Blood pressure measurement device, blood pressure measurement method, program

Abstract

This technology provides a way to enable highly accurate blood pressure estimation regardless of the timing of the cessation of pressurization. [Solution] A blood pressure measuring device that estimates blood pressure values ​​using a trained model comprises a cuff for compressing the part to be measured, a sensor for detecting the pressure of the cuff, a model storage unit for storing multiple trained models, a pulse wave acquisition unit for acquiring pulse wave data from the output signal of the sensor, a pressurization progress determination unit for determining the progress of pressurization based on the pulse wave data acquired up to the present while gradually pressurizing the cuff, a model selection unit for selecting a trained model from the multiple trained models that corresponds to the pressurization progress determined by the pressurization progress determination unit, and a blood pressure estimation unit for estimating blood pressure values ​​using the pulse wave data acquired up to the present and the selected trained model.

Description

[Technical Field]

[0001] This invention relates to a blood pressure measuring device that estimates blood pressure values ​​using a trained model. [Background technology]

[0002] In recent years, technologies for managing health by measuring personal physical information such as blood pressure values ​​using measuring devices, recording the measurement results, and analyzing them have become widespread. One example of a measuring device is a blood pressure monitor that is configured to estimate blood pressure values, including minimum blood pressure (diastolic blood pressure: DBP) and maximum blood pressure (systolic blood pressure: SBP), based on the measured blood pressure and pulse wave data. For example, Patent Document 1 proposes a method for estimating DBP and SBP by determining a pulse score from the profile of the pressure and pulse waves of multiple heartbeats measured most recently, and inputting the envelope of the pressure and pulse waves into a neural network when the pulse score exceeds a threshold. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] U.S. Patent No. 9750419 [Overview of the project] [Problems that the invention aims to solve]

[0004] As proposed in Patent Document 1, by stopping the pressurization of the cuff midway and inputting the pulse wave data obtained up to that point into a trained model to estimate the blood pressure value, it becomes possible to complete the measurement with less pressure and in a shorter time than with conventional blood pressure monitors.

[0005] However, this type of method has the challenge of making it difficult to improve or stabilize the accuracy of blood pressure estimation. Since the changes in pulse waves due to pressurization are influenced to some extent by the user's physical characteristics and the circumstances of use, it is difficult to predict when the cuff pressurization will stop during measurement (in other words, up to what point the pulse wave data will be used as input data for the trained model) when generating (training) a pre-trained model. Therefore, it is necessary to prepare variations of pulse wave data with different stopping timings as training data and perform machine learning. The pre-trained model acquired in this way has the versatility to obtain reasonably valid estimation results regardless of when the pressurization is stopped, but as a trade-off, there is a limit to the accuracy of blood pressure estimation. On the other hand, for example, if only pulse wave data where pressurization is stopped near the peak of the envelope is used as training data, highly accurate blood pressure estimation is possible under the same conditions. However, if pressurization is stopped at any other timing, the accuracy will decrease, and as a result, the stability of the overall blood pressure estimation accuracy will be impaired.

[0006] The present invention aims to provide a technology that enables highly accurate blood pressure estimation regardless of the timing of pressurization cessation. [Means for solving the problem]

[0007] This disclosure relates to a blood pressure measuring device that estimates blood pressure values ​​using a trained model, comprising: a cuff for compressing the part to be measured; a sensor for detecting the pressure of the cuff; a model storage unit for storing multiple trained models; a pulse wave acquisition unit for acquiring pulse wave data from the output signal of the sensor; a pressurization progress determination unit for determining the progress of pressurization based on the pulse wave data acquired up to the present while gradually pressurizing the cuff; and a model selection unit for selecting a trained model from the multiple trained models that corresponds to the pressurization progress determined by the pressurization progress determination unit. The blood pressure measuring device includes a blood pressure estimation unit that estimates a blood pressure value using pulse wave data acquired up to the present time and the selected trained model.

[0008] Thus, by adopting an architecture in which multiple pre-trained models are prepared in advance and the pre-trained model corresponding to the progress of pressurization (the degree of pressurization) is selected to estimate the blood pressure value, it becomes easy to train or tune each pre-trained model to match the corresponding pressurization progress. Therefore, highly accurate blood pressure estimation is possible regardless of when pressurization is stopped.

[0009] The process of pressurizing the cuff may be divided into multiple levels, and the multiple trained models may be machine-trained to correspond to each of these multiple levels. The "pressurization process" refers to the entire period from the start of pressurization (lower limit of cuff pressure) to the end of pressurization (upper limit of cuff pressure) when pressurization is performed gradually without stopping midway. By dividing the pressurization process into multiple levels and associating each level (each division) with a trained model, the correspondence between the progress of pressurization and the trained model becomes clear, making model selection easier. Another advantage is that it becomes clear which period of pulse wave data should be used as training data when machine-training the model.

[0010] The pre-trained model is designed to output data that directly or indirectly represents blood pressure values ​​when given pulse wave features as input data. If one or more feature points appear in the change curve representing the change in the features during the process of pressurizing the cuff, the multiple levels may be divided at the position of the feature points in the change curve. A "feature point" is, for example, a point where the correlation between the features and cuff pressure changes discontinuously, and may also be called a singularity. When the change curve is divided at the position of the feature point, the feature points are not included in each divided change curve, making it easier to formulate (model) the correlation between the features and cuff pressure, and consequently the correlation between the features and blood pressure values, within each level. Therefore, it becomes easier to improve the accuracy of the pre-trained model corresponding to each level. For example, the feature points may include inflection points and / or discontinuities in the change curve. Also, the feature points may appear as a result of changes in the physical properties of the blood vessels due to the compression of the cuff.

[0011] The feature quantity may be the amplitude of the pulse wave, and the change curve may be the envelope of the pulse wave amplitude. In this case, the feature points may include points on the envelope of the pulse wave amplitude corresponding to SBP (systolic blood pressure), MBP (mean blood pressure), and DBP (diastolic blood pressure), respectively. SBP, MBP, and DBP are inflection points on the envelope of the pulse wave amplitude.

[0012] The process of pressurizing the cuff is divided into four levels: Level 1 (below DBP), Level 2 (DBP to MBP), Level 3 (MBP to SBP), and Level 4 (above SBP). The pressurization progress determination unit may determine which of Levels 1 to 4 the pressurization progress is at based on the pulse wave data acquired up to that point. When divided into these four levels, it is expected that highly accurate blood pressure estimation can be achieved at any level.

[0013] The system further includes a second blood pressure estimation unit that estimates a second blood pressure value from the pulse wave data acquired up to the present time using a method different from that of the trained model, and the blood pressure estimation unit may estimate the blood pressure value using the pulse wave data acquired up to the present time, the selected trained model, and the second blood pressure value estimated by the second blood pressure estimation unit. This can further improve the accuracy of blood pressure estimation.

[0014] The second blood pressure estimation unit may determine at least one of SBP, MBP, or DBP as the second blood pressure value based on features extracted from the pulse wave data acquired up to that point. For example, the second blood pressure estimation unit may use a conventional algorithm that does not use a trained model. A second blood pressure value may be estimated using the sm. Conventional algorithms such as the Fixed Ratio (FR) method and the Derivative method may be used.

[0015] The blood pressure estimation unit may calculate a final blood pressure value from the second blood pressure value obtained by the second blood pressure estimation unit and the first blood pressure value estimated by the trained model.

[0016] The blood pressure estimation unit may estimate the blood pressure value by providing the second blood pressure value obtained by the second blood pressure estimation unit as one of the input data to the trained model.

[0017] The system further includes a pressure control unit that stops pressurizing the cuff when the pressurization progress determined by the pressurization progress determination unit reaches a predetermined target value. After the pressurization of the cuff is stopped by the pressure control unit, the model selection unit selects a trained model corresponding to the predetermined target value, and the blood pressure estimation unit estimates the blood pressure value. The "predetermined target value" may be predetermined, or it may be specified or changed by the user.

[0018] The system may further include a pressure control unit that stops pressurizing the cuff when the pressure of the cuff reaches a predetermined pressure, and after the pressure control unit stops pressurizing the cuff, the pressurization progress determination unit determines the progress of pressurization, the model selection unit selects a trained model corresponding to the progress of pressurization, and the blood pressure estimation unit estimates the blood pressure value.

[0019] This disclosure includes a blood pressure measurement method for estimating blood pressure values ​​using a trained model, comprising the steps of: acquiring an output signal from a sensor that detects the pressure of a cuff compressing the part to be measured; acquiring pulse wave data from the output signal of the sensor; determining the progress of pressurization based on the pulse wave data acquired up to the present while gradually pressurizing the cuff; selecting a trained model from among a plurality of trained models pre-stored in a model storage unit that corresponds to the determined progress of pressurization; and estimating a blood pressure value using the pulse wave data acquired up to the present and the selected trained model.

[0020] The present disclosure includes a program for causing a processor of a blood pressure measurement device that estimates a blood pressure value using a trained model to execute the steps of: acquiring an output signal from a sensor that detects a pressure of a cuff compressing a measured part; acquiring pulse wave data from the output signal of the sensor; determining progress of pressurization based on pulse wave data acquired up to a current point in time while gradually pressurizing the cuff; selecting a trained model corresponding to the determined progress of pressurization from among a plurality of trained models stored in advance in a model storage unit; and estimating a blood pressure value using the pulse wave data acquired up to the current point in time and the selected trained model.

[0021] The present invention may also be interpreted as a blood pressure measurement device, a sphygmomanometer, a biological information acquisition device, or the like having at least a part of the above configuration. Further, the present invention can also be interpreted as a blood pressure measurement method, a blood pressure estimation method, a pressurization control method including at least a part of the above processing, or a program for implementing such a method and a recording medium storing the program. Each of the above means and processes can be combined with each other as much as possible to constitute the present invention.

Effects of the Invention

[0022] According to the present invention, highly accurate blood pressure estimation is possible regardless of the pressurization stop timing.

Brief Description of Drawings

[0023] [Figure 1] Figure 1 is a diagram illustrating the basic configuration of a blood pressure measurement device according to an embodiment of the present invention. [Figure 2] Figure 2A is a diagram showing the feature amount RAV, Figure 2B is a diagram showing the feature amount WID, and Figure 2C is a diagram showing the feature amount DFN. [Figure 3] Figure 3 is a diagram schematically showing an example of the hardware configuration of the blood pressure measurement device. [Figure 4] Figure 4 is a block diagram showing an example of the functional configuration of the blood pressure measurement device. [Figure 5] Figure 5 is a diagram showing an example of the correspondence between pressurization level classifications and models. [Figure 6] Figure 6 is a flowchart showing the overall processing steps of the blood pressure measurement method in the first embodiment. [Figure 7] Figure 7 is a schematic diagram showing the process from steps S603 to S604 in Figure 6. [Figure 8] Figure 8 is a schematic diagram illustrating the training process of the machine learning model used for blood pressure estimation. [Figure 9] Figures 9A and 9B illustrate design example 1 of the pressure level index P. [Figure 10] Figure 10 illustrates design example 2 of the pressure level index P. [Figure 11] Figure 11 is a flowchart showing the overall processing steps of the blood pressure measurement method in the second embodiment. [Figure 12] Figure 12 is a block diagram showing an example of the functional configuration of a blood pressure measuring device in the third embodiment. [Figure 13] Figure 13 is a schematic diagram illustrating the blood pressure estimation process in the third embodiment. [Modes for carrying out the invention]

[0024] <Features of blood pressure measuring devices> Referring to Figure 1, the basic configuration of a blood pressure measuring device according to an embodiment of the present invention will be described. The blood pressure measuring device 1 is a device that estimates blood pressure values ​​using pulse wave data measured by a sensor and a trained model.

[0025] A cuff is attached to the body part being measured on the person being measured, and the cuff pressure is monitored by a sensor while the cuff is gradually pressurized. The top graph in Figure 1 shows the sensor output signal. The horizontal axis is time, and the vertical axis is cuff pressure. It is an example of the sensor output signal when the cuff pressure is linearly increased from the lower limit CPmin to the upper limit CPmax. The pressure pulse wave signal at the body part being measured is superimposed on the sensor output signal. The second graph from the top in Figure 1 shows the pressure pulse wave signal (also called pulse wave data) extracted (separated) from the sensor output signal, with cuff pressure on the horizontal axis. For example, a high-pass filter is used to extract the pressure pulse wave signal. It can be seen that as the cuff pressure is increased, the amplitude of the pulse wave gradually increases, reaching a maximum value around the midpoint between the lowest blood pressure (diastolic blood pressure: DBP) and the highest blood pressure (systolic blood pressure: SBP), and then gradually decreasing thereafter. It is known that the blood pressure value at which the pulse wave amplitude reaches its maximum value is MBP (mean arterial pressure).

[0026] The third graph from the top in Figure 1 shows the change in features extracted from pulse wave data with respect to cuff pressure. The features can be arbitrarily selected as long as they are indicators used to estimate blood pressure values ​​(details will be discussed later). In Figure 1, looking at the change in features during the cuff pressurization process (the period from the lower limit cuff pressure CPmin to the upper limit cuff pressure CPmax), initially, the features increase monotonically with increasing cuff pressure, but when the cuff pressure exceeds CP1, the features begin to decrease monotonically, and then when the cuff pressure exceeds CP2, the features begin to increase monotonically again. In other words, the correlation between features and cuff pressure changes discontinuously at feature points FP1 and FP2, which correspond to cuff pressures CP1 and CP2, respectively. Since these feature points FP1 and FP2 appear solely due to changes in the physical properties of the blood vessels caused by cuff compression, they can appear similarly regardless of who the subject is (as long as the measured area is the same). In the example in Figure 1, feature points FP1 and FP2 appear at positions corresponding to DBP and SBP, but this is just one example, and feature points may also appear at positions other than DBP and SBP (depending on the type of feature).

[0027] If the correlation between feature variables and cuff pressure changes discontinuously during the cuff pressurization process, formulating (modeling) such a correlation between feature variables and blood pressure values ​​is not easy. Because a model capable of solving complex problems is required, not only is model design, training, and optimization difficult, but improving estimation accuracy becomes extremely challenging.

[0028] Therefore, in this method, when one or more feature points appear in the curve of change of features during the cuff pressurization process, the cuff pressurization process is divided into multiple levels (also called pressurization levels) at the location of each feature point, and multiple trained models are prepared corresponding to each of the multiple levels. In the example in Figure 1, the pressurization process is divided into three pressurization levels, from level 1 to level 3, at two locations: feature point FP1 (cuff pressure CP1) and feature point FP2 (cuff pressure CP2). Then, three models, Models 1 to 3, corresponding to levels 1 to 3, are created and stored in the model memory. Model 1 is a trained model that has been machine-learned using pulse wave data obtained during the period of pressurization level 1 (the interval where the cuff pressure is CPmin to CP1) as training data, and it is a model that expresses the correlation between features and blood pressure values ​​during the period of pressurization level 1. Model 2 is a pre-trained model developed using machine learning with pulse wave data obtained during the pressurization level 2 period (cuff pressure between CP1 and CP2) as training data, and represents the correlation between features and blood pressure values ​​during the pressurization level 2 period. Model 3 is a pre-trained model developed using machine learning with pulse wave data obtained during the pressurization level 3 period (cuff pressure between CP2 and CPmax) as training data, and represents the correlation between features and blood pressure values ​​during the pressurization level 3 period. By preparing a dedicated pre-trained model for each pressurization level, which is divided by feature points in this way, the problems that each model needs to solve are simplified, making it easier to achieve high accuracy.

[0029] (Estimation of blood pressure) The blood pressure estimation process is as follows. As an example, we will explain the blood pressure measurement method using the blood pressure measuring device 1, assuming a case where blood pressure is estimated after time tx has elapsed since the start of cuff pressurization. The blood pressure measuring device 1 determines the progress of pressurization based on the pulse wave data acquired up to time tx (pulse wave data acquired from the start of pressurization until the cuff pressure reaches CPtx). In this example, the progress of pressurization is determined to be level 2. Next, the blood pressure measuring device 1 selects model 2, which is a trained model corresponding to level 2, from the model storage unit, and estimates the blood pressure value by performing blood pressure estimation using the pulse wave data acquired up to time tx and model 2. With this method, highly accurate blood pressure estimation is possible based on the pulse wave data up to time tx.

[0030] (Features) The features can be arbitrarily selected as long as they are indicators used to estimate blood pressure values. For example, the features may include features obtained from the envelope of the pulse wave amplitude. Alternatively, the features may include features obtained from the pulse wave (pulse) for each heartbeat. Examples of envelope features include the height of the envelope, the maximum amplitude of the envelope, the full width at half maximum of the envelope, the peak cuff pressure of the envelope (the cuff pressure at which the envelope amplitude is maximum), the area of ​​the envelope, the standard deviation of the envelope amplitude, the skewness of the envelope, the kurtosis of the envelope, and the parameters obtained when a predetermined model (curve) is fitted to the envelope (for example, the parameters of the Gaussian function fitted to the envelope). Examples of pulse wave (pulse) features may include statistics such as the maximum value, minimum value, variance, standard deviation, nth percentile value, skewness, and suboptimality of the pulse wave (pulse) for multiple heartbeats. Alternatively, features obtained from the pulse wave (pulse) for the most recent heartbeat may be used. As feature quantities for a single heartbeat pulse wave, the maximum amplitude of the pulse, the full width at half maximum of the pulse, the peak cuff pressure of the pulse, the pulse area, the standard deviation of the pulse amplitude, the pulse skewness, and the pulse kurtosis can be used. Alternatively, feature quantities such as those shown in Figures 2A and 2C (RAV, WID, DFN) may be used. RAV is calculated by normalizing the pulse area for a single heartbeat, shown by the shaded area in Figure 2A, by the amplitude. For example, RAV is (pulse area / pulse amplitude) × It may be calculated using 100. WID is the peak (Maximum) as shown in Figure 2B. The waveform width is defined as the time interval it takes for the amplitude to decrease from a certain point to a threshold. These are calculated by normalizing by the pulse wave period. For example, WID can be calculated as (wavelength / pulse wave period) × 100. DFN is calculated by normalizing the minimum value of the first derivative of the pulse wave by the amplitude, as shown in Figure 2C. For example, DFN can be calculated as (minimum value of the first derivative of the pulse wave from 0 / amplitude of the first derivative of the pulse wave) × 100. RAV, WID, and DFN represent the area, width, and slope of the pulse wave for one heartbeat. Note that the features listed here are just examples, and other features may be used.

[0031] (Model) Each model is composed of a machine learning model. The machine learning model is configured to have one or more computational parameters that can be adjusted by machine learning. One or more computational parameters are used for the calculation of the desired inference (in this embodiment, estimation of blood pressure values). The machine learning model may be composed of, for example, a neural network, a regression model, a decision tree model, a support vector machine, or other functional equations (computational models). The machine learning method may be appropriately selected depending on the machine learning model adopted (for example, backpropagation). Note that the structures of Models 1 to 3 may be the same (only the training data differs) or they may be different (both the structure and the training data differ).

[0032] In one example, each model may include a neural network. The structure of the neural network is not particularly limited and may be determined as appropriate depending on the embodiment. The structure of the neural network may be specified, for example, by the number of layers from the input layer to the output layer, the type of each layer, the number of nodes (neurons) included in each layer, and the connection relationships between the nodes in each layer. In one example, the neural network may include any mechanism such as a recursive structure, a self-attention mechanism, or an autoregressive model. Furthermore, the neural network may include any layer such as a fully connected layer, a convolutional layer, a pooling layer, an inverse convolutional layer, an unpooling layer, a normalization layer, a dropout layer, or an LSTM (Long Short-Term Memory). The neural network may include any type of model such as a diffusion model, a Transformer model, or a generative model. The weights of the connections between each node included in the neural network and the threshold values ​​of each node are examples of computational parameters.

[0033] Machine learning involves adjusting (optimizing) the values ​​of the computational parameters of a machine learning model using each training data set. Typically, a trained machine learning model (pre-trained model) may be generated by supervised learning using multiple training data sets. That is, the values ​​of the computational parameters of the machine learning model may be adjusted so that the output obtained from the machine learning model, given each training data set, fits the true value (blood pressure value in this embodiment) indicated by the corresponding label. For example, in the case of a neural network, the values ​​of the computational parameters of the machine learning model may be adjusted using backpropagation. Note that the training method of the machine learning model is not limited to this example and may be appropriately modified depending on the embodiment, as long as it is possible to acquire the ability to estimate blood pressure values. Furthermore, the machine learning model may be trained online and offline, or at least one of the other. Tuning such as transfer learning, retraining, and additional learning may be performed on the machine learning model as appropriate.

[0034] The input and output formats of each model may be determined as appropriate depending on the embodiment. In one example, the input data may be provided to the model directly or after preprocessing. In another example, the output of the model may be configured to directly or indirectly show the estimated blood pressure. If the output of the model is configured to indirectly show the estimated blood pressure, the estimated blood pressure may be obtained by performing arbitrary information processing (such as interpretation processing) on ​​the output of the model.

[0035] <First Embodiment> (Configuration of a blood pressure measurement device) Figure 3 schematically shows an example of the hardware configuration of the blood pressure measuring device 1 according to this embodiment. The blood pressure measuring device 1 according to this embodiment includes a control unit 10, a storage unit 11, an oscillation circuit 121, a pump drive circuit 122, a valve drive circuit 123, a cuff 130, a pressure sensor 131, a pressurizing pump 132, an exhaust valve 133, an air tube 134, an operation switch 14, a display unit 15, and a power supply 16. The control unit 10 and the storage unit 11 are examples of the computer portion of the blood pressure measuring device 1.

[0036] The cuff 130 includes an air bag 1301 that contains air. The cuff 130 is equipped with a pressure sensor 131, a pressurizing pump 132, and an exhaust valve 133 via an air tube 134. The pressure sensor 131 is configured to detect the pressure (cuff pressure) inside the air bag 1301 of the cuff 130. The pressurizing pump 132 is configured to supply air into the air bag 1301. The exhaust valve 133 is provided as a boundary between the internal and external spaces of the air bag 1301 and is configured to be openable and closable. By closing the exhaust valve 133, air can be trapped inside the air bag 1301 and the pressure inside the air bag 1301 can be maintained. On the other hand, by opening the exhaust valve 133, air can be released from the air bag 1301 and the pressure can be reduced.

[0037] The control unit 10 includes a hardware processor such as a CPU and is configured to perform information processing based on a program and various data. The control unit 10 (CPU) is an example of the processor resources of the estimation device. The oscillation circuit 121 outputs a signal with an oscillation frequency corresponding to the output value of the pressure sensor 131 to the control unit 10. In this embodiment, the control unit 10 processes the signal from the oscillation circuit 121 to obtain cuff pressure and pulse wave data. The pump drive circuit 122 is configured to control the drive of the pressurizing pump 132 based on a control signal output from the control unit 10. The valve drive circuit 123 is configured to control the opening and closing of the exhaust valve 133 based on a control signal output from the control unit 10.

[0038] The memory unit 11 may be composed of, for example, a semiconductor memory. The memory unit 11 is an example of a memory resource of the estimation device. In this embodiment, the memory unit 11 stores various information such as the program 110 and model data 111. The program 110 includes a program that causes the CPU of the blood pressure measuring device 1 to execute information processing (described later), such as cuff pressurization control processing, pressure pulse wave acquisition processing, pressurization progress (pressurization level) determination processing, blood pressure value estimation processing, and measurement result display processing. The program 110 includes a series of instructions for said information processing. Note that each processing may be composed of a different program module, or all processing may be executed with a single program. The model data 111 is a trained model. The configuration of the model data 111 is not particularly limited and may be determined appropriately according to the embodiment, as long as the trained model can be used when determining the pressurization level or estimating blood pressure. The model data 111 may be incorporated into the program 110. The memory unit 11 may appropriately store other information such as the results of blood pressure measurement (cuff pressure, pulse wave data, pressurization level value, estimated blood pressure, etc.).

[0039] The operation switch 14 is used to perform operations such as starting blood pressure measurement. The operation switch 14 may consist of at least one of a physical switch and a virtual switch. The display unit 15 is configured to display various information such as the blood pressure measurement results. The operation switch 14 and the display unit 15 may be integrally configured as a touch panel display. The power supply 16 is configured to supply power to each part, such as the control unit 10.

[0040] Regarding the specific hardware configuration of the blood pressure measuring device 1, components can be omitted, replaced, and added as appropriate depending on the embodiment. For example, the control unit 10 may include multiple hardware processors. The hardware processors may consist of microprocessors, FPGAs, DSPs, ASICs, etc. The blood pressure measuring device 1 may be equipped with a communication interface. For example, the blood pressure measuring device 1 may be equipped with a proximity wireless communication module. The blood pressure measuring device 1 may be configured to communicate with a user terminal such as a smartphone. This allows the blood pressure measuring device 1 to acquire arbitrary data via the user terminal. At least one of the program 110 and the model data 111 may be stored on a storage medium of an external computer, such as a NAS. The blood pressure measuring device 1 may acquire at least one of the program 110 and the model data 111 from the external computer. The blood pressure measuring device 1 may acquire data directly from the external computer, or it may acquire data indirectly from the external computer via the user terminal. The blood pressure measuring device 1 may be configured as a general blood pressure monitor, or as a wearable device such as a wristwatch.

[0041] Figure 4 is a block diagram showing an example of the functional configuration of the blood pressure measuring device 1 according to this embodiment.

[0042] The blood pressure measuring device 1 has the following main functions related to blood pressure measurement: a pressure detection unit 40, a feature extraction unit 41, a pressurization progress determination unit 42, a model selection unit 43, a pressurization control unit 44, a blood pressure estimation unit 45, and a model storage unit 46. These functions are realized by the CPU of the control unit 10 reading and executing a program stored in the storage unit 11. The storage area of ​​the model storage unit 46 is also provided within the storage unit 11.

[0043] The pressure detection unit 40 receives the signal from the oscillation circuit 121. The pressure detection unit 40 detects the oscillation frequency of the signal input from the oscillation circuit 121 and converts the detected oscillation frequency into a pressure value signal. The pressure detection unit 40 includes an HPF (High Pass Filter) unit 401 that extracts and outputs a pressure pulse wave signal by processing the pressure value signal with an HPF (High Pass Filter), and an LPF (Low Pass Filter) unit 402 that extracts and outputs a cuff pressure signal by processing the pressure value signal with an LPF (Low Pass Filter). The time-series data of the pressure pulse wave output from the PF unit 401 and the time-series data of the cuff pressure output from the LPF unit 402 are stored in the memory of the control unit 10. That is, the HPF unit 401 functions as a pulse wave acquisition unit that acquires pulse wave data from the output signal of the pressure sensor 131, and the LPF unit 402 functions as a cuff pressure acquisition unit that acquires cuff pressure data from the output signal of the pressure sensor 131. The sampling rate for the pulse wave and cuff pressure is, for example, about 250 Hz.

[0044] The feature extraction unit 41 has the function of extracting features from the pulse wave data acquired up to the present while gradually pressurizing the cuff 130. The feature extraction unit 41 preferably extracts features from the pulse wave data that will be used in the subsequent pressurization progress determination process and the blood pressure estimation process. The extracted features are not limited to one type, and multiple types of features may be extracted. The features for the pressurization progress determination process and the features for the blood pressure estimation process may include the same type of features, or they may be completely different. In this embodiment, for example, the features for the pressurization progress determination process are extracted from the envelope of the pulse wave (temporal change in amplitude) and the pulse wave (pulse) of the most recent beat, and the features for the blood pressure estimation process are extracted from the envelope of the pulse wave.

[0045] The pressurization progress determination unit 42 has the function of determining the progress of pressurization based on the pulse wave data acquired up to the present while gradually pressurizing the cuff 130. In this embodiment, as shown in Figure 5, the pressurization process of the cuff 130 is divided into four levels, from Level 1 to Level 4. Specifically, the envelope 50 of the pulse wave amplitude is used as the curve of change of feature quantities, and three inflection points 51, 52, and 53 that appear on this envelope 50 are considered as feature points. The first inflection point 51 corresponds to the lowest blood pressure (DBP), the second inflection point 52 corresponds to the peak (MBP) of the envelope 50, and the third inflection point 53 corresponds to the highest blood pressure (SBP). With the positions of these inflection points (feature points) 51, 52, and 53 as boundaries, the entire pressurization process is divided into four levels: "Level 1, lower than DBP", "Level 2, from DBP to MBP", "Level 3, from MBP to SBP", and "Level 4, above SBP". The pressurization progress determination unit 42 determines the pressure based on the envelope of the pulse wave extracted by the feature extraction unit 41 and the features extracted from the pulse wave of the most recent beat. Next, it is determined whether the current cuff pressure is at level 1 through 4. Details of the pressurization progress determination process will be described later.

[0046] The model storage unit 46 has the function of storing multiple trained models. As shown in Figures 4 and 5, in this embodiment, four models M1 to M4 are provided, corresponding to levels 1 to 4, respectively.

[0047] The model selection unit 43 has the function of selecting a model from among multiple trained models M1 to M4 stored in the model storage unit 46 that corresponds to the progress of pressurization (pressurization level) determined by the pressurization progress determination unit 42.

[0048] The pressurization control unit 44 has the function of pressurizing and depressurizing the cuff 130 by sending control signals to the pump drive circuit 122 and the valve drive circuit 123 and controlling the operation of the pressurizing pump 132 and the exhaust valve 133.

[0049] The blood pressure estimation unit 45 has the function of estimating blood pressure values ​​using pulse wave data acquired up to the present time and a trained model selected by the model selection unit 43. The blood pressure values ​​to be estimated should include systolic blood pressure (SBP) and diastolic blood pressure (DBP).

[0050] In this embodiment, the functions shown in Figure 4 are implemented by a software program, but all or part of these functions may be replaced by circuits such as ASICs or FPGAs. Alternatively, the functions shown in Figure 4 may be implemented through collaboration with other computers connected via a network (such as a cloud server, a user's PC, or a smartphone).

[0051] (Operation of the blood pressure measuring device) Figure 6 is a flowchart showing the overall processing steps of the blood pressure measurement method using the blood pressure measurement device 1. The processing shown in Figure 6 is achieved by the CPU of the control unit 10 reading and executing the program 110 from the memory unit 11. Note that all or part of the program 110 may be read into the blood pressure measurement device 1 from a computer-readable external storage medium (such as flash memory), or it may be acquired via a network.

[0052] When measuring blood pressure, the person being measured has already wrapped the cuff 130 around the part to be measured. The following explanation describes an example where the upper arm is the part to be measured, but the part to be measured is not limited to the upper arm; it may also be the wrist or other parts. Furthermore, it will be explained assuming that the person being measured has made the necessary settings using the operation switch 14 and has given the instruction to start blood pressure measurement. When the blood pressure measuring device 1 receives the instruction to start blood pressure measurement, it performs the necessary initialization, such as opening the exhaust valve 133 and setting the cuff pressure to atmospheric pressure (initial pressure).

[0053] When blood pressure measurement is started, the pressurization control unit 44 starts pressurization control to pressurize the cuff 130 (step S600). During pressurization control, the cuff pressure is increased at a constant rate.

[0054] The pressure detection unit 40 acquires the pulse wave and cuff pressure while gradually increasing the cuff pressure (step S601). The pressure detection unit 40 monitors the cuff pressure and waits until the cuff pressure reaches a predetermined lower limit (step S602). When the cuff pressure reaches the lower limit and the part to be measured is compressed with a certain force by the cuff 130, a reliable pulse wave can be measured, and the process proceeds to step S603.

[0055] In step S603, the feature extraction unit 41 extracts features from the pulse wave data of multiple heartbeats acquired up to that point, which will be used in the subsequent pressurization progress determination process (step S604) and blood pressure value estimation process (step S609). For example, the feature extraction unit 41 extracts pulse wave data of multiple heartbeats. Based on the wave data, the envelope of the pulse wave up to the present time is obtained. A known method can be used to generate the envelope. For example, the envelope may be obtained by simply connecting the peak positions of the pulses of each heartbeat, or by fitting a curve to the sequence of points of the peak positions of the pulses of each heartbeat. Although not shown in the figure, the feature extraction unit 41 may perform preprocessing such as noise reduction on the pulse wave data before obtaining the envelope. Subsequently, the feature extraction unit 41 calculates the features of the envelope and the features of the pulse wave.

[0056] In step S604, the pressurization progress determination unit 42 determines the pressurization progress (pressurization level) of the cuff 130 based on the envelope features up to the present time and the features of the most recent heartbeat pulse. Figure 7 schematically shows the processing in steps S603 to S604. A trained model, which has been trained to output a pressurization level value when given the envelope features and the features of one heartbeat pulse as input, is used for the pressurization progress determination process. Any machine learning algorithm can be used, such as multiple regression, random forest, neural network, support vector regression, Lasso regression, Ridge regression, or naive Bayes (Gaussian distribution).

[0057] In step S605, the pressurization control unit 44 compares the pressurization level determined in step S604 with a predetermined target value. The target value may be preset in the storage unit 11 of the blood pressure measuring device 1, or it may be changeable by the user.

[0058] If the pressurization level has reached the target value (Y in step S605), the pressurization control unit 44 stops pressurizing the cuff 130 at that point (step S606). Even if the pressurization level has not reached the target value (N in step S605), if the cuff pressure has reached a predetermined upper limit (Y in step S607), it is assumed that some kind of abnormality may have occurred, and pressurization of the cuff 130 is stopped. As long as the cuff pressure is below the upper limit (N in step S607), the pressurization progress determination process is repeatedly executed at predetermined intervals until the pressurization level reaches the target value.

[0059] In step S608, the model selection unit 43 selects a model from among the four learned models M1 to M4 stored in the model storage unit 46 that corresponds to the pressure level determined in step S604.

[0060] In step S609, the blood pressure estimation unit 45 estimates the blood pressure values ​​(SBP, DBP) using the features obtained from the pulse wave data acquired up to the point when the pressurization of the cuff 130 is stopped, and the model selected in step S608. The measurement results obtained from the above measurement process (blood pressure values, heart rate, pressurization level, etc.) are displayed on the display unit 15 (step S610).

[0061] (Training of models M1-M4) Figure 8 is a schematic diagram illustrating the training process of machine learning models M1 to M4 used in blood pressure estimation. This training process can be performed by the control unit 10 of the blood pressure measurement device 1, but typically it is performed by a separate training device. The training device consists of a general-purpose computer equipped with a CPU, GPU, memory, storage, etc., and the training process described below is realized by loading the training program stored in storage into memory and executing the program using the CPU and GPU.

[0062] First, the learning device acquires training data, specifically pressure pulse wave data and blood pressure values ​​(true values ​​as training information) (step S80). At this time, it is advisable to prepare a large amount of pressure pulse wave data and blood pressure values ​​obtained from a large number of subjects.

[0063] The learning device first selects one pulse wave data point as the target, and then processes the target pulse wave data using an arbitrary cuff. The cuff is cut off by pressure (step S81). The cuff pressure at this point will be referred to as the "cut cuff pressure" below. This cutting process simulates the state in which measurements have been taken up to the point of the cut cuff pressure.

[0064] The learning device extracts envelope features from the pressure pulse wave data extracted in step S81 (step S82). These features are then input into the machine learning model being trained to obtain an estimated blood pressure value (step S83).

[0065] The learning device calculates the difference between the estimated blood pressure obtained in step S83 and the true value obtained as training information (step S84), and uses that difference (error) to modify the machine learning model (step S85).

[0066] In training model M1 corresponding to pressurization level 1, steps S81 to S85 are repeated while varying the cut cuff pressure within a range lower than the diastolic blood pressure (DBP). Furthermore, the same process is repeated for each of the numerous pressure pulse wave data provided as training data. This allows the model to learn the correlation between pulse wave data at pressurization level 1 and blood pressure values.

[0067] To train model M2, which corresponds to pressurization level 2, the process of steps S81 to S85 should be repeated while varying the cut cuff pressure within the range of DBP to MBP. To train model M3, which corresponds to pressurization level 3, the process of steps S81 to S85 should be repeated while varying the cut cuff pressure within the range of MBP to SBP. To train model M4, which corresponds to pressurization level 4, the process of steps S81 to S85 should be repeated while varying the cut cuff pressure within the range of MBP or higher. By using the method described above, a trained model with high estimation accuracy can be obtained for each pressurization level.

[0068] (Example of a design for a pressure level index) Referring to Figures 9A and 9B, we show Design Example 1 of an index P representing the pressurization level (referred to as the "pressurization level index"). In Design Example 1, the envelope is modeled using a Gaussian function, and the relationship between the pressurization level index P and the cuff pressure is defined using the parameters of the Gaussian function.

[0069] First, prepare the pressure pulse wave data. As shown in Figure 9A, use the pressure pulse wave data for the range from cuff pressure less than the diastolic blood pressure (DBP) to cuff pressure exceeding the systolic blood pressure (SBP). Next, obtain the envelope from the pressure pulse wave. For example, you can obtain the envelope by connecting the peak points of the pulses for each heartbeat. Then, fit the envelope with a Gaussian function represented by equation (1) below. In equation (1), the variable x is the cuff pressure, and A, B, and C are parameters that define the shape of the Gaussian function. The pressurization level index P is defined as shown in equation (2) below, using the parameters B and C of the Gaussian function obtained by fitting. Equation (2) defines the relationship (conversion) between index P and cuff pressure x.

[0070]

number

[0071] Figure 9B shows the relationship between the value of index P according to equation (2) and the amplitude of the envelope. Index P is 0 at the peak position where the amplitude of the envelope is at its maximum value, and the absolute value of index P is (2log2) at the position where the amplitude of the envelope is half of the maximum value. 1 / 2 Take the following: Here, log is the natural logarithm.

[0072] Statistically, it is known that the position where the envelope amplitude is half of its maximum value roughly corresponds to DBP and SBP. Therefore, when using the index P in design example 1, P = -(2log2) 1 / 2 The position of the arrow roughly indicates the position of DBP, and P = (2log2) 1 / 2 The position of will roughly indicate the position of SBP. For example, if the trained model shown in Figure 7 is designed to output the pressurization level index P of design example 1, the pressurization progress determination unit 42 should determine the pressurization level as follows, according to the value of index P estimated by the trained model.

[0073] P <- (2log2) 1 / 2 In this case → Pressurization level 1 -(2log2) 1 / 2When ≦P<0 → Pressurization level 2 0≦P<(2log2) 1 / 2 When → Pressurization level 3 (2log2) 1 / 2 When ≦P → Pressurization level 4

[0074] Next, as another example, Design Example 2 of the pressurization level index P will be described. In Design Example 2, the relational expression between the index P and the cuff pressure is defined by normalizing the envelope curve in the interval between the diastolic blood pressure and the systolic blood pressure.

[0075] Specifically, the true value of systolic blood pressure SBP gt and the true value of diastolic blood pressure DBP gt may be used to define the relationship (conversion) between the index P and the cuff pressure x as shown in the following formula (3). In the case of Design Example 2, since it is not necessary to calculate the envelope curve or fit a mathematical formula to the envelope curve as required in Design Example 1, the relational expression for the index P can be obtained through simple processing.

[0076]

Math

[0077] FIG. 10 shows the relationship between the index P and the cuff pressure x according to formula (3). That is, the index P takes a value of -1 at the position corresponding to the diastolic blood pressure DBP on the envelope curve, and takes a value of +1 at the position corresponding to the systolic blood pressure SBP on the envelope curve. Furthermore, by setting that the envelope curve peaks between the diastolic blood pressure and the systolic blood pressure, the index P becomes 0 at the peak position of the envelope curve. For example, if the trained model shown in FIG. 7 is designed to output the pressurization level index P of Design Example 2, the pressurization progress determination unit 42 may determine the pressurization level as follows according to the value of the index P estimated by the trained model.

[0078] When P<-1 → Pressurization level 1 When -1≦P<0 → Pressurization level 2 When 0≦P<1 → Pressurization level 3 When 1≦P → Pressurization level 4

[0079] (Advantages of this embodiment) According to the blood pressure measuring device 1 and blood pressure measuring method of this embodiment described above, the optimal trained model is selected according to the progress of pressurization to estimate the blood pressure value, so highly accurate blood pressure estimation is possible regardless of when pressurization is stopped. Furthermore, by introducing the concepts of pressurization level and pressurization level index P, the pressurization process is divided into multiple levels, and each level 1 to 4 is assigned a one-to-one correspondence to trained models M1 to M4, so the correspondence between the progress of pressurization and the trained model becomes clear, and model selection becomes easy.

[0080] <Second Embodiment> In the first embodiment, the pressurization is stopped and blood pressure estimation is performed when the pressurization level reaches a predetermined target value, whereas in the second embodiment, the pressurization is stopped and blood pressure estimation is performed when the cuff pressure reaches a predetermined pressure. The following description will focus on the configurations and processes specific to the second embodiment.

[0081] Figure 11 is a flowchart showing the overall procedure for blood pressure measurement using blood pressure measurement device 1. The same steps are used for the same processes as in the flowchart in Figure 6.

[0082] When blood pressure measurement is started, the pressurization control unit 44 starts pressurization control to pressurize the cuff 130 (step S600). During pressurization control, the cuff pressure is increased at a constant rate.

[0083] The pressure detection unit 40 acquires the pressure pulse wave and cuff pressure while gradually increasing the cuff pressure (step S601). The pressure detection unit 40 monitors the cuff pressure, and if the cuff pressure reaches a predetermined target value (Y in step S1100), the pressurization control unit 44 stops pressurizing the cuff 130 at that point (step S1101). The target value may be preset in the storage unit 11 of the blood pressure measuring device 1, or it may be changeable by the user.

[0084] In step S603, the feature extraction unit 41 extracts features from the pulse wave data of multiple heartbeats acquired up to that point to be used in the subsequent pressurization progress determination process (step S604) and blood pressure value estimation process (step S609). In step S604, the pressurization progress determination unit 42 determines the pressurization progress (pressurization level) of the cuff 130 based on the envelope features up to that point and the pulse features of the most recent heartbeat.

[0085] In step S608, the model selection unit 43 selects a model from among the four trained models M1 to M4 stored in the model memory unit 46 that corresponds to the pressurization level determined in step S604. In step S609, the blood pressure estimation unit 45 estimates the blood pressure values ​​(SBP, DBP) using the features obtained from the pulse wave data acquired up to the point when pressurization of the cuff 130 is stopped, and the model selected in step S608. The measurement results obtained from the above measurement process (blood pressure values, heart rate, pressurization level, etc.) are displayed in the display unit 15 (step S610).

[0086] According to the method of the second embodiment, since pressurization stops at a predetermined cuff pressure, there is an advantage in that the feeling of pressure on the user and the time required for blood pressure measurement can be kept roughly constant.

[0087] <Third Embodiment> In the third embodiment, a second blood pressure estimation unit is provided that estimates blood pressure values ​​in a different way than the blood pressure estimation unit 45. The second blood pressure value estimated by this second blood pressure estimation unit is used in the blood pressure estimation processing of the blood pressure estimation unit 45 to further improve the accuracy of blood pressure estimation. The following description will focus on the configuration and processing specific to the third embodiment.

[0088] Figure 12 is a block diagram showing an example of the functional configuration of the blood pressure measuring device 1 of the third embodiment. It differs from the configuration of the first embodiment (Figure 4) in that it has a second blood pressure estimation unit 47.

[0089] Figure 13 is a schematic diagram illustrating the blood pressure estimation process in the third embodiment.

[0090] When pulse wave data for multiple heartbeats is acquired, the feature extraction unit 41 extracts a first feature used in the blood pressure estimation process of the blood pressure estimation unit 45 and a second feature used in the blood pressure estimation process of the second blood pressure estimation unit 47. The first and second feature are of the same type. It's fine to use the same type of feature, or different types of features.

[0091] The second blood pressure estimation unit 47 estimates a second blood pressure value using a second feature. The blood pressure estimation algorithm used by the second blood pressure estimation unit 47 can be any algorithm, as long as it is different from that used by the blood pressure estimation unit 45. For example, the Fixed Retio (FR) method or the Derivative method, which are conventional algorithms with a proven track record in oscillometric blood pressure monitors, may be used. The FR method considers the values ​​ΔOd (=Rd×ΔOm) and ΔOs (=Rs×ΔOm), obtained by multiplying the peak of the envelope of the pulse wave amplitude (i.e., the maximum value of the pulse wave amplitude) ΔOm by predetermined ratios Rd and Rs, as the amplitude of the diastolic blood pressure DBP and the amplitude of the systolic blood pressure SBP, respectively, and calculates DBP from the cuff pressure Pd at which the pulse wave amplitude is ΔOd, and SBP from the cuff pressure Ps at which the pulse wave amplitude is ΔOs. The Derivative method calculates DBP from the cuff pressure Pd at which the slope (first derivative) of the pulse wave amplitude envelope is maximized, and SBP from the cuff pressure Ps at which it is minimized. In both the FR method and the Derivative method, MBP can be calculated from the peak position of the pulse wave amplitude envelope.

[0092] The second blood pressure value calculated by the second blood pressure estimation unit 47 is used in the blood pressure estimation process in the blood pressure estimation unit 45. In the case of the FR method, since the envelope peak ΔOm must be used in the calculation, the second blood pressure value cannot be calculated from the pulse wave data at pressurization levels 1 and 2. DBP and MBP can be calculated at pressurization level 3, and DBP, MBP, and SBP can be calculated at pressurization level 4. On the other hand, in the case of the Derivative method, the second blood pressure value cannot be calculated at pressurization level 1, but DBP can be calculated at pressurization level 2, DBP and MBP can be calculated at pressurization level 3, and DBP, MBP, and SBP can be calculated at pressurization level 4. Thus, since the type of second blood pressure value that can be used in the processing of the blood pressure estimation unit 45 changes depending on the pressurization level, the blood pressure estimation unit 45 does not require the second blood pressure value as a mandatory parameter, and only uses it as an auxiliary parameter in the blood pressure estimation process if the second blood pressure value can be obtained.

[0093] The method of using the second blood pressure value by the blood pressure estimation unit 45 is not particularly limited. For example, the blood pressure estimation unit 45 may estimate the first blood pressure value using a trained model, as in the embodiment described above, and then calculate the final blood pressure value by correcting this first blood pressure value with the second blood pressure value. For example, the blood pressure estimation unit 45 may use a simple average or weighted average of the first and second blood pressure values ​​as the final blood pressure value, or it may calculate the final blood pressure value by substituting the first and second blood pressure values ​​into a predetermined function. As another example, the blood pressure estimation unit 45 may provide the second blood pressure value as one of the input data to the trained model to estimate the blood pressure value. In this case, the trained model is designed and trained to estimate a highly accurate blood pressure value using a first feature quantity extracted from the pulse wave data and a second blood pressure value estimated by the second blood pressure estimation unit 47.

[0094] With the configuration and processing of this embodiment described above, further improvements in accuracy can be expected compared to the first embodiment. Note that the configuration of this embodiment may be combined with the configuration of the second embodiment.

[0095] <Other> The above embodiments are merely illustrative examples illustrating the configuration of the present invention. The present invention is not limited to the above specific forms, and various modifications are possible within the scope of its technical idea. For example, in the above embodiments, the blood pressure estimation process was performed after pressurization was stopped, but blood pressure estimation may be performed at predetermined cycles during the process of pressurizing the cuff. Also, in the above embodiments, the pressurization level was divided into four stages, but it may be divided into any number of stages, as long as there are two or more. Furthermore, in the above embodiments, an example of division at positions corresponding to DBP, MBP, and SBP was shown, but the boundary positions of the pressurization levels are not limited to these. Depending on the feature quantity of interest, feature points may appear at positions other than DBP, MBP, and SBP, and in such cases, the boundary positions of the pressurization levels will be set at positions other than DBP, MBP, and SBP. [Explanation of symbols]

[0096] 1: Blood pressure measuring device

Claims

1. In a blood pressure measurement device that estimates blood pressure values ​​using a trained model, A cuff for compressing the area to be measured, A sensor for detecting the pressure of the cuff, A model memory unit that stores multiple trained models, A pulse wave acquisition unit that acquires pulse wave data from the output signal of the aforementioned sensor, A pressurization progress determination unit determines the progress of pressurization based on pulse wave data acquired up to the present while gradually pressurizing the cuff, A model selection unit selects a trained model from among the multiple trained models that corresponds to the progress of pressurization determined by the pressurization progress determination unit, A blood pressure estimation unit that estimates blood pressure values ​​using the pulse wave data acquired up to the present time and the selected trained model, A blood pressure measuring device having the following features.

2. The process of pressurizing the cuff is divided into multiple levels. The aforementioned multiple trained models are machine-learned to correspond to each of the aforementioned multiple levels. The blood pressure measuring device according to claim 1.

3. The aforementioned trained model is designed to output data that directly or indirectly represents blood pressure values ​​when pulse wave features are given as input data. When one or more feature points appear in the change curve representing the change in the feature quantity during the process of pressurizing the cuff, the multiple levels are divided by the position of the feature points in the change curve. The blood pressure measuring device according to claim 2.

4. The aforementioned characteristic points include inflection points and / or discontinuities of the change curve. The blood pressure measuring device according to claim 3.

5. The aforementioned characteristic is that it arises from a change in the physical properties of the blood vessel caused by compression of the cuff. The blood pressure measuring device according to claim 3.

6. The aforementioned feature points include points corresponding to SBP, MBP, and DBP, respectively, on the envelope of the pulse wave amplitude. The blood pressure measuring device according to claim 3.

7. The process of pressurizing the cuff is divided into four levels: Level 1, which is lower than DBP; Level 2, which is between DBP and MBP; Level 3, which is between MBP and SBP; and Level 4, which is above SBP. The aforementioned pressurization progress determination unit determines, based on the pulse wave data acquired up to that point, whether the pressurization progress is at level 1 to 4. The blood pressure measuring device according to claim 6.

8. The system further includes a second blood pressure estimation unit that estimates a second blood pressure value from the pulse wave data acquired up to the present time using a method different from that of the trained model. The blood pressure estimation unit estimates a blood pressure value using the pulse wave data acquired up to the present time, the selected trained model, and the second blood pressure value estimated by the second blood pressure estimation unit. The blood pressure measuring device according to claim 7.

9. The second blood pressure estimation unit determines at least one of SBP, MBP, or DBP as the second blood pressure value based on the feature quantities extracted from the pulse wave data acquired up to that point. The blood pressure measuring device according to claim 8.

10. The blood pressure estimation unit calculates a final blood pressure value from the second blood pressure value obtained by the second blood pressure estimation unit and the first blood pressure value estimated by the trained model. The blood pressure measuring device according to claim 8.

11. The blood pressure estimation unit provides the second blood pressure value obtained by the second blood pressure estimation unit as one of the input data to the trained model and estimates the blood pressure value. The blood pressure measuring device according to claim 8.

12. The system further includes a pressure control unit that stops pressurizing the cuff when the pressurization progress determined by the pressurization progress determination unit reaches a predetermined target value. After the pressurization of the cuff is stopped by the pressurization control unit, the model selection unit selects a trained model corresponding to the predetermined target value, and the blood pressure estimation unit estimates the blood pressure value. The blood pressure measuring device according to claim 1.

13. The system further includes a pressure control unit that stops pressurizing the cuff when the pressure in the cuff reaches a predetermined pressure. After the pressurization of the cuff is stopped by the pressurization control unit, the pressurization progress determination unit determines the progress of pressurization, the model selection unit selects a trained model corresponding to the progress of pressurization, and the blood pressure estimation unit estimates the blood pressure value. The blood pressure measuring device according to claim 1.

14. In a blood pressure measurement method that estimates blood pressure values ​​using a trained model, The steps include: acquiring an output signal from a sensor that detects the pressure of a cuff compressing the part to be measured; The steps include: acquiring pulse wave data from the output signal of the aforementioned sensor; The steps include gradually pressurizing the cuff and determining the progress of pressurization based on the pulse wave data acquired up to that point, The steps include selecting a trained model from among several trained models pre-stored in the model memory unit that corresponds to the determined progress of pressurization, The steps include: estimating blood pressure values ​​using the pulse wave data acquired up to the present time and the selected trained model; A blood pressure measurement method having [a certain feature].

15. The processor of a blood pressure measuring device that estimates blood pressure values ​​using a trained model, The steps include: acquiring an output signal from a sensor that detects the pressure of a cuff compressing the part to be measured; The steps include: acquiring pulse wave data from the output signal of the aforementioned sensor; The steps include gradually pressurizing the cuff and determining the progress of pressurization based on the pulse wave data acquired up to that point, The steps include selecting a trained model from among several trained models pre-stored in the model memory unit that corresponds to the determined progress of pressurization, The steps include: estimating blood pressure values ​​using the pulse wave data acquired up to the present time and the selected trained model; A program to execute.

Images