corresponding group values ​​of the normalized 2PWS include the normalized 2PWS values ​​appearing at corresponding additional lag time values ​​relative to a reference point for each identified cardiac cycle; and (vi) forming a first average cardiac cycle waveform at a first time point in the first time period, wherein the first average cardiac cycle waveform is formed by the values ​​of the plurality of n-order polynomial fits at the first time point. Therefore, this aspect provides an improved average cardiac cycle waveform that takes into account the trend of the PWS within the first time period and allows for improved analysis of the PWS obtained using, for example, single-point measurement techniques.

[0018] In some embodiments, the method further includes forming a second average cardiac cycle waveform at a second time point in the first time period, wherein the second average cardiac cycle waveform is formed by the values ​​fitted by the plurality of nth-order polynomials at the second time point.

[0019] In these embodiments, the method may further include comparing the first average cardiac cycle waveform and the second average cardiac cycle waveform to determine the change in the average cardiac cycle waveform between the first time point and the second time point. These embodiments provide a means to assess the change in the average cardiac cycle waveform over a first time period, for example, to assess how cardiac cycle-related characteristics change.

[0020] In these embodiments, the method may further include determining a measure of the subject's blood pressure or a measure of the subject's blood pressure variation based on the first mean cardiac cycle waveform and the second mean cardiac cycle waveform.

[0021] In some embodiments, the method further includes processing the first mean cardiac cycle waveform to determine a measure of the subject's blood pressure.

[0022] In some embodiments, each of the first lag time value and the one or more other lag time values ​​is equal to or less than the duration of the object's cardiac cycle.

[0023] In some embodiments, the reference point for each identified cardiac cycle is the start of the object's pulse wave. This reference point is useful because it is relatively easy to detect the first or second derivative of the PWS with respect to time.

[0024] In some embodiments, step (ii) is performed before or as part of step (i), and step (i) may include identifying a plurality of cardiac cycles and a corresponding reference point for each identified cardiac cycle as a local maximum in 2PWS.

[0025] In an alternative embodiment, step (ii) is performed before or as part of step (i), and step (i) may include detecting a peak in the first derivative of the PWS with respect to time (1PWS); and identifying multiple cardiac cycles and a corresponding reference point for each identified cardiac cycle as a local maximum in the 2PWS within a corresponding search window defined by the detected peak in the 1PWS.

[0026] In some embodiments, n is 1. In other embodiments, n is 2.

[0027] In some embodiments, the PWS is a photoplethysmography (PPG) signal.

[0028] According to a second aspect, an apparatus is provided for analyzing pulse wave signals (PWS) obtained from an object. The PWS includes pulse wave measurements of multiple cardiac cycles of the object during a first time period. The apparatus is configured to (i) analyze the PWS to identify multiple cardiac cycles and a corresponding reference point for each identified cardiac cycle; (ii) determine a 2PWS as the second derivative of the PWS with respect to time; (iii) determine a normalized 2PWS by normalizing the portion of the 2PWS corresponding to each identified cardiac cycle relative to the amplitude of the 2PWS at the identified reference point of the cardiac cycle; and (iv) for a first lag time value, determine an n-order polynomial fit for a first set of values ​​of the normalized 2PWS, wherein the first set of values ​​of the normalized 2PWS includes values ​​at a first lag time relative to the reference point of each identified cardiac cycle. (v) The normalized 2PWS values ​​appearing at time values, where n is equal to or greater than 1; (v) One or more additional iterations of operation (iv) are performed for one or more additional lag time values ​​to determine corresponding additional n-order polynomial fits for corresponding group values ​​of the normalized 2PWS, wherein the corresponding group values ​​of the normalized 2PWS include the normalized 2PWS values ​​appearing at corresponding additional lag time values ​​at a reference point for each identified cardiac cycle; and (vi) A first average cardiac cycle waveform is formed at a first time point in the first time period, wherein the first average cardiac cycle waveform is formed by the values ​​of the plurality of n-order polynomial fits at the first time point. Therefore, this aspect provides an improved average cardiac cycle waveform that takes into account the trend of PWS within the first time period and allows for improved analysis of PWS obtained using, for example, single-point measurement techniques.

[0029] In some embodiments, the apparatus is further configured to form a second average cardiac cycle waveform at a second time point in the first time period, wherein the second average cardiac cycle waveform is formed by the values ​​fitted by the plurality of n-order polynomials at the second time point.

[0030] In these embodiments, the apparatus may also be configured to compare the first average cardiac cycle waveform and the second average cardiac cycle waveform to determine the change in the average cardiac cycle waveform between the first time point and the second time point. These embodiments provide a means to assess the change in the average cardiac cycle waveform over a first time period, for example, to assess how characteristics related to the cardiac cycle change.

[0031] In these embodiments, the device may also be configured to determine a measure of the subject's blood pressure or a measure of the subject's blood pressure variation based on the first mean cardiac cycle waveform and the second mean cardiac cycle waveform.

[0032] In some embodiments, the apparatus may also be configured to process the first average cardiac cycle waveform to determine a measure of the subject's blood pressure.

[0033] In some embodiments, each of the first lag time value and the one or more other lag time values ​​is equal to or less than the duration of the object's cardiac cycle.

[0034] In some embodiments, the reference point for each identified cardiac cycle is the start of the object's pulse wave. This reference point is useful because it is relatively easy to detect the first or second derivative of the PWS with respect to time.

[0035] In some embodiments, operation (ii) is performed before or as part of operation (i), and operation (i) may include identifying a plurality of cardiac cycles and a corresponding reference point for each identified cardiac cycle as a local maximum in 2PWS.

[0036] In an alternative embodiment, operation (ii) is performed before or as part of operation (i), and operation (i) may include detecting a peak in the first derivative of the PWS with respect to time (1PWS); and identifying a plurality of cardiac cycles and a corresponding reference point for each identified cardiac cycle as a local maximum in the 2PWS within a corresponding search window defined by the detected peak in the 1PWS.

[0037] In some embodiments, n is 1. In other embodiments, n is 2.

[0038] In some embodiments, the PWS is a photoplethysmography (PPG) signal.

[0039] In some embodiments, the apparatus further includes a pulse wave sensor for obtaining the PWS from the object. In an alternative embodiment, the apparatus is configured to receive the PWS from the pulse wave sensor.

[0040] According to a third aspect, a computer program product including a computer-readable medium is provided, the computer-readable medium having computer-readable code configured to cause, when executed by a suitable computer or processor, to cause the computer or processor to perform the method according to the first aspect or any embodiment thereof.

[0041] These and other aspects will become apparent with reference to one or more embodiments described below and will be explained with reference to one or more embodiments described below. Attached Figure Description

[0042] Exemplary embodiments will now be described by way of example only, with reference to the following figures, wherein:

[0043] Figure 1 (a) shows the PPG signal in a 1-second segment, while Figure 1 (b) and (c) show the first and second derivatives of the PPG signal, respectively;

[0044] Figure 2 (a) shows the measurement of mean arterial pressure over a 5-hour time period, while Figure 2 (b) shows an example of single-point PPG measurement and PWV derivation according to Equation (1) within the same time period;

[0045] Figure 3 This is a block diagram of an apparatus according to various exemplary embodiments;

[0046] Figure 4 This is a flowchart illustrating a method for analyzing pulse wave signals obtained from an object according to various embodiments;

[0047] Figure 5 The PPG signal in a 1-second segment is shown, while Figure 5 (b) and 5(c) show the first and second derivatives of the PPG signal, respectively;

[0048] Figure 6 (a) shows a graph of arterial blood pressure within a 1-minute time window, while Figure 6 (b) shows a graph of the second derivative of the pulse wave signal during a 60-second time window;

[0049] Figure 7 It is a graph showing multiple superimposed waveforms identified in the second derivative of the pulse wave signal;

[0050] Figure 8 It is shown Figure 7 A graph of the 0th-order average values ​​of the multiple waveforms shown;

[0051] Figure 9 It is a graph showing the values ​​of multiple waveforms identified in the second derivative of the pulse wave signal at a specific lag time value from the common reference point, and three nth-order polynomial fits for the displayed values.

[0052] Figure 10 It is a graph showing the first-order average waveform at several time points during a 60-second time window; and

[0053] Figure 11 This is a functional block diagram illustrating various operations for analyzing pulse wave signals according to various embodiments. Detailed Implementation

[0054] The technique described herein applies averaging over multiple cardiac cycles to reduce or remove noise in the resulting average, thereby improving the analysis of reflected pulse waves and / or other features of the cardiac cycle waveform. Analysis of reflected pulse waves can be used as an alternative to blood pressure measurement; however, it should be understood that analysis of the averaged cardiac cycle waveform can provide additional information about the subject's health. The described technique is particularly useful for so-called single-point measurement techniques that apply a single sensor to a subject.

[0055] For a pulse wave signal (PWS) that includes information about pulse changes / pulse waves at measurement points on the subject's body, a reference point is identified in the PWS for each cardiac cycle to be smoothed. The PWS can be, for example, a PPG signal or a pulse wave signal obtained using intraocular pressure measurement. The reference point to be identified preferably relates to the initial up-flank of the pulse wave, which should be unaffected by reflections and therefore should be a stable reference point (i.e., independent of blood pressure changes). However, it should be understood that different reference points can be used if desired. Next, the average cardiac cycle waveform is calculated using the various times and amplitudes of occurrence of the reference points. Normal averaging of all cardiac cycle waveforms across a time window does not allow for time-varying cases. Such normal averaging is described in WO2015 / 044010. The averaging technique described herein extends the averaging to allow for (linear) variation in the time of each lag in the averaging process (where lag or lag time is the time relative to the identified reference point for each cardiac cycle, e.g., the time relative to an identified reference point a for each cardiac cycle).

[0056] Figure 3 This is a block diagram of an apparatus 30 for analyzing PWS according to various embodiments of the technology described herein. Figure 3 The diagram illustrates a pulse wave sensor 32 for measuring pressure waves at a single point on a subject's body and outputting a pulse wave sensor (PWS). The pulse wave sensor 32 can be a PPG sensor, a sensor based on intraocular pressure measurement, or a hydraulic sensor pad that can apply pressure to the subject's (upper arm) and use any type of pressure sensor (e.g., the MPXV6115 series integrated silicon pressure sensor from NXP Semiconductors). In some embodiments, the pulse wave sensor 32 can be part of or integrated with device 30. In other embodiments, device 30 can be connected directly (e.g., wired) or indirectly (e.g., using wireless communication technologies such as Bluetooth, WiFi, cellular communication protocols, etc.). In alternative embodiments, device 30 may not be connected to the pulse wave sensor 32; instead, device 30 may obtain the PWS from another device or apparatus (e.g., a server or database).

[0057] As is well known, a PPG sensor 32 can be placed on the body of an object (e.g., on an arm, leg, earlobe, finger, etc.) and can provide an output signal (“PPG signal”) related to the volume of blood passing through that part of the body. The volume of blood passing through that part of the body is related to the pressure of the blood in that part of the body. The PPG sensor 32 typically includes a light sensor and one or more light sources. The PPG signal output by the PPG sensor 32 can be a raw measurement signal from the light sensor (e.g., the PPG signal can be a signal representing the light intensity over time). Alternatively, the PPG sensor 32 can perform some preprocessing of the light intensity signal, such as to reduce noise and / or compensate for motion artifacts; however, it should be understood that such preprocessing is not necessary for implementing the techniques described herein.

[0058] Device 30 may be in the form of a computing device (e.g., a server, desktop computer, laptop computer, tablet computer, smartphone, smartwatch, etc.) or a type of device typically found in clinical settings (e.g., a patient monitoring device for monitoring (and optionally displaying) various physiological characteristics of a subject / patient (e.g., a bedside monitoring device for a patient in a clinical setting)).

[0059] The device 30 includes a processing unit 34 that controls the operation of the device 30 and can be configured to run or perform the methods described herein to analyze the PWS. The processing unit 34 can be implemented in a variety of ways, using software and / or hardware, to perform the various functions described herein. The processing unit 34 may include one or more microprocessors or digital signal processors (DSPs) that can be programmed using software or computer program code to perform desired functions and / or components that control the processing unit 34 to perform desired functions. The processing unit 34 may be implemented as a combination of dedicated hardware performing some functions (e.g., amplifiers, preamplifiers, analog-to-digital converters (ADCs) / or digital-to-analog converters (DACs)) and processors performing other functions (e.g., one or more programmed microprocessors, controllers, DSPs, and associated circuitry). Examples of components that may be employed in various embodiments of this disclosure include, but are not limited to, conventional microprocessors, DSPs, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), hardware for implementing neural networks, and / or so-called artificial intelligence (AI) hardware accelerators (i.e., one or more processors or other hardware specifically designed for AI applications that can be used in conjunction with a main processor).

[0060] Processing unit 34 is connected to storage unit 36, which may store data, information, and / or signals for use by processing unit 34 in controlling the operation of device 30 and / or in running or performing the methods described herein. In some embodiments, storage unit 36 ​​stores computer-readable code that can be executed by processing unit 34 to cause processing unit 34 to perform one or more functions (including the methods described herein). In particular embodiments, the program code may be in the form of an application for a smartwatch, smartphone, tablet, laptop, or computer. Storage unit 36 ​​may include any type of non-transient machine-readable medium, such as cache or system memory, including volatile and non-volatile computer memory, such as random access memory (RAM), static RAM (SRAM), dynamic RAM (DRAM), read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), and electrically erasable PROM (EEPROM), and storage unit 36 ​​may be implemented in the form of storage chips, optical discs (such as optical discs (CD), digital versatile discs (DVD), or Blu-ray discs), hard disks, magnetic tape storage solutions, or solid-state devices (including memory sticks, solid-state drives (SSDs), memory cards, etc.).

[0061] In some embodiments, device 30 includes a user interface 38, which includes one or more components that enable a user of device 30 to input information, data, and / or commands into device 30 and / or enable device 30 to output information or data to the user of device 30. Information that can be output by user interface 38 may include an indication or diagram of the mean cardiac cycle waveform at one or more time points, and / or information derived from the mean cardiac cycle waveform. User interface 38 may include one or more suitable input components, including but not limited to a keyboard, keypad, one or more buttons, switches or dial pads, mouse, trackpad, touchscreen, stylus, camera, microphone, etc., and / or user interface 38 may include one or more suitable output components, including but not limited to a display screen, one or more lamps or lamp elements, one or more speakers, vibration elements, etc.

[0062] It should be understood that actual implementations of device 30 may include Figure 3 Additional components to those components shown. For example, device 30 may also include a power source (e.g., a battery) or components enabling device 30 to connect to a mains power source. Device 30 may also include interface circuitry for enabling data connectivity and / or data exchange with other devices, including pulse wave sensor 32 (in embodiments where pulse wave sensor 32 is separate from device 30), servers, databases, user equipment, and / or other sensors.

[0063] Figure 4 The flowchart illustrates exemplary methods for analyzing PWS obtained from an object, according to various embodiments. In some embodiments, the processing unit 34 in apparatus 30 may be configured to implement... Figure 4 The method described above. In other embodiments, computer-readable code may be provided that, when executed by the computer or processing unit 34, causes the computer or processing unit 34 to perform... Figure 4 The method.

[0064] The pulse wave sensor 32, located at a single measurement point on the object, receives the pulse wave measurement results from multiple cardiac cycles (i.e., heartbeats) of the object. The sampling rate F... s Obtain PWS. Figure 4 The method described below refers to PWS in the form of PPG signals; however, it should be understood that this method can be applied to other forms of PWS. In some embodiments, Figure 4 The methods described herein can be executed periodically or continuously on the PWS when it is received or measured from the object. In some embodiments, Figure 4 The method can operate on windowed portions of longer PWS (e.g., a 1-minute window of a PWS covering a 1-hour time period). However, it should be understood that the method can be applied to PWS of any desired length covering any number of cardiac cycles. In the following description, references to operations or steps performed on a PWS refer to performing those operations or steps on a portion of the PWS of interest (e.g., a portion corresponding to a 1-minute time period).

[0065] In the first step (step 40) of this method, the PWS is analyzed to identify cardiac cycles, and a corresponding reference point is identified for each cardiac cycle. The reference point is a point in each cardiac cycle that can be used in subsequent steps to "align" the cardiac cycles and enable the determination of average values.

[0066] As referenced above Figure 1Each cardiac cycle in the described “acceleration waveform” (the second derivative of the PWS with respect to time) can be considered to include five “reference points” or “reference points”. Reference point a marks the beginning of the direct pulse wave, which is the start of the pulse wave (i.e., the pulsation of blood caused by the heartbeat), while point b marks the end of the direct pulse wave. Point e marks the end of systole (aortic valve closure), and for the purposes of this disclosure, it is assumed that point c marks the beginning of the reflected wave, and point d marks the end of the reflected wave. Preferably, in step 40, the reference point for each cardiac cycle is the beginning of the pulse wave (the direct portion of the pulse wave), which can be considered as the beginning of systole, i.e., reference point a. However, it should be understood that in other embodiments, other reference points among the reference points can be identified in step 40, or in fact, related to… Figure 1 (c) shows different reference points a to e.

[0067] In the following text, the signal corresponding to the first derivative of PWS with respect to time is denoted as "1PWS", and the signal corresponding to the second derivative of PWS with respect to time is denoted as "2PWS". When described with reference to specific examples of PPG signals, 1PWS is also referred to as the "velocity waveform" (v-PPG), and 2PWS is also referred to as the "acceleration waveform" (a-PPG).

[0068] Some embodiments of step 40 provide detection of the onset of the pulse wave by detecting local maxima in each cardiac cycle represented in the 2PWS. However, in Figure 1 As can be seen in (c), the double-differential PPG signal may have low quality due to noise or quantization in the original PPG signal.

[0069] Therefore, in a more preferred embodiment, a two-stage process is used to detect the onset of the pulse wave. Figure 5 (a) shows Figure 1 (a) shows the same 1-second PPG signal. Figure 5 (b) shows the first derivative of the PPG signal with respect to time (v-PPG), while Figure 5 (c) shows the second derivative of the PPG signal with respect to time (a-PPG). Figure 5 In (a), the start of the pulse wave for two cardiac cycles is indicated by a point marked 50; the purpose of this embodiment is to identify these points. In the first phase, peak detection is performed on the first derivative of the PWS with respect to time (1PWS). Figure 5 As can be seen in (b), the peak value in the v-PPG signal will roughly correspond to the steepest flank in the original PPG waveform (e.g., Figure 5 (as shown in (a)). The peak value detected in the v-PPG signal is marked as 52.

[0070] After detecting the maximum velocity peak 52 in 1PWS, then in the second stage, a narrowed (local) search window is applied to 2PWS based on the timing of each maximum velocity peak 52 (e.g., Figure 5 (c) shows a-PPG), and detects the maximum peak value within the narrowed search window on the 2PWS. The narrowed search window can have a duration of 30-100 ms. These maximum peak values ​​correspond to the desired a-reference point, i.e., the beginning of the upper pulse wave in the PWS signal. Figure 5 In the a-PPG waveform shown in (c), the narrowing search window is indicated by a horizontal line 54, which ends at the timing of the detected velocity peak 52. The largest peak identified in those narrowing search windows on the a-PPG waveform is marked as 56 and corresponds to the desired a reference point.

[0071] Since 2PWS may be noisy in the case of poorly quantized signals, it may be beneficial to perform some smoothing before detecting peaks in 1PWS. This smoothing can be applied to the PWS before differentiation or to the 1PWS before peak detection. In some embodiments, smoothing can be achieved by using filtering (e.g., Savitzky-Golay filtering). Figure 5 The smoothing line in (c) shows the smoothing process and the differentiation of the smoothed PWS / 1PWS to determine the result of 2PWS.

[0072] exist Figure 5 In (c), the maximum / minimum values ​​of the b, c, d, and e waves of the first cardiac cycle are also shown. It can be seen that the c and d waves are very small (even in this optimal case) and are generally difficult to detect robustly. This illustrates the preference for detecting the a reference point, but as mentioned above, other reference points can be targeted in step 40. Hereinafter, the detected a reference points for multiple cardiac cycles are used to perform averaging across multiple cardiac cycles to determine the average cardiac cycle waveform.

[0073] Next, in steps 42, 44, and 46, an averaging technique is applied to the PWS to determine the average cardiac cycle waveform at one or more time points Y within the time period covered by the PWS. See below for reference. Figure 6 Steps 42, 44, and 46 are described in the exemplary 2PWS shown in (b) (in the form of a-PPG). Figure 6 The 2PWS in (b) is derived from the PPG signal covering a 1-minute time period for a specific subject. This 1-minute time period covers 74 cardiac cycles of the subject. Figure 6 (b) marks the corresponding reference points identified in step 40 for each cardiac cycle in the a-PPG waveform. Figure 6 (a) shows the target and Figure 6 (b) Measurements of arterial blood pressure (ABP - thin line) and mean arterial pressure (MAP - thick line) of the same subject and time period related to the a-PPG signal. Figure 6 (a) shows an ABP measurement to provide Figure 6 (b) is the background of the a-PPG waveform, and it should be understood that ABP measurements are generally unavailable for objects being acquired for single-point PWS measurements. Figure 6 As can be seen in (a), the mean arterial pressure increased by approximately 20 mmHg between 20 and 40 seconds. This is observed in the a-PPG waveform ( Figure 6 In (b)), there are some visible changes at approximately 30 seconds (e.g., high breathing modulation from the a reference point from 0 to 30 seconds and low breathing modulation from the a reference point from 30 to 60 seconds). The averaging technique described herein can be used to identify and / or analyze morphological changes in the a-PPG waveform within and / or throughout the 60-second window.

[0074] To demonstrate how the morphology of a-PPG changes over a 60-second duration, Figure 7 It is shown Figure 6 (b) A graph of seven cardiac cycles of superimposed a-PPG waveforms. Figure 7 The seven cardiac cycles included Figure 6 (b) Marked as 1 to 7, they are relatively evenly spaced within a 1-minute time window. It should be noted that these 7 cardiac cycles are from... Figure 6 (b) The cardiac cycles identified are arbitrarily selected merely to provide a representation of how the cardiac cycle changes. Figure 7 In this study, the a-PPG waveforms of seven selected cardiac cycles overlap with the α-reference point of each aligned a-PPG waveform, and the corresponding a-PPG waveforms are normalized such that each a-PPG waveform has the same amplitude (amplitude = 1) at the aligned α-reference point, and at other times, the a-PPG waveforms have corresponding amplitudes typically in the range of -1 to 1. The time measured from the aligned reference point is called the "lag time" or "lag time value" and is denoted as T. Δk The aligned reference point corresponds to the lag time T. Δk = 0 seconds. Figure 7 Each of the seven waveforms shown is referred to herein as a “normalized” a-PPG waveform, meaning that they are normalized near a common portion of the cardiac cycle (e.g., a reference point “a” for each cardiac cycle in this example), where the amplitude of the a-PPG waveform matches at that reference point.

[0075] exist Figure 7As can be seen, the overall shape (after reference point a) changes over time due to variations in blood pressure. It can also be seen that, partly due to noise in the second derivative calculations, the waveform is difficult to identify or analyze independently.

[0076] Because the shape of the pulse wave can change continuously, such as Figure 7 As shown in the example, simply applying an average to multiple waveforms would lose high-frequency information relevant to the analysis of the reflected pulse wave. Therefore, an averaging method that can account for linear variations is needed. The averaging technique disclosed in this paper is based on the technique in WO2015 / 044010, which is extended to accommodate linear variations.

[0077] To describe the algorithm below, vectors x Defined as having N x / F s a-PPG data within a time window lasting seconds. Waveform selection. x The reference point (which is the peak location in the following working example, a reference point) is listed as having a length N. p vector p , where N p This refers to the number of reference points for cardiac cycles / recognition. Figure 6 and 7 In the example, N p It is 74. The waveform value at the detected peak position is determined by... This represents the expression where j = 0, ..., N. p -1 is the index of the peak value in each cardiac cycle. Values ​​from the peak location will be represented later in a shorter format.

[0078]

[0079] Where j = 0, ..., N p -1 is the index of the peak value.

[0080] The next step in the averaging process is to analyze N. p Each neighboring sample (to the left and right of the initial peak (reference point a)) is used to calculate the average change (decrease) compared to the initial peak level. Similar to the peak, neighboring positions relative to the peak location are defined in short form as:

[0081]

[0082] Where Δ k It is a lag exponent relative to the peak position, and it can be positive or negative. Lag exponent Δ k With lag time T Δk The relationship is as follows:

[0083] Δ k =TΔk ·F s (4)

[0084] For all adjacent positions relative to the peak position within a 1-minute window, the average level change (decline) is calculated as:

[0085]

[0086] Where N equals the value in the region [0...N] x The number of averages in [-1]

[0087] This means that N in the averaging process is not necessarily equal to the number of peaks N. p And it will be 1, 2, or less, depending on the value ( p ) j +Δ k Are some of them in a sequence with length N? x Outside the window.

[0088] For Δ k The calculation of the model using several negative and positive values ​​is called iteration or repetition. All averages in the average model can be calculated independently. Since only the average model values ​​during the systolic phase of the cardiac cycle are of interest, Δ can be applied. k The boundary between negative and positive values ​​for Δ. k The negative value of Δ can include only the negative value that is part of the start of systole. Because the peak position is very close to this start of cardiac contraction, Δ k The negative value can be restricted to, for example, -0.1 seconds. For Δ k The positive value includes the lag time that is part of the rest of the heart's contraction. Δ k The typical maximum positive value can be chosen to be equal to, for example, a lag time T of 0.4 seconds. Δk Now it is possible to target all lagging indicators Δ k Average normalized waveform w The calculation is as follows:

[0089]

[0090] in Calculate using equation (5).

[0091] Figure 8 It shows the superimposed on from Figure 7 The normalized a-PPG waveform shown is the average normalized waveform over seven cardiac cycles (thicker line 80). In fact, it is obtained by calculating the normalized a-PPG waveform at each lag time value (T...). Δk The average normalized waveform 80 from the above process is obtained by taking the average value (mean) of each normalized cardiac cycle at ) . For example, at the lag time TΔk = the value of the average waveform 80 at +100 ms (i.e., the normalized amplitude) is given by the mean (average) of the individual normalized a-PPG waveform values at lag time T Δk = +100 ms (i.e., normalized in time and amplitude with respect to the identified a-reference point for each cardiac cycle). This average waveform 80 can also be described as a 0th order polynomial fit for the normalized a-PPG cardiac cycle.

[0092] However, as Figure 8 shown, since the morphology of the waveform changes over time, this average over the entire 1-minute window does not fit well for all normalized cardiac cycles in the entire 1-minute window.

[0093] Therefore, the above averaging process is generalized to accommodate time-varying situations by calculating not simply the 0th order average of the N Δk values at lag time T p but rather a 1st order or higher polynomial fit of the N Δk values at lag time T p Since this polynomial fit of order n (where n is equal to or greater than 1) will have (lag) time on the x-axis and the level drop value on the y-axis, the averaged (or curve fit) level drop for polynomial orders greater than 0 will also have a dependence on time. For each lag time Δ k , m (m < n) polynomial curve fit coefficients can be calculated. These coefficients a0, …, a m can be calculated in such a way as to obtain a minimization in the least squares sense:

[0094]

[0095] where is the linear fit model:

[0096]

[0097] Next, based on the polynomial coefficients a0(Δ k ), …, a m (Δ k ), the averaged level change (drop) can be calculated directly via the linear fit model as:

[0098]

[0099] It can be seen that for n = 0, we obtain the averaged level change (drop) as given by equation (5), and by using equation (6), we obtain what is described in WO2015 / 044010 and in Figure 8The average fit model shown will not depend on the actual location within the time window (e.g., a 1-minute window in this example). However, for n>0, the average fit model will depend on the actual location within the time window. This is in Figure 9 The diagram in the middle shows, Figure 9 The normalized a-PPG waveform values ​​at a lag time of +200 ms are shown for the identified reference point 'a' relative to 74 cardiac cycles in the example above. The normalized a-PPG waveform values ​​are indicated by line 90. Therefore, each value in line 90 corresponds to the normalized a-PPG waveform value at a lag time of +200 ms for one of the cardiac cycles within the time window. For example, the value of line 90 for the 10th cardiac cycle is the normalized a-PPG waveform value for the 10th cardiac cycle at a lag time of +200 ms from the reference point 'a', the value of line 90 for the 25th cardiac cycle is the normalized a-PPG waveform value for the 25th cardiac cycle at a lag time of +200 ms from the reference point 'a', and so on. Figure 9 Line 90 in the diagram illustrates how the value of the normalized a-PPG waveform at +200 ms after the reference point changes within the time window across the cardiac cycle in the PWS.

[0100] Figure 9 The zeroth-order average of the normalized a-PPG waveform values ​​at a lag time of +200 ms is also shown (line 92 - which corresponds to the conventional method of averaging (derived average) all values ​​at that lag time only), the first-order average of the normalized a-PPG waveform values ​​at a lag time of +200 ms (line 94), and the second-order average of the normalized a-PPG waveform values ​​at a lag time of +200 ms (line 96).

[0101] Determining a series of lag time values Figure 9 The corresponding version (i.e., the corresponding fitted model). That is to say, Figure 9 The corresponding version is formed by the value of the normalized a-PPG waveform at the corresponding lag time. For example, for each lag time such as 10ms, 50ms, 100ms, etc., there can be... Figure 9 The corresponding version. Figure 9 These versions will show how the values ​​of the normalized a-PPG waveform at the corresponding lag time values ​​change over time across the cardiac cycle in the PWS. For Figure 9 For each corresponding version, the corresponding 0th, 1st, and / or subsequent order averages can be determined.

[0102] Figure 10 The graph in the figure shows the lag time value (T) for n=1. Δk The average (time-varying) fitted model within the range of ) Figure 6The average normalized cardiac cycle waveform of Y at several time points in the a-PPG signal in (b). Therefore, Figure 10 The display shows the lag time (T) between -100ms and +400ms. Δk The range of values ​​determines the average waveform generated by the first-order average of the normalized a-PPG waveform values. This is just an example. Figure 10 The average waveforms at time point Y = 0 seconds (labeled 102) and time point Y = 1 minute (labeled 104) are shown, along with the corresponding waveforms. Figure 6 (b) shows the average waveform at time points 1-7 marked in the diagram. Time point Y = 0 seconds corresponds to... Figure 6 The PWS in the image covers the start of a 60-second time period, and the time point Y = 60 seconds corresponds to... Figure 6 The end of the time period covered by PWS in the PWS.

[0103] The average normalized cardiac cycle waveform is based on the lag time value T. Δk The scope (e.g., in) Figure 6-10 The example in the text (between -100ms and +400ms) is derived. The values ​​forming the average normalized cardiac cycle waveform are taken from the lag time value T. Δk The set of average (time-varying) fitted models (expected values ​​for n) within a range. In other words, for a selected time point Y where the cardiac cycle waveform needs to be averaged and normalized, the values ​​that form the averaged and normalized cardiac cycle waveform in the range of -100ms to +400ms are for a lag time T. Δk Fitting models within this range (e.g., Figure 9 The value at time point Y in lines 94 and 96.

[0104] As an example, consider a lag time of +200ms. Figure 9 Furthermore, other lag time values ​​within the range of -100ms to +400ms are considered. Figure 9 The corresponding version. The expected value of n is 1, and the average normalized cardiac cycle waveform (which is shortened to "average waveform" in this paper) will be derived for the time point Y = 0 seconds (i.e., at the beginning of the time period covered by PWS). The value of the average waveform at the lag time + 200 ms is Figure 9 The value of line 94 at time point Y=0 (or at the cardiac cycle index corresponding to Y=0). From Figure 9 and Figure 10A comparison of the average cardiac cycle waveform values ​​at a lag of +200 ms shows that for the average waveform at time point Y = 0, the value at a lag of +200 ms is -0.3 (i.e., for Y = 0 (or a cardiac cycle with index 0), the first-order polynomial fit 94 has a value of -0.13), and for the average waveform at time point Y = 1 minute, the value at a lag of +200 ms is approximately 0 (i.e., for Y = 1 minute (or a cardiac cycle with index 73), the first-order polynomial fit 94 has an approximation of 0). This is repeated for other lag values ​​(i.e., using...). Figure 9 Other versions (of which) are used to derive the fully averaged waveform in the range of -100ms to +400ms. Therefore, the average waveform at Y=0 has a lag time T. Δk The value at -100ms corresponds to line 94 (for a lag time T). Δk = -100ms) at the value of Y = 0 (or the cardiac cycle with index 0), while the average waveform at Y = 0 is at a lag time T. Δk The value at +100ms is for the lag time T. Δk = +100ms corresponds to the value of line 94 at Y=0, etc.

[0105] Therefore, the average cardiac cycle waveform for a selected time within a 1-minute time window can be derived from the corresponding polynomial fit at each lag time value. Furthermore, it can be seen that all intermediate average waveform results at different times within the 1-minute window can also be derived (for example, for...). Figure 6 (b) The normalized a-PPG waveforms marked 1 to 7 correspond to the corresponding times. Therefore, by using an order n ≥ 1, time-varying variations are adapted during the averaging process. It should be noted that although n can be any integer value equal to or greater than 1, n = 1 has yielded good results in practice for PWS representing the cardiac cycle, where the PWS is analyzed over relatively short time periods and the potential changes in the waveform are straightforward. If greater changes in morphology are expected over time, n can be set to a value higher than 1.

[0106] Figure 4 Steps 42, 44, and 46 implement the aforementioned averaging technique, as shown below. According to the aforementioned averaging technique, steps 42, 44, and 46 operate on the "normalized 2PWS". The normalized 2PWS is obtained by individually normalizing each portion of the 2PWS corresponding to the cardiac cycle at an identified reference point relative to the corresponding cardiac cycle identified in step 40. That is, for a specific cardiac cycle identified in step 40, the amplitude of the 2PWS corresponding to that cardiac cycle is normalized around the amplitude of the 2PWS at the identified reference point of that cardiac cycle.

[0107] In step 42, the first lag time value T is measured for the reference point identified by the reference. Δk An nth-order polynomial fit is determined for the first set of normalized 2PWS values. As mentioned above, n is equal to or greater than 1. The first set of normalized 2PWS values ​​includes the 2PWS values ​​occurring at a distance of a first lag time value from each identified reference point of the cardiac cycle. That is, in step 42, for a lag time value of X ms, the first set of values ​​is the normalized 2PWS value, which is X ms away from each reference point identified in step 40. Figure 9 In the example shown, the first set of values ​​corresponds to line 90. Therefore, for example, for a lag time value T of 200 ms... Δk ,For example, Figure 9 The value of the 20th identified cardiac cycle is the normalized 2PWS value of the 20th identified cardiac cycle at 200 ms. Line 90 is formed by the normalized 2PWS value of each identified cardiac cycle at 200 ms. The nth-order polynomial fit (n>1) determined in step 42 corresponds to line 94 for n=1 and line 96 for n=2.

[0108] In step 44, step 42 is repeated once or more for one or more additional lag time values. Therefore, in step 44, one or more additional iterations of step 42 are performed for one or more additional lag time values ​​to determine corresponding additional n-order polynomial fits 94, 96 for the corresponding group values ​​of the normalized 2PWS. Each of the corresponding group values ​​of the normalized 2PWS includes a normalized 2PWS value occurring at a distance from the reference point of each identified cardiac cycle at the corresponding additional lag time value. Therefore, step 44 results in a result for the corresponding lag time value T. Δk Export Figure 9 One or more corresponding versions.

[0109] As described below with reference to step 46, the number of times step 42 is repeated determines the temporal resolution of the average cardiac cycle waveform determined in step 46. The more times step 42 is repeated, the smoother and more resolution the resulting average cardiac cycle waveform will be. In some embodiments, step 42 may be repeated for lag time values ​​in the range of -100 ms to +400 ms. It should be understood that the upper limit of the lag time value range may be affected by the heart rate of the subject (higher heart rates shorten the range of lag time values, while lower heart rates make the range of lag time values ​​wider). The lag time value range should cover one cardiac cycle or fewer (but enough cardiac cycles for pulse wave characteristics, such as reflected pulse waves observed in the resulting average cardiac cycle waveform).

[0110] Next, in step 46, a first average cardiac cycle waveform is formed for a first time point Y within the time period covered by the PWS. For example, for the coverage derived from it... Figure 6 (b) The PPG signal of the 2PWS for a 1-minute time period, the first time point Y can be any of the following: 0 seconds (i.e., at the beginning of the 1-minute time period), 25 seconds (i.e., through the middle of the 1-minute time period), 60 seconds / 1 minute (i.e., at the end of the 1-minute time period), etc.

[0111] In step 46, instead of evaluating equation (8) above, the coefficient a0(Δ) is evaluated at time point Y. k ),…,a m (Δ k ):

[0112]

[0113] Where parameter F s The sampling rate of the PWS signal is used to convert time point Y into the number of samples, similar to equation (8).

[0114] The first average cardiac cycle waveform is formed by the (normalized amplitude) values ​​of multiple n-order polynomial fits 94, 96 at the first time point. Therefore, for time point Y in the time period covered by the PWS, the value of the average cardiac cycle waveform at time point Y is given by the (normalized average) value of the n-order polynomial fits 94, 96 of the first set of values ​​determined in step 42 at time point Y (i.e., the value of the n-order polynomial fits 94, 96 of the first lag time value at time point Y) and the corresponding value of each of the other n-order polynomial fits 94, 96 determined in step 44 at time point Y (i.e., the value of the n-order polynomial fits 94, 96 of the other lag time values ​​at time point Y).

[0115] In the specific example where time point Y = 0 and n = 1, the average cardiac cycle waveform is formed by the (normalized amplitude) value at time point Y in the first-order polynomial fit 94 for each lag time value. For example, step 46 can produce, for instance, the waveform of the average cardiac cycle. Figure 10 The average cardiac cycle waveforms 102 and 104 are shown. More generally, step 46 is as described above (refer to the above reference). Figure 9 and 10 It was implemented as described.

[0116] In some embodiments, the average cardiac cycle waveform formed in step 46 may be analyzed to determine information about the subject's health status. In some embodiments, the information about the health status is a measurement or indication of the subject's blood pressure and / or a measurement or indication of changes in the subject's blood pressure.

[0117] In some embodiments, a second average cardiac cycle waveform can be formed for a second time point within the time period covered by the PWS. The second average cardiac cycle waveform can be formed in the same manner as the first average cardiac cycle waveform determined in step 46. For example, one of the first and second time points may be at or near the beginning of the PWS, and the other may be at or near the end of the PWS. In other embodiments, one or more additional average cardiac cycle waveforms can be determined for corresponding time points within the time period covered by the PWS.

[0118] In embodiments of forming a second average cardiac cycle waveform, the method may further include comparing the first and second average cardiac cycle waveforms to determine or identify changes in the average cardiac cycle waveform between a first time point and a second time point. This step may include determining a difference signal representing the difference between the two average cardiac cycle waveforms for all lag time values, and / or determining the difference between one or more specific portions of the average cardiac cycle waveform. For example, the difference between the amplitudes of minimum values ​​in the average cardiac cycle waveform after an identified reference point a may be determined. As another example, the difference between the timings of different reference points in the two average cardiac cycle waveforms may be determined. For example, a reference point e may be identified in each average cardiac cycle waveform, and the time between the corresponding reference points a and e in each average cardiac cycle waveform may be compared.

[0119] In some embodiments, a measure of a subject's blood pressure or a measure of a change in blood pressure can be determined based on a first mean cardiac cycle waveform and a second mean cardiac cycle waveform. In some embodiments, blood pressure or a measure of a change in blood pressure can be determined based on a comparison of the first mean cardiac cycle waveform and the second mean cardiac cycle waveform. In a particular embodiment, a blood pressure surrogate measurement can be determined based on the time between reference point a and reference point c, which corresponds to pulse reflex time (PRT).

[0120] Figure 11 This is a functional block diagram illustrating various operations of the PWS analysis according to various exemplary embodiments. It has a duration N. x / F s The PWS 100 (e.g., the PPG signal) is received by the derivative calculation block 102 and the peak lookup block 104. The derivative calculation block 102 determines the first derivative of the PWS with respect to time (1PWS) and the second derivative of the PWS with respect to time (2PWS). The 1PWS and / or 2PWS signals can be provided to the peak lookup block 104. The peak lookup block 104 evaluates the input signal to identify the cardiac cycle and a reference point corresponding to a portion of the cardiac cycle in the 2PWS (e.g., an early / upper systolic reference point). The operation of the derivative calculation block 102 and the peak lookup block 104 generally corresponds to step 40 described above.

[0121] The 2PWS and the identified reference point are input to averaging block 106. The expected value 108 of n for polynomial fitting is input to averaging block 106. Alternatively, the expected value of n can be predetermined or preset in averaging block 106. Averaging block 106 performs the waveform averaging process described above in steps 42, 44, and 46 based on the input 2PWS and the identified reference point. Averaging block 106 outputs at least one average cardiac cycle waveform at a specific time point. Figure 11 In the diagram, the average block 106 is shown as outputs Y = 0 and Y = -N. x / F s (For example, the average cardiac cycle waveform in the minute before t=0).

[0122] In an optional embodiment, the average cardiac cycle waveform can be input to analysis block 110, which can perform some analysis of the average cardiac cycle waveform to determine information about the subject's health status. In some embodiments, analysis block 110 can determine alternative blood pressure measurements. As an example, analysis block 110 can subtract two average cardiac cycle waveforms as follows:

[0123]

[0124] in: It is the waveform of the first average cardiac cycle at time point Y=0, and It is the second average cardiac cycle waveform at time point Y = 60 seconds.

[0125] The value indicates the time lag T. Δk The energy changes in the pulse wave over a window length (e.g., 1 minute). Since changes in blood pressure will cause changes in the energy in the pulse wave at the lag time of the expected reflection (e.g., a reflection from a branch of the renal artery), this can be utilized. The value is used to estimate the amount of blood pressure change during the window length process.

[0126] Therefore, techniques for improving the averaging of cardiac cycle waveforms are provided.

[0127] Those skilled in the art, through studying the accompanying drawings, disclosure, and claims, will be able to understand and implement variations of the disclosed embodiments in practice with respect to the principles and techniques described herein. In the claims, the word "comprising" does not exclude other elements or steps, and the quantifiers "a" or "an" do not exclude a plurality. A single processor or other unit may implement the functions of several items recited in the claims. Although certain measures are recited in dissimilar dependent claims, this does not imply that combinations of these measures cannot be advantageously used. Computer programs may be stored or distributed on suitable media, such as optical storage media or solid-state media supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems. No reference numerals in the claims should be construed as limiting the scope.

Claims

1. A computer-implemented method for analyzing pulse wave signals (PWS) obtained from an object, wherein, The PWS includes pulse wave measurements of the object over multiple cardiac cycles during a first time period, and the method includes: (i) Analyze the PWS described in (40) to identify multiple cardiac cycles and the corresponding reference point for each identified cardiac cycle; (ii) Define 2PWS as the second derivative of PWS with respect to time; (iii) The normalized 2PWS is determined by normalizing the amplitude of the 2PWS at a reference point of the identification of the cardiac cycle for each part of the 2PWS corresponding to the corresponding identified cardiac cycle. (iv) For the first lag time value, determine (42) an nth-order polynomial fit for the first set of values ​​of the normalized 2PWS, wherein the first set of values ​​of the normalized 2PWS includes the values ​​of the normalized 2PWS that occur at the reference point of each identified cardiac cycle at a distance from the first lag time value, wherein n is equal to or greater than 1. (v) Perform one or more additional iterations of step (iv) of (44) for one or more additional lag time values ​​to determine a corresponding additional n-order polynomial fit for the corresponding group values ​​of the normalized 2PWS, wherein the corresponding group values ​​of the normalized 2PWS include the normalized 2PWS values ​​at the corresponding additional lag time values ​​occurring at the reference point for each identified cardiac cycle; and (vi) Forming (46) the first average cardiac cycle waveform at the first time point in the first time period, wherein the first average cardiac cycle waveform is formed by fitting values ​​of multiple nth-order polynomials at the first time point.

2. The method according to claim 1, wherein, The method further includes: A second average cardiac cycle waveform is formed at a second time point in the first time period, wherein the second average cardiac cycle waveform is formed by fitting values ​​of multiple n-order polynomials at the second time point.

3. The method according to claim 2, wherein, The method further includes: The first average cardiac cycle waveform and the second average cardiac cycle waveform are compared to determine the change in the average cardiac cycle waveform between the first time point and the second time point.

4. The method according to claim 2 or 3, wherein, The method further includes: The measurement of the subject's blood pressure or the measurement of the subject's blood pressure change is determined based on the first average cardiac cycle waveform and the second average cardiac cycle waveform.

5. The method according to claim 1, wherein, The method further includes: The first average cardiac cycle waveform is processed to determine a measure of the subject's blood pressure.

6. The method according to any one of claims 1-3, wherein, Each of the first lag time value and the one or more other lag time values ​​is equal to or less than the duration of the object's cardiac cycle.

7. The method according to any one of claims 1-3, wherein, The reference point for each identified cardiac cycle is the start of the object's pulse wave.

8. An apparatus (30) for analyzing pulse wave signals (PWS) obtained from an object, wherein, The PWS includes pulse wave measurements of multiple cardiac cycles of the object during a first time period, and the device (30) is configured to: (i) Analyze the PWS to identify multiple cardiac cycles and the corresponding reference point for each identified cardiac cycle; (ii) Define 2PWS as the second derivative of PWS with respect to time; (iii) The normalized 2PWS is determined by normalizing the amplitude of the 2PWS at a reference point of the identification of the cardiac cycle for each part of the 2PWS corresponding to the corresponding identified cardiac cycle. (iv) For the first lag time value, determine an n-order polynomial fit for the first set of values ​​of the normalized 2PWS, wherein the first set of values ​​of the normalized 2PWS includes the values ​​of the normalized 2PWS that occur at the reference point of each identified cardiac cycle at a distance from the first lag time value, wherein n is equal to or greater than 1. (v) Perform one or more additional iterations of operation (iv) for one or more additional lag time values ​​to determine a corresponding additional n-order polynomial fit for the corresponding group values ​​of the normalized 2PWS, wherein the corresponding group values ​​of the normalized 2PWS include the normalized 2PWS values ​​at the corresponding additional lag time values ​​occurring at the reference point for each identified cardiac cycle; and (vi) Forming a first average cardiac cycle waveform at a first time point in the first time period, wherein the first average cardiac cycle waveform is formed by fitting values ​​of multiple n-order polynomials at the first time point.

9. The apparatus (30) according to claim 8, wherein, The device (30) is also configured to: A second average cardiac cycle waveform is formed at a second time point in the first time period, wherein the second average cardiac cycle waveform is formed by fitting values ​​of multiple n-order polynomials at the second time point.

10. The apparatus (30) according to claim 9, wherein, The device (30) is also configured to: The first average cardiac cycle waveform and the second average cardiac cycle waveform are compared to determine the change in the average cardiac cycle waveform between the first time point and the second time point.

11. The apparatus (30) according to claim 9 or 10, wherein, The device (30) is also configured to: The measurement of the subject's blood pressure or the measurement of the subject's blood pressure change is determined based on the first average cardiac cycle waveform and the second average cardiac cycle waveform.

12. The apparatus (30) according to claim 8, wherein, The device (30) is also configured to: The first average cardiac cycle waveform is processed to determine a measure of the subject's blood pressure.

13. The apparatus (30) according to any one of claims 8-10, wherein, Each of the first lag time value and the one or more other lag time values ​​is equal to or less than the duration of the object's cardiac cycle.

14. The apparatus (30) according to any one of claims 8-10, wherein, The reference point for each identified cardiac cycle is the start of the object's pulse wave.

15. A computer program product comprising a computer-readable medium, the computer-readable medium having embedded computer-readable code configured to cause the computer or processor, when executed by a suitable computer or processor, to perform the method according to any one of claims 1-7.

Images