Model generating method, estimation program and estimation apparatus

Abstract

To provide a technology for making it possible to implement a stable blood pressure measurement while lowering a cuff pressure adaptively to a subject.SOLUTION: A model generating apparatus according to one aspect of the present invention acquires a plurality of training samples regarding a pulse wave measured from a subject while a cuff is pressurized until an amplitude of the pulse wave assumes a maximum value, and carries out machine learning of an estimation model using the acquired plurality of training samples. Carrying out the machine learning includes training the estimation model so that an estimated value of a maximal blood pressure derived from each training sample by the estimation model matches a true value of the maximal blood pressure when each training sample was measured.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a model generation method, an estimation program, and an estimation device. [Background technology]

[0002] In recent years, technologies for managing health by measuring information about an individual's body, such as blood pressure values, using measuring devices and recording and analyzing the measurement results have become widespread. One example of a measuring device is a sphygmomanometer, which is configured to attach a cuff to a subject's upper arm, wrist, or other part of the body, measure pressure pulse waves while inflating the attached cuff, and measure blood pressure, including systolic blood pressure (SBP), based on the measured pressure pulse waves.

[0003] Patent Document 1 proposes a method for estimating the systolic blood pressure value using a cuff pressure lower than the general systolic blood pressure value. Specifically, the method proposed in Patent Document 1 measures pressure pulse waves up to a certain cuff pressure (e.g., 130 mmHg) and estimates the blood pressure value from the measured pressure pulse wave. Compared to the oscillometric method, which temporarily stops the blood flow, this method allows for a lower cuff pressure during measurement, thereby reducing discomfort caused by the cuff pressure. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] US Patent Application Publication No. 2020 / 0383579 [Patent Document 2] Japanese Patent Application Publication No. 03-280932 Summary of the Invention [Problem to be solved by the invention]

[0005] The inventors of the present invention have found that the above-described conventional method has the following problems. Specifically, in the above-described conventional method, the upper limit of the cuff pressure is set to a fixed value, which does not necessarily result in appropriate blood pressure measurement for each subject. For example, in the case of a subject with low systolic blood pressure, the cuff pressure may exceed the systolic blood pressure value, potentially making it impossible to suppress discomfort caused by the cuff pressure. On the other hand, in the case of a subject with high systolic blood pressure, cuff inflation may stop before the pulse wave amplitude becomes sufficiently large, making it difficult to perform appropriate blood pressure measurement. In other words, in the above-described conventional method, it may be difficult to lower the cuff pressure appropriately for each subject and achieve stable blood pressure measurement.

[0006] In one aspect, the present invention has been made in consideration of the above circumstances, and its purpose is to provide a technology that enables a cuff pressure to be lowered in accordance with the subject and that enables stable blood pressure measurement. [Means for solving the problem]

[0007] In order to solve the above-mentioned problems, the present invention employs the following configurations. Note that the following configurations of the invention can be combined as appropriate.

[0008] A model generation method according to one aspect of the present invention is executed by a computer. The model generation method includes acquiring a plurality of training samples of pulse waves measured from a subject while a cuff is inflated until the amplitude of the pulse wave reaches a maximum value, performing machine learning of an estimation model using the acquired training samples, and outputting the results of the machine learning. Performing the machine learning includes training the estimation model so that an estimated blood pressure value derived by the estimation model from each training sample matches the true value of the blood pressure at the time each training sample was measured.

[0009] This configuration makes it possible to generate a trained estimation model capable of estimating blood pressure values ​​from pulse wave samples up to and including the point at which the amplitude reaches its maximum (maximum point). During the cuff inflation process, the point at which the amplitude reaches its maximum occurs before the point at which the systolic blood pressure and the cuff pressure become equal. Therefore, the cuff pressure at the point at which the amplitude reaches its maximum is lower than the systolic blood pressure. Furthermore, by using a large number of training samples measured under various conditions from many subjects for machine learning, the generated trained estimation model is expected to acquire the ability to appropriately estimate blood pressure. Therefore, the generated trained estimation model can be expected to lower the cuff pressure in accordance with the subject and to perform stable blood pressure measurements.

[0010] In the model generation method according to the above aspect, each of the training samples may include at least one of envelope data and amplitude data of the pulse wave. The envelope data and amplitude data accurately represent pulse wave characteristics. Therefore, with this configuration, by using at least one sample of the envelope data and the amplitude data as input, it is possible to expect the generation of a trained estimation model that can appropriately estimate blood pressure values.

[0011] In the model generation method according to the above aspect, each of the training samples may further include beat-by-beat features of the pulse wave. The beat-by-beat features allow for grasping the characteristics of the pulse wave. Therefore, with this configuration, by further using the beat-by-beat features of the pulse wave as input, it is possible to generate a trained estimation model that can appropriately estimate blood pressure values.

[0012] The present invention is not limited to the above-described model generation method. Another aspect of the present invention may be an estimation program that uses a trained estimation model generated by the above-described model generation method. Furthermore, the application of a configuration for deriving an estimated blood pressure value from a pulse wave while the cuff is inflated until the amplitude of the pulse wave reaches its maximum value is not limited to the application of the trained estimation model generated by the above-described aspects.

[0013] For example, an estimation program according to one aspect of the present invention is a program for causing a computer to execute an estimation method. The estimation method includes acquiring target samples of a subject's pulse wave while the cuff is inflated until the amplitude of the pulse wave reaches a maximum value, deriving an estimate of the subject's blood pressure from the acquired target samples using an estimation model, and outputting information related to the derived estimate. This configuration makes it possible to lower the cuff pressure in accordance with the subject and to expect stable blood pressure measurement.

[0014] In the estimation program according to the above aspect, the target sample may be composed of an actually measured portion obtained by actually measuring the pulse wave until the amplitude of the pulse wave reaches a maximum value. With this configuration, an estimated blood pressure value can be derived using envelope data up to the peak value, which is a characteristic point of the pulse wave envelope. This allows the cuff pressure to be reduced in accordance with the subject, and more accurate blood pressure estimation can be expected.

[0015] In the estimation program according to the above aspect, the target sample may be composed of a measured portion obtained by actually measuring the pulse wave up to a characteristic time point of the pulse wave that appears before the amplitude of the pulse wave reaches a maximum, and a predicted portion obtained by predicting the pulse wave from the characteristic time point to a maximum time point of the pulse wave when the amplitude of the pulse wave reaches a maximum, based on at least a part of the measured portion. With this configuration, it is possible to stop inflating the cuff before the amplitude reaches a maximum value. This enables blood pressure measurement at a lower pressure.

[0016] In the estimation program according to the above aspect, the target sample may include at least one of envelope data and amplitude data of the pulse wave. With this configuration, by using a sample of at least one of the envelope data and the amplitude data as input, it is possible to appropriately estimate the blood pressure value. We can expect accurate estimates.

[0017] In the estimation program according to the above aspect, the target sample may further include beat-by-beat features of the pulse wave. With this configuration, by further using the beat-by-beat features of the pulse wave as an input, it is possible to expect improvement in the accuracy of estimating the blood pressure value.

[0018] Note that the present invention is not limited to the above-described model generation method (information processing method). As another aspect of the model generation method according to each of the above aspects, one aspect of the present invention may be an information processing device (model generation device) that realizes all or part of the above-described configurations, a program, or a storage medium readable by a machine such as a computer on which such a program is stored. A storage medium readable by a machine such as a computer is a medium that stores information such as a program by electrical, magnetic, optical, mechanical, or chemical action. Similarly, the present invention is not limited to the above-described estimation program. As another aspect of the estimation program according to each of the above aspects, one aspect of the present invention may be an information processing device (estimation device) that realizes all or part of the above-described configurations, an information processing method (estimation method), or a storage medium readable by a machine such as a computer on which a program is stored.

[0019] For example, an estimation device according to one aspect of the present invention includes a control unit configured to acquire target samples of a pulse wave of a subject while a cuff is inflated until the amplitude of the pulse wave reaches a maximum value, derive an estimate of the subject's blood pressure from the acquired target samples using an estimation model, and output information related to the derived estimate. [Effects of the Invention]

[0020] According to the present invention, it is possible to provide a technique for lowering the cuff pressure in accordance with the subject and for enabling stable blood pressure measurement. [Brief explanation of the drawings]

[0021] [Figure 1]FIG. 1 shows a schematic diagram of an example of a situation in which the present invention is applied. [Figure 2A] FIG. 2A schematically shows an example of pulse wave data measured while the cuff is being inflated. [Figure 2B] FIG. 2B schematically shows an example of the feature amount of pulse wave data. [Figure 2C] FIG. 2C schematically shows an example of the feature amount of pulse wave data. [Figure 2D] FIG. 2D schematically shows an example of the feature amount of pulse wave data. [Figure 3A] FIG. 3A schematically illustrates an example of how a portion of a target sample is obtained by prediction in an embodiment. [Figure 3B] FIG. 3B schematically illustrates an example of how a portion of a target sample is obtained by prediction in an embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of a hardware configuration of a model generating device according to an embodiment. [Figure 5] FIG. 5 schematically illustrates an example of a hardware configuration of a blood pressure measurement device according to an embodiment. [Figure 6] FIG. 6 is a diagram illustrating an example of a software configuration of the model generating device according to the embodiment. [Figure 7] FIG. 7 schematically illustrates an example of the software configuration of the blood pressure measurement device according to the embodiment. [Figure 8] FIG. 8 is a flowchart illustrating an example of a processing procedure of the model generating device according to the embodiment. [Figure 9] FIG. 9 is a flowchart illustrating an example of a processing procedure of the blood pressure measurement device according to the embodiment. [Figure 10] FIG. 10 is a diagram illustrating an example of an application scene of the estimation device according to the modified example. [Figure 11] FIG. 11 schematically illustrates an example of a hardware configuration of an estimation device according to a modified example. DETAILED DESCRIPTION OF THE INVENTION

[0022] An embodiment according to one aspect of the present invention (hereinafter also referred to as "the present embodiment") will be described below with reference to the drawings. However, the present embodiment described below is merely an example of the present invention in all respects. Various improvements or modifications may be made without departing from the scope of the present invention. In implementing the present invention, a specific configuration according to the embodiment may be appropriately adopted. Note that while data appearing in the present embodiment is described in natural language, more specifically, it is specified using computer-recognizable pseudo-language, commands, parameters, machine language, etc.

[0023] §1 Application Examples 1 is a schematic diagram showing an example of a situation in which the present invention is applied. The system according to this embodiment includes a model generating device 1 and a blood pressure measuring device 2.

[0024] The model generation device 1 according to this embodiment is one or more computers configured to generate a trained estimation model by performing machine learning. The model generation device 1 according to this embodiment acquires multiple training samples 31 of pulse waves measured from a subject SU while the cuff is inflated until the amplitude of the pulse wave reaches its maximum value. The model generation device 1 performs machine learning on an estimation model 40 using the acquired multiple training samples 31. Performing machine learning includes training the estimation model 40 so that the estimated blood pressure value derived by the estimation model 40 from each training sample 31 matches the true blood pressure value at the time each training sample 31 was measured. By performing this machine learning, a trained estimation model 40 can be generated that has acquired the ability to estimate blood pressure values ​​(derive estimated blood pressure values) from pulse wave samples up to the time when the amplitude reaches its maximum value. The model generation device 1 outputs the results of the machine learning.

[0025] On the other hand, the blood pressure measurement device 2 according to this embodiment is configured to include a computer configured to measure blood pressure values. The blood pressure measurement device 2 is an example of an estimation device according to the present disclosure. The blood pressure measurement device 2 according to this embodiment acquires a target sample 51 of the pulse wave of the subject TU while the cuff is inflated until the amplitude of the pulse wave reaches its maximum value. The blood pressure measurement device 2 derives an estimated value 52 of the blood pressure of the subject TU from the acquired target sample 51 using an estimation model 41. In one example, the estimation model 41 may be the trained estimation model 40 generated by the model generation device 1. The blood pressure measurement device 2 outputs information related to the derived estimated value 52.

[0026] During the cuff inflation process, the time when the amplitude reaches its maximum occurs before the time when the cuff pressure equals the systolic blood pressure. Therefore, the cuff pressure at the time when the amplitude reaches its maximum is lower than the systolic blood pressure. Furthermore, by using a large number of training samples 31 measured from many subjects under various conditions for machine learning, the generated trained estimation model 40 can be expected to acquire the ability to appropriately estimate blood pressure. Therefore, the trained estimation model 40 generated by the model generation device 1 according to this embodiment can be expected to lower the cuff pressure in accordance with the subject and to perform stable blood pressure measurement. In the blood pressure measurement device 2, the cuff pressure can be lowered in accordance with the subject TU and to perform stable blood pressure measurement.

[0027] (Estimated target) In this embodiment, the blood pressure values ​​of the subjects estimated by the estimation models (40, 41) are The values ​​may be selected appropriately depending on the embodiment. In one example, the blood pressure value to be estimated may include at least one of the systolic blood pressure, the diastolic blood pressure, and the mean blood pressure. In another example, the blood pressure value to be estimated may include the value of the systolic blood pressure. Furthermore, the blood pressure value to be estimated may further include other blood pressure values ​​other than the systolic blood pressure, such as the diastolic blood pressure and the mean blood pressure. In yet another example, the blood pressure value to be estimated may be only the value of the systolic blood pressure, without including other blood pressure values.

[0028] (Training sample collection method) 2A is a schematic diagram illustrating an example of pulse wave data measured while the cuff is being inflated. As shown in FIG. 2A, when the cuff is first inflated and the cuff pressure is increased, the amplitude of the measured pulse wave (pressure pulse wave) first gradually increases and reaches a maximum value. After the amplitude of the pulse wave reaches its maximum value, as the cuff pressure is further increased, the amplitude of the measured pulse wave gradually decreases.

[0029] The diastolic blood pressure appears after the start of cuff inflation but before the amplitude of the pulse wave reaches its maximum. The diastolic blood pressure corresponds to the cuff pressure at which the value of the first derivative of the envelope of the pulse wave reaches its maximum. After the start of cuff inflation but before the cuff pressure reaches the diastolic blood pressure, a stable point appears as a characteristic time point. The stable point is the time when the value of the second derivative of the envelope reaches its maximum. Generally, the stable point appears around 40 mmHg. On the other hand, the systolic blood pressure appears after the amplitude of the pulse wave reaches its maximum but before the pulse wave completely stops (the amplitude becomes 0).

[0030] In the oscillometric method, systolic and diastolic blood pressures are measured based on the characteristics of the pulse wave that appears during the process of inflating the cuff to a pressure exceeding the systolic blood pressure. Therefore, the cuff pressure increases during measurement. In contrast, in one example of this embodiment, cuff inflation is stopped near the point where the amplitude reaches its maximum, and blood pressure values, including the systolic blood pressure, are estimated from samples of the obtained pulse wave data (pulse wave samples). This allows the cuff pressure during measurement to be reduced.

[0031] Each training sample 31 may be appropriately collected so as to include pulse wave data up to the point where the amplitude reaches a maximum value during the process of properly measuring the blood pressure of the subject SU. In one example, each training sample 31 may be acquired from pulse wave data measured from the start of cuff inflation to the point where the amplitude reaches a maximum value during cuff inflation during the process of measuring the blood pressure of the subject SU using an existing method such as an oscillometric method.

[0032] The pulse wave data is obtained by attaching a cuff to the measurement site and measuring the magnitude of vibrations at the measurement site caused by the pulse (e.g., fluctuations in cuff pressure) while the cuff is compressing the measurement site. The magnitude of vibrations at the measurement site may correspond to the amplitude of the pulse wave. The measurement site may be appropriately selected from the upper arm, wrist, etc. The configuration of the pulse wave data is not particularly limited as long as it shows the characteristics of the pulse wave, and may be appropriately selected depending on the embodiment.

[0033] The amplitude data and envelope data shown in FIG. 2A are examples of pulse wave data. The amplitude data is generated by plotting measured values ​​of the pulse wave amplitude (vibration at the measurement site) in a time series. The amplitude data may be raw data, or may be obtained by applying preprocessing such as filtering to the raw data. The envelope data (peak envelope data) is generated by extracting the maximum values ​​(peaks) of the pulse wave amplitude and plotting the extracted maximum values ​​in a time series. The point at which the pulse wave amplitude reaches its maximum corresponds to the point at which the envelope reaches its peak value.

[0034] Each training sample 31 may include at least one of the envelope data and amplitude data of the pulse wave. As shown in FIG. 2A, the envelope data and amplitude data accurately represent the characteristics of the pulse wave. Therefore, by using at least one sample of the envelope data and amplitude data as input, it is expected that a trained estimation model 40 capable of appropriately estimating blood pressure values ​​can be generated.

[0035] The input data may not be limited to envelope data and amplitude data. In another example, the pulse wave data may be composed of data other than envelope data and amplitude data. The pulse wave data may include at least one of envelope data and amplitude data as well as other data other than envelope data and amplitude data. In one example, each training sample 31 may include at least one of envelope data and amplitude data as well as other data other than envelope data and amplitude data. As an example of other data, each training sample 31 may further include cuff pressure data. Furthermore, each training sample 31 may further include beat-by-beat feature values ​​of the pulse wave. As long as the feature values ​​are configured to indicate the characteristics of the pulse wave for each beat, the type of feature value is not particularly limited and may be selected appropriately depending on the embodiment. The feature values ​​may be calculated by any arithmetic processing. In one example, the feature values ​​may be statistical values ​​such as maximum values, minimum values, variance, standard deviation, n% tile values, skewness, and kurtosis. In another example, the feature values ​​may be values ​​other than statistical values.

[0036] 2B to 2D are schematic diagrams showing examples of beat-by-beat feature quantities (RAV, WID, DFN) in a pulse wave. RAV is calculated by normalizing the pulse wave area for each beat, indicated by the diagonal lines in FIG. 2B, by the amplitude. In one example, RAV may be calculated by (pulse wave area / amplitude of pulse wave within one beat)×100. WID is calculated by dividing the pulse wave area by the amplitude of the pulse wave from the peak (Maximum) to the threshold, as shown in FIG. 2C. The waveform width, which is defined as the time it takes for the amplitude to decrease to the threshold, is called the pulse wave period. The RAV, WID, and DFN indicate the area, width, and slope of the pulse wave for one beat. Each training sample 31 may further include at least one of the RAV, WID, and DFN for each beat. The differential and integral values ​​(area) of the pulse wave may be calculated by any signal processing, such as the difference or sum between adjacent sample points. A known method, such as that disclosed in Patent Document 2, may be used to calculate each value.

[0037] The starting point of the measurement range from which samples are obtained is not particularly limited and may be determined as appropriate depending on the embodiment. In one example, each training sample 31 may include pulse wave data from immediately after the start of cuff inflation. In another example, each training sample 31 may be configured to include pulse wave data from a predetermined time point, such as at a predetermined cuff pressure, after an arbitrary time has elapsed since the start of cuff inflation.

[0038] For example, "until the amplitude of the pulse wave reaches its maximum value" may be "until the amplitude of the pulse wave reaches its maximum value." It is known that the cuff pressure at the time when the amplitude reaches its maximum value corresponds to the mean blood pressure. Therefore, if there is pulse wave data up to the maximum amplitude, it is possible to appropriately estimate blood pressure. Therefore, the model generation device 1 can be expected to generate a trained estimation model 40 that has acquired the ability to appropriately estimate blood pressure. Furthermore, the blood pressure measurement device 2 can be expected to perform appropriate blood pressure measurement.

[0039] In this case, the end point of the measurement range from which samples are obtained is not particularly limited as long as it includes the time when the maximum amplitude is reached, and may be set appropriately depending on the embodiment. In one example, the measurement range may extend beyond the time when the maximum amplitude is reached without reaching the time of systolic blood pressure. That is, the time when the maximum amplitude is reached may be set as the end point of the measurement range. In another example, the measurement range may be set so that the end point of the measurement range does not exceed the time when the maximum amplitude is reached. Accordingly, each training sample 31 may include some pulse wave data after the time when the maximum amplitude is reached, or may be configured not to include pulse wave data after the time when the maximum amplitude is reached. Note that it is preferable to have as little pulse wave data as possible after the time when the maximum amplitude is reached.

[0040] In another example, "before the amplitude of the pulse wave reaches its maximum value" may be configured to exclude the point at which the amplitude reaches its maximum value. In this case, the end point of the measurement range for obtaining samples may be appropriately set to include a characteristic time point of the pulse wave that appears before the amplitude reaches its maximum value. The end point of the measurement range may be set to the characteristic time point, or may be set to a time point beyond the characteristic time point without reaching the point at which the amplitude reaches its maximum value. The characteristic time point may be, for example, the time of the diastolic blood pressure or a stable point. Each training sample 31 may be a sample of the pulse wave measured from the subject SU while the cuff is inflated until the characteristic time point of the pulse wave that appears before the amplitude of the pulse wave reaches its maximum value. The characteristic time point can be uniformly identified for each subject. In addition, because pulse wave characteristics are reflected in the pulse wave data up to the characteristic time point, appropriate blood pressure estimation can be expected from the pulse wave data up to this characteristic time point. Therefore, the model generation device 1 can be expected to generate a trained estimation model 40 that has acquired the ability to appropriately estimate blood pressure. Furthermore, appropriate blood pressure measurement can be expected from the blood pressure measurement device 2.

[0041] The true value of blood pressure when each training sample 31 is measured may be appropriately obtained during the process of collecting each training sample 31. As described above, in one example, the true value of blood pressure may be obtained along with each training sample 31 by measuring the target blood pressure of the subject SU using an existing method such as an oscillometric method. In one example, the target blood pressure may include at least one of systolic blood pressure, diastolic blood pressure, and mean blood pressure. When training the estimation model 40 to acquire the ability to estimate multiple types of blood pressure values, the true values ​​of various blood pressures may be obtained.

[0042] The acquired true value may be retained as appropriate. In one example, as shown in FIG. 1, the true value of the subject's blood pressure may be acquired as a label 32, and a training dataset 30 may be configured by combining the training samples 31 and the labels 32. The labels 32 may also be referred to as ground truth data, teacher signals, etc. When training the estimation model 40 to acquire the ability to estimate multiple types of blood pressure values, the labels 32 may be configured to indicate the true values ​​of each of the multiple types of blood pressure.

[0043] The number of subjects SU from which the training samples 31 are collected is not particularly limited and may be determined appropriately depending on the embodiment. The subject TU whose blood pressure is measured in the estimation stage may or may not be included among the subjects SU.

[0044] (Estimated model) The estimation model 40 is configured by a machine learning model. The machine learning model is configured to have one or more calculation parameters that can be adjusted by machine learning. The one or more calculation parameters are used to calculate the desired inference (in this embodiment, estimation of blood pressure values). The machine learning model may be configured by, for example, a neural network, a regression model, a decision tree model, a support vector machine, or other functional formulas (calculation models). The machine learning method may be appropriately selected depending on the machine learning model employed (for example, backpropagation).

[0045] In one example, the estimation model 40 may include a neural network. The structure of the neural network is not particularly limited and may be determined appropriately 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, the connection relationships between the nodes in each layer, etc. In one example, the neural network may include any mechanism such as a recurrent 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, a deconvolutional layer, an unpooling layer, a normalization layer, a dropout layer, or a long short-term memory (LSTM). The neural network may be any type of model such as a diffusion model, a transformer model, or a generative model. The weight of the connections between the nodes included in the neural network and the threshold value of each node are examples of the calculation parameters.

[0046] Machine learning involves adjusting (optimizing) the values ​​of computational parameters of a machine learning model using each training sample 31. Typically, a trained machine learning model (estimation model 40) may be generated by performing supervised learning using multiple datasets 30. That is, the values ​​of computational parameters of the machine learning model may be adjusted so that the output obtained from the machine learning model by providing the training samples 31 of each dataset 30 matches the true value indicated by the corresponding label 32. In one example, when the estimation model 40 is configured using a neural network, the values ​​of the computational parameters of the machine learning model may be adjusted using a backpropagation algorithm. Note that as long as the ability to estimate blood pressure values ​​can be acquired, the training method for the machine learning model is not limited to this example and may be changed as appropriate depending on the embodiment. Furthermore, the machine learning model may be trained online or offline. The machine learning model may be tuned as appropriate by transfer learning, re-learning, additional learning, etc.

[0047] The input / output format of the estimation model 40 may be determined appropriately depending on the embodiment. In one example, the input data may be provided to the estimation model 40 as is, or may be provided after being preprocessed. In another example, the output of the estimation model 40 may be configured to directly or indirectly indicate an estimated value of blood pressure. When the output of the estimation model 40 is configured to indirectly indicate an estimated value of blood pressure, the estimated value of blood pressure may be obtained by performing any information processing (such as interpretation processing) on ​​the output of the estimation model 40.

[0048] Furthermore, the estimation model 40 may be configured to be able to appropriately derive an estimated value of blood pressure. In one example, the estimation model 40 may be configured to directly derive an estimated value of blood pressure. In another example, the estimation model 40 may be configured to indirectly derive an estimated value of the target blood pressure by predicting the time point of the target blood pressure, assuming that the increase in cuff pressure complies with a predetermined condition (e.g., is constant). In this case, the estimated value of the target blood pressure can be derived by calculating the value of the cuff pressure at the time point predicted by the estimation model 40.

[0049] The estimation model 41 used in the estimation stage may be configured by any computational model that can execute inference processing (computation) equivalent to that of the trained estimation model 40. In one example, the trained estimation model 40 may be used as the estimation model 41. In another example, the estimation model 41 may be configured by a computational model other than the trained estimation model 40. The other computational model may be generated manually or otherwise as appropriate.

[0050] (Target sample) The target sample 51 may be configured similarly to the training sample 31, except that it is obtained from a subject TU whose blood pressure has been measured. In one example, the target sample 51 may include at least one of envelope data and amplitude data of a pulse wave. By using a sample of at least one of the envelope data and amplitude data as input, it is possible to expect an appropriate estimation of the blood pressure value.

[0051] In one example, the target sample 51 may further include beat-by-beat feature values ​​of the pulse wave. By further using the beat-by-beat feature values ​​of the pulse wave as input, it is possible to expect an improvement in the accuracy of estimating the blood pressure value. As described above, the type of feature value may be appropriately selected depending on the embodiment. In one example, the target sample 51 may further include at least one of the RAV, WID, and DFN for each beat. In another example, the target sample 51 may further include cuff pressure data.

[0052] In the blood pressure measurement device 2, the target sample 51 may be acquired as needed. "Until the amplitude of the pulse wave reaches its maximum value" may mean "until the amplitude of the pulse wave reaches its maximum value." In one example of this case, the blood pressure measurement device 2 monitors the amplitude of the measured pulse wave, stops inflating the cuff and measuring the pulse wave when the amplitude exceeds the maximum value, and acquires the target sample 51 from the pulse wave data measured up to that point. That is, the target sample 51 may be obtained entirely by actual measurement. Accordingly, the target sample 51 may be composed of an actual measurement portion obtained by actually measuring the pulse wave until the amplitude of the pulse wave reaches its maximum value. In this example, it may be acceptable for the time when cuff inflation is stopped to be after the time when the amplitude reaches its maximum value, and the target sample 51 may include a data portion beyond the time when the amplitude reaches its maximum value. According to one example of the present embodiment, the cuff pressure can be reduced in accordance with the subject TU, and more accurate blood pressure estimation can be expected.

[0053] Note that when the target sample 51 is obtained entirely through actual measurement, the method for stopping cuff inflation is not limited to this example and may be modified as appropriate depending on the embodiment. In another example, the blood pressure measurement device 2 may predict the time of maximum amplitude from measured pulse wave data before the amplitude reaches its maximum value, and stop cuff inflation and pulse wave measurement at the predicted time. For example, the blood pressure measurement device 2 may predict the time of maximum pulse wave amplitude from a characteristic time point of the pulse wave that appears before the pulse wave amplitude reaches its maximum value. A characteristic time point is a time point at which any characteristic appears in the pulse wave. In other words, a characteristic time point is a time point that can be uniformly detected for each subject SU and each object TU when the pulse wave data satisfies a predetermined condition. The time point of the minimum blood pressure and the stable point are examples of characteristic time points.

[0054] In one example, the blood pressure measurement device 2 may calculate the first or second derivative of the measured pulse wave data (envelope data) and monitor the calculated derivative value to detect the time of the diastolic blood pressure or a stable point as a characteristic time point. The blood pressure measurement device 2 may predict the cuff pressure at the time of the maximum amplitude from the cuff pressure at the diastolic blood pressure or the stable point. The method for predicting the time of the maximum amplitude from information up to the characteristic time point is not particularly limited and may be selected appropriately depending on the embodiment. In a simple example, the blood pressure measurement device 2 may obtain, as a predicted value of the cuff pressure at the time of the maximum amplitude, the product calculated by multiplying the cuff pressure at the time of the diastolic blood pressure or the stable point by a predetermined value. In another example, the blood pressure measurement device 2 may predict the time of the maximum amplitude from the pulse wave data and cuff pressure up to the characteristic time point using a computational model. The computational model may be configured as a trained machine learning model such as a regression model.

[0055] In another example, the blood pressure measurement device 2 may acquire a portion of the target sample 51 by actual measurement and then predict the remaining portion of the target sample 51 from the actually measured portion. The prediction range may be determined appropriately depending on the embodiment. In one example, the target sample 51 may be composed of an actually measured portion obtained by actually measuring the pulse wave up to a characteristic time point of the pulse wave that appears before the amplitude of the pulse wave reaches its maximum, and a predicted portion obtained by predicting the pulse wave from the characteristic time point to the maximum time point at which the amplitude of the pulse wave reaches its maximum, based on at least a portion of the actually measured portion. That is, the blood pressure measurement device 2 may actually measure the pulse wave up to the characteristic time point and predict pulse wave data for the unmeasured portion from the measured pulse wave data, thereby obtaining pulse wave data from the start of cuff inflation to the maximum time point. The blood pressure measurement device 2 may acquire the target sample 51 from the obtained pulse wave data. In this case, the time at which measurement is stopped may coincide with the characteristic time point or may be slightly beyond the characteristic time point.

[0056] 3A and 3B are schematic diagrams showing an example of a mode in which a portion of the target sample 51 is obtained by prediction in this embodiment. FIG. 3A shows an example of a scene in which the pulse wave is actually measured up to the point of minimum blood pressure, and the remaining portion is predicted. FIG. 3B shows an example of a scene in which the pulse wave is actually measured up to a stable point, and the remaining portion is predicted. The method of predicting the pulse wave is not particularly limited, and may be selected appropriately depending on the embodiment. In one example, a calculation model may be used to generate the predicted portion. The calculation model may be configured by a trained machine learning model. For example, the calculation model may be configured to predict the maximum blood pressure from the start of cuff inflation. Training pulse wave data may be collected by actually measuring pulse waves up to time t. The training pulse wave data may be the same as the training sample 31. Then, a regression model (trained machine learning model) that predicts pulse wave data at time t+n+1 from pulse wave data from time t to time t+n may be generated by performing regression analysis on the collected pulse wave data. n is a natural number greater than or equal to 1. In this case, the blood pressure measurement device 2 can generate a predicted portion by repeatedly predicting pulse wave data using the generated regression model. In one example, the blood pressure measurement device 2 may further employ the above-described configuration for predicting the maximum time from a characteristic time point and generate pulse wave data (predicted portion) from the characteristic time point to the maximum time by repeatedly predicting pulse wave data using the regression model up to the predicted maximum time. The predicted pulse wave data may be at least one of envelope data and amplitude data. When the target sample 51 is configured to further include a feature value for each beat, the blood pressure measurement device 2 may acquire the feature value of the predicted portion by calculating the feature value for each beat from the generated predicted portion. Note that the configuration of the trained machine learning model is not limited to this example, and may be modified as appropriate depending on the embodiment, as long as it is capable of generating a prediction portion. According to one example of this embodiment, during blood pressure measurement, the cuff pressure can be stopped before the amplitude reaches its maximum value. This allows blood pressure measurement to be performed at a lower cuff pressure.

[0057] In another example, "until the amplitude of the pulse wave reaches its maximum value" may be configured to not include the time when the amplitude reaches its maximum value. In one example of this case, the target sample 51 may be a sample of a pulse wave measured from the subject TU while the cuff is inflated up to a characteristic time when the pulse wave appears before the amplitude of the pulse wave reaches its maximum value. The target sample 51 may be obtained entirely by actual measurement. Alternatively, the target sample 51 may be obtained by actually measuring a portion of the target sample 51 and predicting the remainder from the actually measured portion. According to one example of this embodiment, blood pressure measurement is possible at a lower cuff pressure.

[0058] §2 Configuration example [Hardware configuration] (Model generation device) 4 schematically shows an example of the hardware configuration of the model generation device 1 according to this embodiment. The model generation device 1 according to this embodiment is a computer in which a control unit 11, a storage unit 12, a communication interface 13, an input device 14, an output device 15, and a drive 16 are electrically connected.

[0059] The control unit 11 includes a CPU (Central Processing Unit) which is a hardware processor, The control unit 11 includes a RAM (Random Access Memory), a ROM (Read Only Memory), etc., and is configured to execute information processing based on programs and various data. The control unit 11 (CPU) is an example of a processor resource.

[0060] The storage unit 12 may be configured, for example, with a hard disk drive, a solid state drive, a semiconductor memory, or the like. The storage unit 12, RAM, and ROM are examples of memory resources. In this embodiment, the storage unit 12 stores various information such as a generation program 81, multiple datasets 30, and learning result data 400. The generation program 81 is a program for causing the model generation device 1 to execute information processing (see FIG. 8, described below) related to machine learning of the estimation model 40. The generation program 81 includes a series of instructions for the information processing.

[0061] The learning result data 400 indicates information related to the trained estimation model 40. As long as the trained estimation model 40 can be reproduced when estimating blood pressure, the configuration of the learning result data 400 is not particularly limited and may be determined appropriately depending on the embodiment. In one example, the learning result data 400 may include information indicating values ​​of calculation parameters adjusted by machine learning. In some cases, the learning result data 400 may include information indicating the configuration of the estimation model 40 (for example, the neural network In this embodiment, the learning result data 400 is generated as a result of executing instructions included in the generation program 81.

[0062] The communication interface 13 is configured to perform wired or wireless data communication via a network. The communication interface 13 may be configured, for example, by a wired LAN (Local Area Network) module, a wireless LAN module, etc. In this embodiment, the model generation device 1 may perform data communication with another computer via the communication interface 13.

[0063] The input device 14 is a device for inputting, for example, a mouse, a keyboard, an operator, etc. The output device 15 is a device for outputting, for example, a display, a speaker, etc. An operator can operate the model generation device 1 by using the input device 14 and the output device 15. The input device 14 and the output device 15 may be connected via an external interface. The input device 14 and the output device 15 may be integrally configured, for example, by a touch panel display, etc.

[0064] The drive 16 is a device for reading various information, such as a program, stored in a storage medium 91. At least one of the generation program 81, the plurality of data sets 30, and the learning result data 400 may be stored in the storage medium 91 instead of or together with the storage unit 12. The storage medium 91 is configured to accumulate various information (such as the stored program) by electrical, magnetic, optical, mechanical, or chemical action so that a machine, such as a computer, can read the information. The model generation device 1 may acquire at least one of the generation program 81 and the learning result data 400 from the storage medium 91. The storage medium 91 may be a disk-type storage medium, such as a CD or DVD, or a non-disk-type storage medium, such as a semiconductor memory (e.g., a flash memory). The storage medium 91 may be a disk-type storage medium, such as a CD or DVD, or a non-disk-type storage medium, such as a semiconductor memory (e.g., a flash memory). The type of the drive 16 may be appropriately selected depending on the type of the storage medium 91. The drive 16 may be connected via an external interface or a communication interface 13. The external interface may be configured as appropriate to connect to an external device via a wired or wireless connection using, for example, a USB (Universal Serial Bus) port, a dedicated port, or the like.

[0065] Regarding the specific hardware configuration of the model generating device 1, components can be omitted, replaced, or added as appropriate depending on the embodiment. For example, the control unit 11 may include multiple hardware processors. The hardware processors may be a microprocessor, a field-programmable gate array (FPGA), a digital signal processor (DSP), a GPU (Gateway Processor), a 3D processor, a 3D image ... The communication interface 13, the input device 14, and the output device 15 may be configured by a PU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), etc. and drive 16 may be omitted. The model generation device 1 may be configured with multiple computers. In this case, the hardware configurations of the computers may be the same or may be at least partially different. At least one of the generation program 81, the multiple datasets 30, and the learning result data 400 may be stored in a storage medium of an external computer, such as a NAS (Network Attached Storage). The model generation device 1 may acquire at least one of the generation program 81 and the multiple datasets 30 from an external computer. Furthermore, the model generation device 1 may be an information processing device designed specifically for the services provided, as well as a general-purpose server device, a general-purpose PC (Personal Computer), etc.

[0066] (blood pressure measuring device) FIG. 5 schematically shows an example of the hardware configuration of the blood pressure measurement device 2 according to this embodiment. The blood pressure measuring device 2 according to this embodiment includes a control unit 21, a memory unit 22, an oscillator circuit 231, a pump drive circuit 232, a valve drive circuit 233, a cuff 240, a pressure sensor 241, a pressure pump 242, an exhaust valve 243, an air tube 244, an operation switch 25, a display unit 26, and a power supply 27. The control unit 21 and the memory unit 22 are an example of a computer part of the blood pressure measuring device 2.

[0067] Cuff 240 includes an air bag 2401 containing air. Cuff 240 is provided with a pressure sensor 241, a pressurizing pump 242, and an exhaust valve 243 via an air tube 244. Pressure sensor 241 is configured to detect the pressure (cuff pressure) inside air bag 2401 of cuff 240. Pressurizing pump 242 is configured to supply air into air bag 2401. Exhaust valve 243 is provided as a boundary between the internal space and the external space of air bag 2401, and is configured to be openable and closable. By closing exhaust valve 243, air is trapped inside air bag 2401, and the pressure inside air bag 2401 can be maintained. On the other hand, by opening exhaust valve 243, air can be discharged from air bag 2401, thereby reducing the pressure.

[0068] The control unit 21 includes a hardware processor such as a CPU, and is configured to execute information processing based on programs and various data. The control unit 21 (CPU) is an example of a processor resource of the estimation device. The oscillation circuit 231 outputs a signal having an oscillation frequency corresponding to the output value of the pressure sensor 241 to the control unit 21. In this embodiment, the control unit 21 processes the signal from the oscillation circuit 231 to obtain cuff pressure and pulse wave data. The pump drive circuit 232 is configured to control the drive of the pressure pump 242 based on a control signal output from the control unit 21. The valve drive circuit 233 is configured to control the opening and closing of the exhaust valve 243 based on a control signal output from the control unit 21.

[0069] The storage unit 22 may be configured, for example, with a semiconductor memory. The storage unit 22 is an example of a memory resource of the estimation device. In this embodiment, the storage unit 22 stores various information related to the estimation program 82, model data 410, and the like. The estimation program 82 is a program for causing the blood pressure measurement device 2 to execute information processing related to blood pressure value estimation (see FIG. 9, which will be described later). The estimation program 82 includes a series of instructions for the information processing. The model data 410 indicates information related to the estimation model 41. As with the learning result data 400, the configuration of the model data 410 is not particularly limited as long as the estimation model 41 can be reproduced during blood pressure estimation, and may be determined appropriately depending on the embodiment. In an embodiment in which a trained estimation model 40 is used as the estimation model 41, the model data 410 may be the learning result data 400. The model data 410 may be incorporated into the estimation program 82. The storage unit 22 may also store other information, such as blood pressure measurement results (cuff pressure, pulse wave data, estimated blood pressure values, etc.), as appropriate.

[0070] The operation switch 25 is used to perform operations such as starting blood pressure measurement. The operation switch 25 may be configured as at least one of a physical switch and a virtual switch. The display unit 26 is configured to display various information such as the results of blood pressure measurement. The operation switch 25 and the display unit 26 may be integrated into a touch panel display. The power supply 27 is configured to supply power to each unit such as the control unit 21.

[0071] Note that, with regard to the specific hardware configuration of the blood pressure measurement device 2, components can be omitted, replaced, or added as appropriate depending on the embodiment. For example, the control unit 21 may include multiple hardware processors. The hardware processor may be configured with a microprocessor, FPGA, DSP, ASIC, etc. The blood pressure measurement device 2 may be equipped with a communication interface. For example, the blood pressure measurement device 2 may be configured to be able to communicate data with a user terminal such as a smartphone by being equipped with a near-field communication module. This allows the blood pressure measurement device 2 to acquire any data via the user terminal. At least one of the estimation program 82 and the model data 410 may be stored in a storage medium of an external computer, such as a NAS. The blood pressure measurement device 2 may store the estimation program 82 and the model data 410. At least one of the data 410 and the data 411 may be acquired from an external computer. The blood pressure measurement device 2 may acquire data directly from the external computer, or may acquire data indirectly from the external computer via the user terminal as described above. The blood pressure measurement device 2 may be configured as a general blood pressure monitor, or may be configured as a wearable device such as a wristwatch.

[0072] [Software configuration] (Model generation device) 6 schematically shows an example of the software configuration of the model generation device 1 according to this embodiment. The control unit 11 of the model generation device 1 executes instructions included in the generation program 81 stored in the storage unit 12 using the CPU. As a result, the model generation device 1 operates as a computer including an acquisition unit 111, a learning processing unit 112, and an output processing unit 113 as software modules. That is, in this embodiment, each software module of the model generation device 1 is realized by the control unit 11 (CPU).

[0073] The acquisition unit 111 is configured to acquire a plurality of training samples 31 of pulse waves measured from the subject SU while the cuff is inflated until the amplitude of the pulse wave reaches its maximum value. The learning processing unit 112 is configured to perform machine learning of the estimation model 40 using the acquired plurality of training samples 31. Performing machine learning includes training the estimation model 40 so that an estimated value of blood pressure derived by the estimation model 40 from each training sample 31 matches the true value of blood pressure at the time each training sample 31 was measured. The output processing unit 113 is configured to output the results of the machine learning.

[0074] (blood pressure measuring device) 7 schematically shows an example of the software configuration of the blood pressure measurement device 2 according to this embodiment. The control unit 21 of the blood pressure measurement device 2 executes instructions included in the estimation program 82 stored in the storage unit 22 using the CPU. As a result, the blood pressure measurement device 2 operates as a computer including an acquisition unit 211, an estimation unit 212, and an output processing unit 213 as software modules. That is, in this embodiment, each software module of the blood pressure measurement device 2 is also realized by the control unit 21 (CPU).

[0075] The acquisition unit 211 is configured to acquire a target sample 51 of the pulse wave of the subject TU while the cuff is inflated until the amplitude of the pulse wave reaches its maximum value. The estimation unit 212 has an estimation model 41 by holding model data 410. The estimation unit 212 is configured to derive an estimated value 52 of the blood pressure of the subject TU from the acquired target sample 51 using the estimation model 41. The output processing unit 213 is configured to output information related to the derived estimated value 52.

[0076] (others) In the present embodiment, an example is described in which each software module of the model generating device 1 and the blood pressure measurement device 2 is implemented by a general-purpose CPU. However, some or all of the software modules may be implemented by one or more dedicated processors or chipsets. Each module may be implemented as a hardware module. With regard to the software configuration of the model generating device 1 and the blood pressure measurement device 2, modules may be omitted, replaced, or added as appropriate depending on the embodiment.

[0077] §3 Example of operation [Model generation device] 8 is a flowchart showing an example of the processing procedure of the model generating device 1 according to this embodiment. The following processing procedure is an example of a model generating method executed by a computer. However, the following processing procedure of the model generation device 1 is merely an example, and each step may be changed as much as possible. Furthermore, steps in the following processing procedure may be omitted, replaced, or added as appropriate depending on the embodiment.

[0078] (Step S101) In step S101, the control unit 11 operates as the acquisition unit 111 and acquires multiple training samples 31 of pulse waves measured from the subject SU while the cuff is inflated until the amplitude of the pulse wave reaches its maximum value. In one example, each training sample 31 may include at least one of pulse wave envelope data and amplitude data. In one example, each training sample 31 may further include cuff pressure data. In another example, each training sample 31 may further include beat-to-beat features of the pulse wave. For example, each training sample 31 may further include at least one of beat-to-beat RAV, WID, and DFN. In one example, the control unit 11 may acquire multiple data sets 30, each of which is composed of a combination of the training samples 31 and the labels 32. Each data set 30 may be collected by the model generation device 1 or by a computer other than the model generation device 1. When at least some of the multiple data sets 30 are collected on another computer, the control unit 11 may acquire at least some of the multiple data sets 30 from the other computer via a network, a storage medium 91, etc. After acquiring the multiple training samples 31, the control unit 11 proceeds to the next step S102.

[0079] (Step S102) In step S102, the control unit 11 operates as the learning processing unit 112 and performs machine learning on the estimation model 40 using the acquired multiple training samples 31. In the machine learning process, the control unit 11 trains the estimation model 40 so that the estimated blood pressure value derived by the estimation model 40 from each training sample 31 matches the true value of the blood pressure when each training sample 31 was measured. In one example, the control unit 11 may adjust the values ​​of calculation parameters of the estimation model 40 so that the output (estimated blood pressure value) obtained from the estimation model 40 by providing the training sample 31 of each dataset 30 matches the true value indicated by the corresponding label 32. This machine learning can generate a trained estimation model 40 that has acquired the ability to estimate blood pressure values ​​from samples of a pulse wave while the cuff is inflated until the amplitude of the pulse wave reaches its maximum value.

[0080] In one example, the blood pressure to be estimated may include at least one of systolic blood pressure, diastolic blood pressure, and mean blood pressure. In another example, the estimation model 40 may be configured to estimate multiple types of blood pressure values. Accordingly, the true values ​​of each of the multiple types of blood pressure may be provided in machine learning. For example, the label 32 may be configured to indicate the true value of the systolic blood pressure as well as the true values ​​of other blood pressures. This makes it possible to generate a trained estimation model 40 capable of estimating the systolic blood pressure and other blood pressure values ​​from pulse wave samples obtained while the cuff is inflated until the amplitude of the pulse wave reaches its maximum value. The above-mentioned machine learning computation may be executed by the model generation device 1 or an external computer. That is, performing machine learning may include executing the machine learning computation within the model generation device 1 and issuing an instruction to an external computer to have the external computer execute the machine learning computation. When machine learning is complete, the control unit 11 proceeds to the next step S103.

[0081] (Step S103) In step S103, the control unit 11 operates as the output processing unit 113 to output the results of the machine learning. The output destination and the content of the information to be output may be selected appropriately depending on the embodiment.

[0082] In one example, the control unit 11 may generate learning result data 400 indicating the results of the machine learning as the output process and store the generated learning result data 400 in a predetermined storage area. The predetermined storage area may be, for example, RAM within the control unit 11, the storage unit 12, an external storage device, a storage medium, or a combination thereof. The storage medium may be, for example, a CD, a DVD, a semiconductor memory, or the like. The external storage device may be, for example, a data server such as a NAS. The external storage device may be, for example, an external storage device. When the machine learning calculation process is performed on an external computer, the learning result data 400 may be generated by the external computer. In another example, the control unit 11 may output the convergence result of the machine learning training as the output process. The output destination may be, for example, RAM within the control unit 11, the output device 15, an external computer, an external storage device, a storage medium, or a combination thereof. In yet another example, the control unit 11 may output the progress of the machine learning training as the output process. In this case, the process of step S103 may be executed together with step S102.

[0083] When the output of the machine learning results is completed, the control unit 11 ends the processing procedure of the model generation device 1 according to this operation example.

[0084] When the blood pressure measurement device 2 uses the trained estimation model 40 as the estimation model 41, the generated learning result data 400 may be provided to the blood pressure measurement device 2 from the model generation device 1 or an external computer at any timing and by any method. In one example, the learning result data 400 may be provided to the blood pressure measurement device 2 via a network, a storage medium, or the like. In another example, the learning result data 400 may be incorporated into the blood pressure measurement device 2 in advance.

[0085] Furthermore, the control unit 11 may update or newly generate the learning result data 400 by periodically or irregularly repeating the processes of steps S101 to S103. During this repetition, at least a portion of the multiple training samples 31 (multiple data sets 30) may be changed, modified, added, deleted, or the like as appropriate. Then, the control unit 11 may provide the updated or newly generated learning result data 400 to the blood pressure measurement device 2, thereby updating the model data 410 (learning result data 400) held by the blood pressure measurement device 2.

[0086] [Blood pressure measuring device] FIG. 9 is a flowchart showing an example of the processing procedure of the blood pressure measurement device 2 according to this embodiment. The following processing procedure is an example of an estimation method executed by a computer. However, the processing procedure of the blood pressure measurement device 2 is merely an example, and each step may be changed as much as possible. Furthermore, steps in the following processing procedure may be omitted, replaced, or added as appropriate depending on the embodiment.

[0087] (Step S201) In step S201, the control unit 21 operates as the acquisition unit 211 and acquires a target sample 51 of the pulse wave of the subject TU while the cuff 240 is inflated until the amplitude of the pulse wave reaches its maximum value.

[0088] In one example, when measuring blood pressure, the subject TU may wrap the cuff 240 around the measurement site. The subject TU may operate the operation switch 25 to start measuring blood pressure. In response to receiving an instruction to start blood pressure measurement via the operation switch 25, the control unit 21 may execute initialization processing, such as opening the exhaust valve 243 and setting the cuff pressure to atmospheric pressure (initial pressure). The control unit 21 may also control the pressure pump 242 to supply air to the air bag 2401, thereby increasing the cuff pressure. At the same time, the control unit 21 may detect the oscillation frequency of the input signal (output of the oscillation circuit 231) and convert the detected oscillation frequency into a pressure value signal. The control unit 21 may extract a pulse wave (pressure pulse wave) signal by applying a high-pass filter to the pressure value signal. Amplitude data can be obtained by plotting the extracted pulse wave signal. The pulse wave signal The control unit 21 may extract the maximum value of the pressure value signal and plot the extracted maximum value to obtain envelope data. The control unit 21 may also extract the cuff pressure signal by applying a low-pass filter to the pressure value signal. The cuff pressure data can be obtained by plotting the extracted cuff pressure signal.

[0089] In one example, the control unit 21 may acquire at least one of amplitude data and envelope data as the target sample 51. In another example, the control unit 21 may acquire the target sample 51 further including cuff pressure data. In another example, the control unit 21 may calculate a feature amount for each beat from the obtained pulse wave data, and acquire the target sample 51 further including the calculated feature amount for each beat. For example, the target sample 51 may further include at least one of RAV, WID, and DFN for each beat.

[0090] In one example, the target sample 51 may be obtained by actually measuring the entire pulse wave up to the maximum point at which the amplitude of the pulse wave reaches its maximum. That is, the target sample 51 may be composed of the actual measurement portion obtained by actually measuring the pulse wave up to the maximum point. In another example, the target sample 51 may be obtained by actually measuring a portion and predicting the remaining portion from the actual measurement portion. For example, the target sample 51 may be composed of an actual measurement portion obtained by actually measuring the pulse wave up to a characteristic time point of the pulse wave that appears before the amplitude of the pulse wave reaches its maximum, and a predicted portion obtained by predicting the pulse wave from the characteristic time point to the maximum point at which the amplitude of the pulse wave reaches its maximum, based on at least a portion of the actual measurement portion. In one example, the characteristic time point may be at least one of the diastolic blood pressure and a stable point. After the actual measurement portion of the target sample 51 is obtained, the control unit 21 may quickly stop inflating the cuff 240 by the pressure pump 242, open the exhaust valve 243, and release the air in the air bag 2401 to reduce the cuff pressure. In yet another example, the target sample 51 may be a sample of a pulse wave measured from the subject TU while the cuff is inflated up to a characteristic time point of a pulse wave that appears before the amplitude of the pulse wave reaches its maximum value. After acquiring the target sample 51, the control unit 21 proceeds to the next step S202.

[0091] (Step S202) In step S202, the control unit 21 operates as the estimation unit 212, and derives an estimated value 52 of the blood pressure of the subject TU from the acquired subject sample 51 using the estimation model 41.

[0092] In one example, the control unit 21 provides the acquired target sample 51 to the estimation model 41 and executes the calculation process of the estimation model 41. As a result of this calculation, the control unit 21 may obtain a derived result of an estimated blood pressure value 52 from the estimation model 41. In one example, the estimation model 41 may be the trained estimation model 40. In one example, the estimation model 41 may be configured to estimate multiple types of blood pressure values. In response to this, the control unit 21 may derive the estimated value 52 for each of the multiple types of blood pressure using the estimation model 41. After deriving the estimated blood pressure value 52, the control unit 21 proceeds to the next step S203.

[0093] (Step S203) In step S203, the control unit 21 operates as the output processing unit 213 and outputs information relating to the derived estimated value 52.

[0094] The output destination and the content of the information to be output may be selected as appropriate depending on the embodiment. In one example, the control unit 21 may display the derived blood pressure estimate 52 on the display unit 26 as the output process. When multiple types of blood pressure estimates have been derived, the control unit 21 may display at least some of the multiple types of blood pressure estimates on the display unit 26. In another example, the control unit 21 may store the derived blood pressure estimate 52 in the memory unit 22 as the output process. When multiple types of blood pressure estimates have been derived, the control unit 21 may store at least some of the multiple types of blood pressure estimates in the memory unit 22.

[0095] When the output of the information is completed, the control unit 21 ends the processing procedure of the blood pressure measurement device 2 according to this operation example. In response to the operation of the operation switch 25, the control unit 21 may execute a series of processes from step S201 to step S203.

[0096] [Features] In this embodiment, steps S101 and S102 described above can generate a trained estimation model 40 that has acquired the ability to estimate blood pressure values ​​from pulse wave samples up to the point at which the amplitude reaches its maximum. This trained estimation model 40 can be expected to lower the cuff pressure in accordance with the subject, and to perform stable blood pressure measurement. Furthermore, steps S201 and S202 described above can be expected to lower the cuff pressure in accordance with the subject TU, and to perform stable blood pressure measurement.

[0097] §4 Variations Although the embodiments of the present invention have been described in detail above, the above description is merely an example of the present invention in every respect. The processes and means described in this disclosure can be freely combined and implemented as long as no technical contradiction occurs. Furthermore, various improvements or modifications may be made to the above embodiments as appropriate. For example, the following modifications are possible. Note that, in the following, the same reference numerals are used for components similar to those in the above embodiments, and descriptions of the same points as those in the above embodiments are omitted as appropriate.

[0098] <4.1> In the above embodiment, the blood pressure measurement device 2 is an example of an estimation device. However, the form of the estimation device is not limited to this example and may be changed as appropriate depending on the embodiment. In another example, the estimation device may be configured by a computer prepared separately from the blood pressure measurement device.

[0099] 10 schematically shows an example of an application scene of an estimation device 7 according to a modified example. The estimation device 7 according to this modified example does not include a pulse wave measurement unit such as the cuff 240 of the blood pressure measurement device 2, and is configured to acquire a target sample 51 directly from a blood pressure measurement device 700 or indirectly via a network, a storage medium, another computer, or the like. Except for this point, the estimation device 7 may be configured similarly to the blood pressure measurement device 2. Note that the blood pressure measurement device 700 may be configured similarly to the blood pressure measurement device 2, or may be configured similarly to a general sphygmomanometer.

[0100] 11 schematically shows an example of the hardware configuration of the estimation device 7 according to this modification. The estimation device 7 according to this modification is a computer in which a control unit 71, a storage unit 72, a communication interface 73, an input device 74, an output device 75, and a drive 76 are electrically connected.

[0101] The control unit 71 to the drive 76 and the storage medium 97 of the estimation device 7 may be configured similarly to the control unit 11 to the drive 16 and the storage medium 91 of the model generation device 1. The control unit 71 (CPU) is an example of a processor resource of the estimation device 7, and the storage unit 72 (and RAM, ROM) is an example of a memory resource of the estimation device 7. In this modification, the storage unit 72 stores various information such as an estimation program 87 and model data 410.

[0102] The estimation program 87 may be configured to include instructions similar to those of the estimation program 82, except that the target sample 51 is acquired from the blood pressure measurement device 700. At least one of the estimation program 87 and the model data 410 may be stored in a storage medium 97 instead of or together with the storage unit 72. The estimation device 7 may acquire at least one of the estimation program 87 and the model data 410 from the storage medium 97.

[0103] The estimation device 7 may perform data communication with other computers (e.g., the model generating device 1 and the blood pressure measuring device 700) via the communication interface 73. An operator can operate the estimation device 7 by using the input device 74 and the output device 75.

[0104] Note that, with regard to the specific hardware configuration of the estimation device 7, components may be omitted, replaced, or added as appropriate depending on the embodiment. For example, the control unit 71 may include multiple hardware processors. The hardware processor may be configured with a microprocessor, FPGA, DSP, GPU, ASIC, etc. At least one of the communication interface 73, the input device 74, the output device 75, and the drive 76 may be omitted. The estimation device 7 may be provided with an external interface and may be connected to the blood pressure measurement device 700 via the external interface. The estimation device 7 may be configured with multiple computers. In this case, the hardware configurations of the computers may or may not be the same. The estimation device 7 may be an information processing device designed specifically for the service provided, as well as a general-purpose server device, a general-purpose PC, a tablet PC, a terminal device (smartphone, etc.), etc.

[0105] The software configuration of the estimation device 7 may be similar to that of the blood pressure measurement device 2. As a result, the estimation device 7 may execute a process for estimating blood pressure using the same processing procedure as the blood pressure measurement device 2. In step S201, the control unit 71 may acquire a target sample 51 of the subject TU directly or indirectly from the blood pressure measurement device 700. In one example, part of the calculation process for acquiring the target sample 51 may be executed by the blood pressure measurement device 700 or another computer, and the remaining calculation process may be executed by the estimation device 7. In another example, the calculation process for acquiring the target sample 51 may be executed by the blood pressure measurement device 700, another computer, or the estimation device 7. In step S202, the control unit 71 may derive an estimated value 52 of the blood pressure of the subject TU from the acquired target sample 51 using the estimation model 41, as in the above embodiment. In step S203, the control unit 71 outputs information related to the derived estimated value 52. In one example, the control unit 71 may output the derived estimated value 52 to at least one of the RAM, the storage unit 72, the output device 75, and another computer. The output destination and the content of the information to be output may be selected appropriately depending on the embodiment. In the estimation device 7 according to this modification, the cuff pressure can be lowered in accordance with the subject TU, and stable blood pressure measurement can be expected. [Explanation of symbols]

[0106] 1...Model generation device, 11...control unit, 12...storage unit, 81...generation program, 91...storage medium, 111...acquisition unit, 112...learning processing unit, 113...output processing unit, 2...blood pressure measuring device, 21...control unit, 22...storage unit, 82... Estimated Program, 31...training samples, 40·41…Estimation model, 51...target sample, 52...estimated value

Claims

1. 1. A computer-implemented method for generating a model, comprising: The model generation method includes: obtaining a plurality of training samples of pulse waves measured from the subject while inflating the cuff until the amplitude of the pulse wave reaches a maximum; performing machine learning of an estimation model using the obtained plurality of training samples; and outputting the results of the machine learning; Equipped with performing the machine learning includes training the estimation model so that an estimated value of blood pressure derived from each of the training samples by the estimation model matches a true value of the blood pressure at the time when each of the training samples was measured; Model generation method.

2. Each of the training samples includes at least one of envelope data and amplitude data of the pulse wave. The model generation method of claim 1 .

3. Each of the training samples further includes a feature value for each beat of the pulse wave. The model generation method according to claim 2 .

4. An estimation program for causing a computer to execute an estimation method, The estimation method includes: obtaining a subject sample of a pulse wave in the subject while inflating the cuff until the amplitude of the pulse wave reaches a maximum; deriving an estimate of the subject's blood pressure from the acquired subject samples using an estimation model; and outputting information about the derived estimate; Including, Estimation program.

5. the target sample is constituted by an actually measured portion obtained by actually measuring the pulse wave until the amplitude of the pulse wave reaches a maximum value; The estimation program according to claim 4 .

6. The subject sample is an actually measured portion obtained by actually measuring the pulse wave up to a characteristic time point of the pulse wave that appears before the amplitude of the pulse wave reaches a maximum value; and a predicted portion obtained by predicting the pulse wave from the characteristic time point to a maximum time point at which the amplitude of the pulse wave is maximum, based on at least a part of the actually measured portion; It consists of The estimation program according to claim 4 .

7. The target sample includes at least one of envelope data and amplitude data of the pulse wave. The estimation program according to claim 4 .

8. The target sample further includes a feature amount for each beat of the pulse wave. The estimation program according to claim 7 .

9. An estimation device including a control unit, The control unit obtaining a subject sample of a pulse wave in the subject while inflating the cuff until the amplitude of the pulse wave reaches a maximum; deriving an estimate of the subject's blood pressure from the acquired subject samples using an estimation model; and outputting information about the derived estimate; configured to perform Estimation device.

Images