KR102627661B1 - Method for analyzing photoplethysmography data and recording medium storing program to implement the method - Google Patents

Abstract

A method for analyzing photoplethysmographic data is disclosed. The photoplethysmography data analysis method is implemented by a computer including a memory storing computer-executable instructions and a processor that executes the computer-executable instructions, and the photoplethysmography data analysis method is implemented by receiving PPG (photoplethysmography) data. PPG data receiving step; A differential data generation step of differentiating the PPG data to generate differential data; A signal processing step of generating analysis data by signal processing the PPG data and differential data; A feature point detection step of detecting feature points present in the analysis data using a pre-learned analysis algorithm; and an additional learning step of training the analysis algorithm using feature point data detected by the analysis algorithm.

Description

Photoplethysmography data analysis method and recording medium {METHOD FOR ANALYZING PHOTOPLETHYSMOGRAPHY DATA AND RECORDING MEDIUM STORING PROGRAM TO IMPLEMENT THE METHOD}

This disclosure relates to a method for analyzing photoplethysmographic data.

Photoplethysmography (PPG) is a signal that measures changes in blood volume within blood vessels according to heartbeat using absorption, reflection, and scattering of light, and is one of the most frequently measured biosignals along with electrocardiogram (ECG). These photoplethysmographic pulse waves can measure parameters such as heart rate (Herat Rate, HR), heart rate variation (HRV), and oxygen saturation (SpO ₂ ) by monitoring the waveform variation period of the sensed PPG signal. At this time, the measured parameters are applied to various diagnoses such as arrhythmia, blood pressure changes, heart disease, and changes in the autonomic nervous system. In this regard, Republic of Korea Patent Publication No. 10-2021-0097512 proposes a technology for obtaining the user's heart rate by analyzing PPG data.

However, photoplethysmography can provide diagnostic parameters regarding abnormal blood flow within blood vessels, but when measuring biosignals using a photoplethysmography sensor, noise generated from heartbeat and motion noise generated from subject movement Due to this, errors occurred and reliability was low when analyzing the waveform of the PPG signal.

Meanwhile, methods for analyzing photoplethysmography using artificial intelligence technology are also being studied, but this method requires a very large amount of learning data for artificial intelligence analysis and requires expensive resources such as high-performance computers for machine learning. Not only was this time-consuming, but real-time analysis was difficult because a lot of input data was used. Errors occurred when analyzing the waveform of the PPG signal and the reliability of the results was low, so the development of new technology was required.

Public Patent Publication No. 10-2021-0007368 (2021.01.20.) Public Patent Publication No. 10-2020-0054723 (2020.05.20.) Public Patent Publication No. 10-2016-0000810 (2016.01.05.)

The technical idea of the present disclosure is to solve the above-mentioned problems, and its purpose is to provide a photoplethysmographic data analysis technology that automatically analyzes the waveform of the PPG signal and increases the reliability of the measurement results.

In addition, the technical idea of the present disclosure has another purpose of providing a photoplethysmography data analysis technology that enables real-time data processing by reducing the amount of calculation of input data and can be implemented even in low-performance hardware.

The problem to be solved by the present invention is not limited to the above-mentioned problem, and other technical problems not mentioned will be clearly understood by those skilled in the art from the contents described later.

In order to achieve this object, in one embodiment of the present invention, the photoplethysmographic data analysis method is implemented by a computer including a memory storing computer-executable instructions and a processor executing the computer-executable instructions, The plethysmographic data analysis method includes a PPG data reception step of receiving PPG (photoplethysmography) data; A differential data generation step of differentiating the PPG data to generate differential data; A signal processing step of generating analysis data by signal processing the PPG data and differential data; A feature point detection step of detecting feature points present in the analysis data using a pre-learned analysis algorithm; and an additional learning step of training the analysis algorithm using feature point data detected by the analysis algorithm.

In addition, in the feature point detection step, the feature points include S point, which is the maximum value point due to heart contraction among the PPG data, O point, which is the minimum value point just before heart contraction among the PPG data, and an area where heart rate cannot be measured among the PPG data. It may include at least one of the ES point, which is the starting point of the error area, the EE point, which is the end point of the error area, the W point, which is the maximum value point among the differential data, and the Z point, which is the first peak caused by a reflected wave among the differential data. there is.

And, in the feature point detection step, the analysis algorithm includes a deep neural network (DNN) filter including a plurality of hidden layers between the input layer and the output layer, and the DNN filter may be provided in single or plural numbers depending on the type of the feature point. You can.

In addition, in the signal processing step, a window with a certain size moves by a preset time unit and signal processes PPG data and differential data corresponding to the area of the window to generate analysis data. The time unit for which the window moves is By being set small compared to the size of the window, some areas of continuously generated analysis data may overlap with each other.

In addition, in the feature point detection step, the analysis data is input to the DNN filter and analysis is performed to detect feature points included in the analysis data. However, when analyzing analysis data with some areas overlapping with each other, the same feature points are overlapped. It is detected, and the decision results for the same feature point can be accumulated and stored.

And, in the additional learning step, the DNN filter is further trained using information about the feature points detected by the DNN filter in the PPG data and information about feature points detected by the DNN filter in the differential data as learning data. You can.

Meanwhile, the photoplethysmography data analysis method includes a biometric information derivation step of deriving biometric information using the feature point data detected in the feature point detection step; and a reliability evaluation step of evaluating the reliability of the measurement value of a measurement device that measures the user's biometric information, wherein the reliability evaluation step includes the biometric information value derived in the biometric information derivation step and the measurement of the measurement device. By comparing the values, if the difference value is less than a predetermined standard value, the reliability of the measuring equipment can be evaluated as high, and if the difference value exceeds the predetermined standard value, the reliability of the measuring equipment can be evaluated as low.

In order to achieve this object, as another embodiment of the present invention, a program for performing the above-described photoplethysmographic data analysis method may be recorded on a computer-readable recording medium.

The means for solving the above problems are merely illustrative and should not be construed as intended to limit the present invention. In addition to the exemplary embodiments described above, there may be additional embodiments described in the drawings and detailed description of the invention.

As described above, according to various embodiments of the present invention, the waveform of the PPG signal is automatically analyzed, and the accuracy and reliability of the analysis results are improved as the judgment results for the feature points of the PPG data are accumulated.

In addition, according to various embodiments of the present invention, in the feature point detection step, the same feature point is detected redundantly in two analysis data in which some areas overlap, the decision results for the same feature point are accumulated, and the accumulated decision result is Since DNN filters can be additionally learned using this method, accuracy increases while reducing the number of neurons and data computation compared to existing artificial neural networks. It can be implemented quickly even on relatively low-spec or low-performance hardware, and has the advantage of being able to process in real time.

In addition, according to various embodiments of the present invention, by installing the analysis algorithm or DNN filter in the server or measurement equipment (e.g., oxygen saturation measurement equipment, blood pressure monitor, etc.), more reliable biometric data can be obtained from abnormally measured PPG data. Information (eg, heart rate, heart rate variability, oxygen saturation, etc.) can be obtained.

Effects according to various embodiments of the present invention are not limited to the effects mentioned above, and other effects not mentioned may be clearly understood by those skilled in the art from the description of the claims.

Figure 1 is a flowchart schematically showing a method of analyzing photoplethysmographic data according to an embodiment of the present invention.
Figure 2 schematically shows the process of signal processing PPG data and differential data and the process of detecting feature points using a DNN filter in one embodiment of the present invention.
Figure 3 schematically shows PPG data and differential data according to an embodiment of the present invention.
Figure 4 schematically shows the process of generating data for analysis while a window moves in one embodiment of the present invention.
Figure 5 is a photograph of normal and abnormal measurements of heart rate using a pulse oximeter to secure normal and abnormal data.
Figure 6 shows the heart rate measured by a pulse oximeter and the heart rate measured by the photoplethysmography data analysis method according to an embodiment of the present invention.
Figure 7 is a result derived by the photoplethysmography analysis method according to an embodiment of the present invention, showing the recognition score for the heartbeat derived during normal and abnormal measurements.

Preferred embodiments of the present invention will be described in more detail with reference to the attached drawings, but technical parts that are already well-known will be omitted or compressed for brevity of explanation.

It should be noted that references herein to “one” or “one” embodiment of the invention do not necessarily refer to the same embodiment, and they refer to at least one.

In the following examples, singular expressions include plural expressions unless the context clearly indicates a different meaning.

In the following embodiments, terms such as include or have mean that the features or components described in the specification exist, and do not exclude in advance the possibility of adding one or more other features or components.

In cases where an embodiment can be implemented differently, a specific process sequence may be performed differently from the described sequence. For example, two processes described in succession may be performed substantially simultaneously, or may be performed in an order opposite to the order in which they are described. That is, each step of the method described herein can be appropriately performed in any order unless otherwise stated in the specification or clearly conflicting from the context.

Since each configuration shown in the drawings is arbitrarily shown for convenience of explanation, the present invention is not necessarily limited to what is shown.

Figure 1 is a flowchart schematically showing a method of analyzing photoplethysmographic data according to an embodiment of the present invention, and Figure 2 is a process of signal processing PPG data and differential data and using a DNN filter in an embodiment of the present invention. This schematically shows the process of detecting feature points. Figure 3 schematically shows PPG data and differential data according to an embodiment of the present invention, and Figure 4 shows the window moving in an embodiment of the present invention. It schematically shows the process of generating data for analysis, and Figure 5 is a photograph of normal and abnormal heart rate measurements using a pulse oximeter to secure normal and abnormal data, and Figure 6 is The heart rate measured by the pulse oximeter and the heart rate measured by the photoplethysmography data analysis method according to an embodiment of the present invention are shown, and Figure 7 shows the heart rate measured by the photoplethysmography data analysis method according to an embodiment of the present invention. As a result, it shows the recognition of heartbeat during normal and abnormal measurements.

The photoplethysmographic data analysis method according to an embodiment of the present invention will be described according to the flowchart shown in Figure 1 and will be described with reference to the drawings shown in Figures 2 to 7, but will be explained in order for convenience. . In one embodiment, the photoplethysmography data analysis method may be implemented by a computer. A computer may include a communication device that communicates through at least one communication network, a display that outputs screen information, a memory in which computer-executable instructions are stored, and a processor that is electrically connected to the memory and executes computer-executable instructions.

In this specification, the analysis algorithm is an algorithm that determines the reliability of data measured by measuring equipment (e.g., oxygen saturation measuring equipment, blood pressure monitor, etc.) using feature point data detected by a neural network filter, and is stored in the form of a program in the computer's memory. Can be installed.

1. PPG data reception step <S101>

In this step, PPG data can be received from the PPG measurement sensor. Here, PPG data is input data input to the computer for feature point analysis. According to one embodiment, the PPG data may include waveform information as shown in the PPG of FIG. 3.

2. Differential data generation step <S102>

In this step, differential data can be generated by differentiating the PPG data received in step S101. Referring to Figure 2, in one embodiment, a window W having a constant size (for example, 1.2 seconds) moves over time and performs first differentiation to generate differential data corresponding to PPG data. You can. In this specification, the size (I) of the window (W) means the width of the window (W), and the width may be applied in units of time.

3. Signal processing step <S103>

In this step, data for analysis can be generated by signal processing the PPG data and differential data. In one embodiment, the signal processing step may include a focusing step and a normalizing step. First, in the focusing step, the size of PPG data and differential data can be reduced. Referring to Figure 2, in the focusing step, the size of the data can be reduced by partially extracting data corresponding to a certain section within the area of the window W. In the normalizing step, the y-axis range of the data on which the focusing step was performed can be normalized to -1 to 1. For example, in the normalizing step, among the data on which the focusing step was performed, the value with the smallest y value is set to -1, the value with the largest y value is set to 1, and the remaining values are between -1 and 1 according to the ratio. It can be generalized to the value of .

Referring to Figure 4, in this step, a window (W) with a certain size (I) (e.g., 1.2 seconds) moves by a preset time unit (T) (e.g., 20 milliseconds) while the window ( Data for analysis can be generated by signal processing the PPG data and differential data corresponding to the area of W). At this time, as the time unit in which the window W moves is set to be smaller than the size I of the window W, some areas of two successively generated analysis data may overlap each other.

4. Feature point detection step <S104>

In this step, feature points existing within the entire area of analysis data can be detected using a pre-learned analysis algorithm. In one embodiment, the analysis algorithm may be pre-trained based on training data that is labeled by an expert and whose answers are predetermined. For example, the analysis algorithm may be pre-trained using learning data labeled with feature point information in PPG data.

Referring to Figure 3, the feature points are the S point, which is the maximum value point due to heart contraction among the PPG data, the O point, which is the minimum value point just before heart contraction among the PPG data, and the starting point of the error area, which is an area where heart rate cannot be measured among the PPG data. It can include at least one of the ES point, the EE point, which is the end point of the error area, the W point, which is the maximum value point among the first-differentiated differential data, and the Z point, which is the first peak caused by a reflected wave among the first-differentiated differential data. there is.

In one embodiment, the analysis algorithm may include a DNN filter including a plurality of hidden layers between the input layer and the output layer. Additionally, in one embodiment, DNN filters may be applied singly or plurally depending on the type of feature point. For example, if there are 6 types of feature points to be detected in this step, a total of 6 DNN filters can be used so that the DNN filter is applied to each feature point.

In addition, in one embodiment, the DNN filter is learned in advance using learning data for which the answer is predetermined (e.g., information about feature points present in specific PPG data, information about feature points present in specific differential data, etc.). It can be applied as a DNN filter.

Also, in this step, when analysis data is input to the DNN filter, the DNN filter can analyze the analysis data and detect feature points included in the analysis data. At this time, since the same feature point may exist in two pieces of analysis data in which some areas overlap each other, the same feature point may be detected repeatedly, and the decision results determined for the same feature point may be accumulated and stored. Referring to Figure 4, some areas of the first analysis data (D1), the second analysis data (D2), and the third analysis data (D3) overlap each other, and the S point, which is the same feature point, within the overlapped area Since (S) exists, the DNN filter can accumulate the decision results of detecting point S (S) and store the decision result value for point S (S) as 3.

5. Additional learning step <S105>

In this step, a process of learning the DNN filter can be performed using the feature point data detected by the DNN filter. According to one embodiment, in this step, the DNN filter can be further trained by using information about the feature points detected in the PPG data by the DNN filter in step S104 and information about the feature points detected in the differential data as learning data. . When learning a DNN filter, weights are added to the neurons included in each layer and adjustments are made repeatedly, and through this, an optimal algorithm that produces high-accuracy judgment results can be constructed.

In addition, depending on implementation, a biometric information derivation step (S106), a scoring step (S107), and a reliability evaluation step (S108) may be further performed after this step. In the biometric information derivation step according to one embodiment, biometric information (e.g., heart beats per minute) is derived using the interval between the previous location and the current location of the feature point (e.g., W point) determined by the DNN filter. You can. Additionally, in the scoring step according to one embodiment, a process may be performed to quantify the degree to which the DNN filter recognizes the heartbeat in the feature point detection step and express it as a quantitative value. For example, in the scoring step, the processor can calculate heart rate recognition using Equation 1 below.

[Equation 1]

RS = S+O+W+(40-ES-EE)

(Here, RS is the heart rate recognition, S is the number of judgments for point S determined by the DNN filter, O is the number of judgments for point O determined by the DNN filter, W is the number of judgments for point W determined by the DNN filter, and ES is the number of judgments for point W determined by the DNN filter. (where EE is the number of judgments for ES points determined by the filter, EE is the number of judgments for EE points determined by the DNN filter)

In the reliability evaluation step according to one embodiment, the communication device of the computer receives the measured value (e.g., heart rate per minute) measured by a measuring device (e.g., a blood pressure monitor) that measures the user's biometric information, and the processor receives the measured value (e.g., heart rate per minute) of the measuring device. If the measured value is compared with the biometric information value (e.g., heart rate per minute) derived in the biometric information derivation step and the difference between the measured value and the biometric information value is less than a predetermined standard value, the reliability of the measurement value measured by the measuring equipment is low. It can be judged to be high. If the difference between the measured value and the biometric information value exceeds the predetermined standard value, the reliability of the measured value of the measuring equipment is evaluated as low, and the measurement is performed through the computer's output device (e.g. display, speaker, etc.). You can output a guidance message recommending re-measurement with the equipment, or print biometric information values in place of unreliable measurement values.

Various embodiments of the present invention can be implemented as computer-readable code on a computer-readable recording medium. Computer-readable recording media include all types of recording devices that store data that can be read by a computer system. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage devices. Additionally, the computer-readable recording medium may be distributed over a computer system connected to a network so that the computer-readable code may be stored and executed in a distributed manner.

Hereinafter, the present invention will be described in more detail through specific experimental examples. The following experimental examples are merely examples to aid understanding of the present invention, and therefore the scope of the present invention is not limited or limited thereto.

<Collection and evaluation of biometric data>

Biometric data was obtained through a clinical trial at Kangwon National University Hospital (Clinical Trial Committee approval number A-2019-11-007-007). The six experiment participants were normal people aged 24 to 36, and SPO ₂ (PPG) data was measured for more than 30 seconds a total of 10 times. As shown in Figure 5, five measurements were made under normal conditions, but for five measurements, the finger to which the sensor was attached was repeatedly flexed and extended at a rate of 100 cycles per minute in the range of 20° to 30°. and measured. Because abnormal measurements were performed immediately after one normal measurement, no physiological changes, such as changes in the patient's heart rate, were observed between normal and abnormal measurements. Afterwards, the PPG waveform and SPO ₂ value measured from the experiment participant, and the HR (heart rate) measured by the SPO ₂ measuring device (BPM-190, Bionics, Inc., Korea) were transferred to the computer, and the PPG waveform was analyzed. Feature points (S point, O point, W point, Z point, ES, EE) were determined.

PPG data was sampled at 62.5Hz, and 32 PPG data and 32 differential data measured within a 1.2 second range were signal processed to create data for analysis and used as input to a deep neural network. 20 S-point data determined by the DNN filter, 20 O-point data and 20 W-point data were also used as input to the deep neural network. The input of each deep neural network consisted of a total of 124 neurons, two hidden layers with 124 neurons each, and one output layer with 21 neurons. HR was calculated using the interval between the previous and current positions of point W determined by the DNN filter, and compared with the HR measured by the SPO ₂ device in normal and abnormal measurement conditions. The DNN filter judges the S, O, and W points repeatedly up to 20 times for each feature point, and calculates the number by increasing one point when detected once. The maximum number of times the S, O, and W points are judged during a heartbeat. The sum was set to 60. To quantitatively indicate the degree to which the DNN filter recognized the heartbeat, the sum of the number of judgments for S, O, and W points was scored. In addition, when no abnormal noise was detected, that is, when ES and EE were not within the heartbeat range, the score was 40 points and added to the heartbeat recognition degree score, and when it was present, it was calculated as -1 point for one detection. In this experiment, the score for recognizing the heartbeat was expressed as recognition score (RS). Therefore, heart rate recognition is calculated from a minimum score of 0 to a maximum of 100 points, and was calculated using Equation 1 described above.

Figure 6 shows the heart rate measured by a pulse oximeter and the heart rate measured by the photoplethysmography data analysis method according to an embodiment of the present invention. Referring to (a) of Figure 6, it can be seen that the HR value measured by the SPO ₂ device varied significantly during normal and abnormal measurements. In other words, it can be seen that motion noise occurs during abnormal measurements, which reduces the accuracy and reliability of heart rate.

Figure 6(b) shows the HR value derived by analyzing PPG measured simultaneously with the same SPO ₂ device using a DNN filter, and there was no significant change in HR between normal and abnormal measurements. Therefore, it can be seen from the results of Figure 6 that the photoplethysmographic wave data analysis method according to an embodiment of the present invention can stably analyze heart rate without being affected by motion noise.

Figure 7 is a result derived by the photoplethysmography analysis method according to an embodiment of the present invention, showing heartbeat recognition derived during normal and abnormal measurements. In Figure 7, the difference in heartbeat recognition derived from normal measurement data and abnormal measurement data is compared. At this time, since the Z point among the feature points represents the characteristics of the patient and the disease and may not be visible in some patients, it was excluded from the calculation of RS.

Referring to Figure 7, when PPG in a normal measurement state was analyzed using a DNN filter, the recognition rate (RS) for all heartbeats was 98.4 (minimum 90.3) on average, but when PPG in an abnormal measurement state was analyzed, the RS The average was low at 48.4 (maximum 78.8). Therefore, if the case where the RS value of the DNN filter is 80 or more is considered a normal measurement state, the RS value of the analysis algorithm can perfectly distinguish between normal and abnormal measurement states, thereby increasing the reliability of the result data derived by the artificial neural network. You can.

As described above, according to various embodiments of the present invention, the waveform of the PPG signal is automatically analyzed, and the accuracy and reliability of the analysis results are improved as the judgment results for the feature points of the PPG data are accumulated.

In addition, according to various embodiments of the present invention, in the feature point detection step, the same feature point is detected redundantly in two analysis data in which some areas overlap, the decision results for the same feature point are accumulated, and the accumulated decision result is Since DNN filters can be learned using DNN, accuracy increases while reducing the number of neurons and data computation compared to existing artificial neural networks, and it has the advantage of being able to be implemented quickly even on relatively low-performance hardware and enabling real-time processing. In addition, the decision results for the same feature point are accumulated and stored, and the information about the feature point with the accumulated decision results is used again to learn the DNN filter, which has the effect of increasing the accuracy of the decision even if the amount of input data is small.

And, according to various embodiments of the present invention, in the signal processing step, a window with a certain size (e.g., 1.2 seconds) moves by a preset time unit (e.g., 20 milliseconds) to generate analysis data. can do. According to one embodiment, instead of making the entire 2-minute PPG data into data for analysis, only a part of the entire signal (for example, a signal from 0 seconds to 1.2 seconds) is made into data for analysis, but the input position of the window is changed to data for analysis. If you create 600 pieces of analysis data by changing them every 2 milliseconds and overlapping a portion of each piece of data for analysis, the amount of one-time calculation is reduced and learning of the overlapped portion is repeated, which can increase the learning effect of the artificial neural network.

As described above, the specific description of the present invention has been made by way of examples with reference to the accompanying drawings, but since the above-described embodiments are only explained by referring to preferred examples of the present invention, the present invention is limited to the above-described embodiments. It should not be understood as being possible, and the scope of rights of the present invention should be understood in terms of the claims described later and their equivalent concepts.

W: window
I: window size
T: unit of time
S: Point S, the maximum value due to heart contraction
D1: Data for first analysis
D2: Data for second analysis
D3: Data for third analysis

Claims (8)

A photoplethysmography data analysis method implemented by a computer including a memory storing computer-executable instructions and a processor executing the computer-executable instructions, comprising:
A PPG data receiving step of receiving PPG (photoplethysmography) data;
A differential data generation step of differentiating the PPG data to generate differential data;
A signal processing step of generating analysis data by signal processing the PPG data and differential data;
A feature point detection step of detecting feature points present in the analysis data using a pre-learned analysis algorithm; and
An additional learning step of training the analysis algorithm using the feature point data detected by the analysis algorithm,
In the feature point detection step, the analysis algorithm includes a deep neural network (DNN) filter including a plurality of hidden layers between an input layer and an output layer, and the DNN filter is provided in a single or plural number depending on the type of the feature point,
In the feature point detection step, the feature points are S point, which is the maximum value point due to heart contraction among the PPG data, O point, which is the minimum value point just before heart contraction among the PPG data, and error, which is an area where heart rate cannot be measured among the PPG data. It includes point ES, which is the starting point of the region, point EE, which is the end point of the error region, point W, which is the maximum value point among the differential data, and point Z, which is the first peak caused by a reflected wave among the differential data,
In the signal processing step, a window with a certain size moves by a preset time unit and signal processes PPG data and differential data corresponding to the area of the window to generate analysis data, and the time unit in which the window moves is the above. By being set small compared to the size of the window, some areas of continuously generated analysis data overlap each other,
In the feature point detection step, the analysis data is input to the DNN filter and analysis is performed to detect feature points included in the analysis data. However, when analyzing analysis data with some areas overlapping with each other, the same feature points are detected redundantly. and the judgment results for the same feature point are accumulated and stored,
In the additional learning step, the DNN filter is additionally trained using information about the feature points detected by the DNN filter in the PPG data and information about feature points detected by the DNN filter in the differential data as learning data. characterized by
Photoplethysmography data analysis method. delete delete delete delete delete According to paragraph 1,
The photoplethysmographic data analysis method is
A biometric information derivation step of deriving biometric information using the feature point data detected in the feature point detection step; and
It further includes a reliability evaluation step of evaluating the reliability of the measurement value of the measuring device that measures the user's biometric information,
In the reliability evaluation step, the biometric information value derived in the biometric information derivation step is compared with the measurement value of the measuring equipment, and if the difference value is less than a predetermined standard value, the reliability of the measuring equipment is evaluated as high, and the difference value is predetermined. If it exceeds the standard value, the reliability of the measuring equipment is evaluated as low.
Photoplethysmography data analysis method. A computer-readable recording medium on which a program for performing the method according to any one of claims 1 and 7 is recorded.

Images