Blood pressure estimation system, blood pressure estimation device, blood pressure estimation method, and blood pressure estimation program

Abstract

This invention provides a blood pressure estimation system that allows for easy and non-invasive blood pressure estimation. [Solution] The blood pressure estimation system 1 according to the present invention comprises a sensor device 11 having a plurality of sheet-type piezoelectric sensors 11a installed at a specific distance from each other to detect biological information including cardiac pulse at each location, and a control device 11b that converts the biological information detected by these sheet-type piezoelectric sensors 11a into biological signal waveform data and transmits it, and a server device 12 having a control unit that estimates blood pressure based on the biological signal waveform data received from the sensor device 11, and a display unit that displays the blood pressure estimation result by the control unit. The control unit estimates the systolic blood pressure P at a specific posture based on the prediction result of the posture of the subject of blood pressure estimation using a learned posture prediction model, and the pulse wave propagation time ΔT calculated from the cardiac pulse waveforms at two locations on the subject's body. max We estimate this.

Description

Technical Field

[0001] The present invention relates to a blood pressure estimation system, a blood pressure estimation device, a blood pressure estimation method, and a blood pressure estimation program for estimating blood pressure in a non-invasive and non-restrictive manner.

Background Art

[0002] Conventionally, the oscillometric method has been used for blood pressure measurement. However, since the oscillometric method measures blood pressure by applying pressure to the upper arm or wrist with a cuff to temporarily block blood flow, the burden on the subject is large. Therefore, in order to continuously measure blood pressure fluctuations for a long time, a blood pressure measurement method with a lower load that does not require pressurization with a cuff is desired.

[0003] As a blood pressure measurement method that does not require pressurization with a cuff, there is a method of calculating from the pulse transit time (PTT: Pulse Transit Time) (see Non-Patent Document 1). According to this method, the systolic blood pressure can be calculated, for example, using the pulse transit time, which is the time difference between the peaks of two pulse waves measured by an electrocardiograph (chest) and a blood oxygen concentration meter (finger).

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the method of calculating systolic blood pressure using the pulse transit time of two pulse waves measured by an electrocardiograph (chest) and a blood oxygen concentration meter (finger), it is necessary to wear the measuring device, and it is difficult to simply and continuously measure blood pressure at home for a long time.

[0006] The present invention has been made in view of the above-mentioned problems, and aims to provide a blood pressure estimation system, blood pressure estimation device, blood pressure estimation method, and blood pressure estimation program that can easily estimate blood pressure in a non-invasive and non-restrictive manner without wearing measuring equipment. [Means for solving the problem]

[0007] The blood pressure estimation system according to the present invention comprises a sensor device having a plurality of sheet-type sensors installed at a specific distance from each other to detect biological information including cardiac movement at each position, and a control device that converts the biological information detected by these sheet-type sensors into biological signal waveform data and transmits it, and a server device having a control unit that estimates blood pressure based on the biological signal waveform data received from the sensor device, and a display unit that displays the blood pressure estimation result by the control unit.

[0008] Alternatively, the blood pressure estimation system according to the present invention may include a sensor device comprising a plurality of sheet-type sensors installed at a specific distance from each other to detect biological information including cardiac resilience at each position, and a control device having a control unit that converts the biological information detected by these sheet-type sensors into biological signal waveform data, estimates blood pressure based on the biological signal waveform data, and transmits the estimation result, and a server device that displays the blood pressure estimation result received from the sensor device.

[0009] Furthermore, the control unit of the server device, or the control unit of the control device in the sensor device, is characterized by comprising: a filtering processing unit that acquires biological waveforms including cardiac pulse waveforms from biological signal waveform data; a posture prediction unit that takes biological waveforms corresponding to a predetermined number of subjects detected while changing their posture as input data, and predicts the posture of a subject by inputting the biological waveform of the subject to be blood pressure estimation into a trained posture prediction model that has been machine-learned using the posture of the subject at the time of detection of the biological information corresponding to these biological waveforms as training data; and a blood pressure estimation unit that estimates the systolic blood pressure at a specific posture based on the prediction result of the subject's posture and the pulse wave propagation time calculated from cardiac pulse waveforms at two locations on the subject's body.

[0010] Furthermore, the blood pressure estimation system according to the present invention preferably uses a supine position as the specific posture, preferably uses two sheet-type sensors, and preferably includes heart sounds and / or respiration as the biological information.

[0011] The blood pressure estimation device according to the present invention includes a control unit that receives biological information, including cardiac pulse, detected by a plurality of sheet-type sensors installed at a specific distance from each other, as biological signal waveform data, and estimates blood pressure based on the biological signal waveform data, and a display unit that displays the blood pressure estimation result by the control unit, wherein the control unit includes a filtering processing unit that acquires a biological waveform including cardiac pulse waveform from the biological signal waveform data, a posture prediction unit that takes biological waveforms corresponding to a predetermined number of subjects detected while changing their posture as input data, and predicts the posture of a subject by inputting the biological waveform of the subject to be blood pressure estimation into a trained posture prediction model that has been machine-learned using the posture of the subject at the time of detection of the biological information corresponding to these biological waveforms as training data, and blood pressure estimation unit that estimates the systolic blood pressure at a specific posture based on the prediction result of the subject's posture and the pulse wave propagation time calculated from cardiac pulse waveforms at two locations on the subject's body.

[0012] Furthermore, the blood pressure estimation device according to the present invention comprises: a plurality of sheet-type sensors installed while maintaining a specific distance and detecting biological information including cardiac pulse at each position; a control device having a control unit that converts the biological information detected by the plurality of sheet-type sensors into biological signal waveform data, estimates blood pressure based on the biological signal waveform data, and transmits the estimation result, wherein the control unit comprises: a filtering processing unit that acquires a biological waveform including cardiac pulse waveform from the biological signal waveform data; a posture prediction unit that takes biological waveforms corresponding to the biological information of a predetermined number of subjects detected while changing their posture as input data, and predicts the posture of a subject by inputting the biological waveform of the subject to be blood pressure estimation into a trained posture prediction model that has been machine-learned using the posture of the subject at the time of detection of the biological information corresponding to these biological waveforms as training data; and a blood pressure estimation unit that estimates the systolic blood pressure at a specific posture based on the prediction result of the subject's posture and the pulse wave propagation time calculated from cardiac pulse waveforms at two locations on the subject's body.

[0013] The blood pressure estimation method according to the present invention is a blood pressure estimation method by a control unit of a blood pressure estimation device that receives biological information, including cardiac pulse, detected by a plurality of sheet-type sensors installed while maintaining a specific distance, as biological signal waveform data, wherein the control unit performs a filtering step of acquiring a biological waveform including cardiac pulse waveform from the biological signal waveform data; a posture prediction step of predicting the posture of a subject by inputting the biological waveform of the subject to be blood pressure estimation into a trained posture prediction model that has been machine-learned using biological waveforms corresponding to the biological information of a predetermined number of subjects detected while changing their posture as input data and the posture of the subjects at the time of detection of the biological information corresponding to these biological waveforms as training data; and a blood pressure estimation step of estimating the systolic blood pressure at a specific posture based on the prediction result of the subject's posture and the pulse wave propagation time calculated from cardiac pulse waveforms at two locations on the subject's body.

[0014] The blood pressure estimation program according to the present invention is a blood pressure estimation program for estimating the blood pressure of a subject, in which the control unit of a blood pressure estimation device that has received biological information including cardiac elasticity detected by a sheet-type sensor as biological signal waveform data causes a computer acting as the control unit of the blood pressure estimation device to execute a series of processes according to each of the steps described above.

[0015] Therefore, according to the blood pressure estimation system, blood pressure estimation device, blood pressure estimation method, and blood pressure estimation program of the present invention, blood pressure can be easily estimated in a non-invasive and non-restrictive manner without wearing measuring devices such as an electrocardiograph or a blood oxygen saturator. [Effects of the Invention]

[0016] According to the blood pressure estimation system, blood pressure estimation device, blood pressure estimation method, and blood pressure estimation program of the present invention, blood pressure can be easily estimated in a non-invasive and non-restrictive manner without wearing any measuring equipment. [Brief explanation of the drawing]

[0017] [Figure 1] Figure 1 shows an example of the configuration of the blood pressure estimation system according to the present invention, where (a) is a diagram showing an example of the block configuration, (b) is a diagram schematically showing the hardware configuration, and (c) is a top view schematically showing the blood pressure estimation process. [Figure 2] Figure 2 shows an example of the hardware configuration of a computer operating as a server device (blood pressure estimation device). [Figure 3] Figure 3 is a functional block diagram showing the functions of the control unit that performs the learning model generation process. [Figure 4] Figure 4 is a flowchart showing an example of the learning model generation process. [Figure 5] Figure 5 shows examples of biosignal waveforms, heart sound waveforms, cardiac elasticity waveforms, and respiratory waveforms. [Figure 6] Figure 6 shows an example of a dataset used to generate a posture prediction model. [Figure 7] FIG. 7 is a functional block diagram showing the functions of a control unit that performs blood pressure estimation processing (blood pressure estimation program). [Figure 8] FIG. 8 is a flowchart showing an example of blood pressure estimation processing. [Figure 9] FIG. 9 is a diagram showing an example of a cardiac ballistocardiogram waveform in the supine position.

Embodiments for Carrying Out the Invention

[0018] Hereinafter, embodiments of a blood pressure estimation system, a blood pressure estimation device, a blood pressure estimation method, and a blood pressure estimation program according to the present invention will be described in detail based on the drawings. Note that the present invention is not limited by this embodiment. Also, in the specification and drawings of the present application, for elements that can be described in the same way, the same reference numerals may be used to omit redundant descriptions.

[0019] <System Configuration> FIG. 1 is a diagram showing a configuration example of a blood pressure estimation system according to the present invention, (a) is a diagram showing an example of a block configuration, (b) is a diagram schematically showing a hardware configuration, and (c) is a top view schematically showing the state of blood pressure estimation.

[0020] In FIG. 1, a blood pressure estimation system 1 of the present embodiment includes a plurality of sheet-type piezoelectric sensors 11a that detect biological information (information such as heart sound, cardiac ballistocardiogram, and respiration) of a subject for blood pressure estimation (hereinafter simply referred to as a subject), and a control device 11b that converts the biological information detected by these sheet-type piezoelectric sensors 11a into waveform data of a biological signal (hereinafter referred to as biological signal waveform data) and transmits it by wireless communication (for example, WIFI, Bluetooth, etc.) or wired communication (for example, wired connection by a USB cable, LAN cable, etc.). The system also includes a server device 12 that functions as a blood pressure estimation device for estimating blood pressure based on the biological signal waveform data of the subject received from the sensor device 11.

[0021] The multiple sheet-type piezoelectric sensors 11a used in this embodiment can be installed in locations where the subject or the subject involved in the generation of the learning model detects biological information, such as on a bed or mattress (see Figures 1(b) and (c)). In this embodiment, in order to detect biological information at two locations on the body, as an example, two sheet-type piezoelectric sensors 11a are installed at a certain distance from each other. Specifically, when the subject or subject is in a supine position, one sheet-type piezoelectric sensor 11a is installed at the chest (hereinafter referred to as the first position), and the other sheet-type piezoelectric sensor 11a is installed at the calf (hereinafter referred to as the second position) (see Figure 1(b)). Figure 1(c) shows an example where the distance between the first position and the second position is 70 cm.

[0022] These sheet-type piezoelectric sensors 11a can be any sheet-type sensor capable of detecting biological information such as heart sounds, heartbeats, and respiration, and there are no particular limitations on the size, material, or detection method of the sheet.

[0023] Furthermore, in this embodiment, the sheet-type piezoelectric sensor 11a is used to detect biological information such as heart sounds, heartbeats, and respiration, but it is not limited to this. For example, the biological information only needs to include information on heartbeats (pulse waves) necessary for blood pressure estimation, and the sheet-type piezoelectric sensor 11a may also be capable of detecting biological information other than heart sounds, heartbeats, and respiration.

[0024] Furthermore, although Figure 1 shows two sheet-type piezoelectric sensors 11a installed at a certain distance from each other, this is not the only option. For example, a sheet-type sensor with multiple sensors (e.g., two) incorporated at a distance from each other could be used. This would allow the distance between the two sheet-type piezoelectric sensors 11a to remain constant at all times.

[0025] Furthermore, although sheet-type piezoelectric sensors 11a are installed at the first position (chest) and the second position (calf) in Figure 1, the installation location is not limited to these positions; any location where a pulse wave can be detected is acceptable, such as the buttocks or head (pillow). As long as a distance sufficient for calculating the pulse wave propagation time, as described later, is ensured, the installation location of the sheet-type piezoelectric sensors 11a is arbitrary.

[0026] <Server configuration> Figure 2 shows an example of the hardware configuration of a computer operating as a server device 12 (blood pressure estimation device). In this embodiment, the server device 12 operates as a host computer that performs the process of estimating blood pressure using the subject's biosignal waveform data received from the sensor device 11 (hereinafter referred to as the blood pressure estimation process), and the process of generating a learning model for predicting the subject's posture (hereinafter referred to as the learning model generation process).

[0027] In Figure 2, the server device 12 comprises a control unit 21 consisting of a CPU (Central Processing Unit) and an FPGA (Field Programmable Gate Array), a storage unit 22 including various memories, an input unit 23 including a user interface such as a keyboard and mouse, an interface (I / F) unit 24 that performs input / output processing such as printing and scanning, a display unit 25 which is a display, and a communication unit 26 that communicates with the outside via a predetermined network. In Figure 2, the server device 12 is shown to include an input unit 23 including a user interface such as a keyboard and mouse, but the server device 12 of this embodiment is not limited to this, and the display unit 25 may be equipped with a touch panel function so that the input unit 23 is not provided, or it may be configured to be used in combination with the input unit 23.

[0028] In Figure 2, the control unit 21 executes, for example, a blood pressure estimation program that estimates the blood pressure of a subject when it receives biosignal waveform data of that subject, and a learning model generation program that generates a learning model (AI model) to predict the posture of the subject, in order to realize the blood pressure estimation process and learning model generation process by the server device 12. The storage unit 22 stores the programs related to the blood pressure estimation process and learning model generation process of this embodiment (blood pressure estimation program, learning model generation program) and various information (such as biosignal waveform data), as well as various data obtained during the processing (such as biosignal waveforms, datasets for machine learning, posture prediction results, and blood pressure estimation results, which will be described later). The control unit 21 executes the blood pressure estimation process and learning model generation process of this embodiment by reading the various programs stored in the storage unit 22.

[0029] Furthermore, the storage unit 22 is not limited to internal memory, but may also be an external storage medium such as a DVD (Digital Versatile Disc), SD memory, or external HD, or even the cloud. Alternatively, it may consist of both internal memory and external storage media or the cloud. Also, for the sake of explanation, the hardware configuration of the server device 12 in this embodiment is listed only for the blood pressure estimation processing and learning model generation processing in this embodiment, and does not represent all the functions of the computer that constitutes the server device 12.

[0030] Furthermore, while the server device 12 is envisioned as a general-purpose PC such as a desktop or notebook computer, it is not limited to these, and may also be a mobile device such as a smartphone or tablet.

[0031] <Learning model generation process> Next, before explaining the blood pressure estimation process using the blood pressure estimation system 1 of this embodiment, we will explain the learning model generation process that is a prerequisite for it. Figure 3 is a functional block diagram showing the functions of the control unit 21 that performs the learning model generation process, and Figure 4 is a flowchart showing an example of the learning model generation process.

[0032] In Figure 3, the control unit 21 includes a filtering processing unit 31 that acquires heart sound waveforms, cardiac pulse waveforms, and respiratory waveforms (hereinafter, heart sound waveforms, cardiac pulse waveforms, and respiratory waveforms may be referred to as biological waveforms) from biological signal waveform data, a dataset generation unit 32 that generates a dataset for machine learning, and a machine learning unit 33 that generates a learning model using a machine learning algorithm. In this embodiment, a posture prediction model is generated as the learning model, which predicts the posture of a subject when the subject's biological waveform is input. This learning model is generated using a known machine learning algorithm, such as a neural network.

[0033] <<Generation of posture prediction model>> Here, the process of generating a posture prediction model using the flowchart shown in Figure 4 will be explained. In this embodiment, as a preprocessing step for generating the posture prediction model, a detection process for biological information (information such as heart sounds, heartbeats, and respiration) is performed on a predetermined number of subjects using a sheet-type piezoelectric sensor 11a installed at a first position. At this time, for example, the biological information detection process is performed for each subject while changing their posture (right lateral recumbent position, supine position, left lateral recumbent position, sitting position, etc.) in 5-minute intervals. The control device 11b of the sensor device 11 converts the detected biological information into biological signal waveform data and transmits this biological signal waveform data to the server device 12 in association with the subject's posture at the time of biological information detection. The server device 12 then stores the biological signal waveform data for each subject received by the communication unit 26 in the storage unit 22 in advance, in association with the subject's posture at the time of biological information detection.

[0034] In this embodiment, the preprocessing for generating the attitude prediction model is performed, for example, using a sheet-type piezoelectric sensor 11a installed at a first position. That is, in this preprocessing, the sheet-type piezoelectric sensor 11a installed at a second position is turned OFF. Alternatively, the sheet-type piezoelectric sensor 11a installed at the first position may be turned OFF, and the attitude prediction model may be generated using the sheet-type piezoelectric sensor 11a installed at the second position.

[0035] After the preprocessing for generating the posture prediction model is completed, if the operator instructs the generation of the posture prediction model by operating the input unit 23, the filtering processing unit 31 of the control unit 21 reads all the biosignal waveform data from the storage unit 22 and extracts biosignal waveforms (heart sound waveform, heartbeat waveform, and respiration waveform) from the biosignal waveform (synthesized waveform) obtained from the biosignal waveform data (Figure 4, step S1). Specifically, the filtering processing unit 31 obtains heart sound waveforms, heartbeat waveforms, and respiration waveforms by extracting frequency components specific to heart sound (10Hz~500Hz), heartbeat (1Hz~20Hz), and respiration (0.05Hz~0.5Hz), respectively, from the biosignal waveform. Figure 5 shows an example of biosignal waveforms, heart sound waveforms, heartbeat waveforms, and respiration waveforms. The filtering processing unit 31 performs filtering on all of each subject's biosignal waveforms and stores the extracted subject's biosignal waveforms in the storage unit 22 in association with the subject's posture.

[0036] Next, in the control unit 21, the dataset generation unit 32 reads all of the subject's bio-waveforms from the storage unit 22, associates all of the bio-waveforms with the subject's posture (right lateral decubitus, supine, left lateral decubitus, or sitting) that is individually associated with those bio-waveforms, and generates a dataset for machine learning (step S2). The dataset generation unit 32 then stores the generated dataset in the storage unit 22. In this embodiment, as an example, four postures are used for the subject, but the type of posture is not limited to the above four, as long as the bio-information (heart sounds, cardiac rhythm, respiration, etc.) can be classified for each posture.

[0037] In other words, the dataset used in the machine learning algorithm is one in which the corresponding training data is individually linked to each input data, with the corresponding training data being the bio-waveforms (heart sound waveform, cardiac pulse waveform, and respiratory waveform) of a predetermined number of subjects detected while changing their posture as explanatory variables (input data), and the subject's posture (right lateral decubitus, supine, left lateral decubitus, or sitting) at the time the bio-information corresponding to these bio-waveforms was detected. Figure 6 shows an example of a dataset for generating a posture prediction model. Here, each posture taken by a specific subject (training data) and one of the bio-waveforms of that subject corresponding to each posture, the heart sound waveform (input data), are shown as an example. A dataset combining such training data and input data is prepared for each subject whose bio-information has been detected.

[0038] Subsequently, in the control unit 21, the machine learning unit 33 reads the dataset from the memory unit 22 and, for example, uses a neural network, which is one of the machine learning algorithms, to perform supervised learning on the dataset, thereby generating a posture prediction model that predicts the posture of a subject when the subject's biological waveform is input (step S3).

[0039] In other words, in this embodiment, a posture prediction model is obtained that predicts the posture of a subject when the subject's biological waveform is input, by using a machine learning algorithm to learn the correlation between the input data (explanatory variables) and the training data (dependent variable) described above.

[0040] There are no particular restrictions on the number of subjects whose biometric information is detected (corresponding to the specified number above), but the number should be such that the correlation between biometric waveforms and posture can be sufficiently learned, taking into account individual differences (height, weight, gender, age, external environment (such as differences in beds, etc., the environment when biometric information is detected)).

[0041] <Machine learning algorithms> In this embodiment, a neural network is used as the machine learning algorithm, and the machine learning unit 33 receives the input data (explanatory variables) and training data (target variable) as described above. The machine learning unit 33 then calculates the error between the output, which is the pose prediction result, and the training data, and performs repeated learning on the dataset over a predetermined number of epochs to minimize this error. In this way, the pose prediction model is trained.

[0042] In this embodiment, a known neural network was used as an example of a machine learning algorithm to generate the above-described pose prediction model. However, the machine learning algorithm used to generate this model is not limited to this. For example, other known machine learning algorithms such as random forests or boosted decision trees can also be used.

[0043] <Blood pressure estimation system> Next, the blood pressure estimation process using the blood pressure estimation system 1 of this embodiment will be described. Figure 7 is a functional block diagram showing the functions of the control unit 21 that performs the blood pressure estimation process (blood pressure estimation program), and Figure 8 is a flowchart showing an example of the blood pressure estimation process. As mentioned above, in this embodiment, two sheet-type piezoelectric sensors 11a are installed at a certain distance from each other so that biological information (cardiac pulsation) from two locations on the body can be detected. Specifically, as shown in Figure 1, one sheet-type piezoelectric sensor 11a is installed at the first position, and the other sheet-type piezoelectric sensor 11a is installed at the second position.

[0044] In Figure 7, the control unit 21 includes a filtering processing unit 41 having the same function as the filtering processing unit 31 described above, a posture prediction unit 42 that functions as a posture prediction model, and a blood pressure estimation unit 43 that estimates the systolic blood pressure at a specific posture based on the prediction result of the subject's posture and the cardiac elasticity waveforms at two locations, a first position and a second position.

[0045] In this embodiment, the sensor device 11 shown in Figure 1 converts the subject's biological information (heart sounds, heartbeats, respiration, etc.) detected by sheet-type piezoelectric sensors 11a installed at two locations, a first position and a second position, into biological signal waveform data and transmits it to the server device 12. Then, in the server device 12, the control unit 21 predicts the subject's posture using a trained posture prediction model, and further estimates, for example, the systolic blood pressure during the supine position based on the predicted posture of the subject and the pulse wave propagation time obtained from the heartbeat waveforms at the first and second positions, which are based on the subject's biological signal waveform data.

[0046] In this process, the control unit 21 of the server device 12 continuously performs blood pressure estimation processing until the subject finishes detecting biological information using the sheet-type piezoelectric sensor 11a.

[0047] The blood pressure estimation process of this embodiment will be described in more detail below with reference to Figure 8.

[0048] In Figure 8, the control device 11b, which constitutes the sensor device 11, converts the subject's biological information (information such as heart sounds, heartbeats, and respiration) detected by the sheet-type piezoelectric sensors 11a at two locations (the first and second locations) into biological signal waveform data, and transmits this biological signal waveform data to the server device 12, for example, using wireless communication (step S11). In this embodiment, using the blood pressure estimation system 1 shown in Figure 1, the detection of biological information by the two sheet-type piezoelectric sensors 11a is continuously performed for a subject who is in a predetermined posture (for example, lying on their right side, on their back, on their left side, sitting, etc.) at a location where biological information is detected (for example, a bed or mattress). That is, each sheet-type piezoelectric sensor 11a constantly performs the detection process of the subject's biological information during the period when the power is ON.

[0049] The server device 12 continuously receives biosignal waveform data corresponding to a first position and biosignal waveform data corresponding to a second position via the communication unit 26 (step S12), and stores this biosignal waveform data in the storage unit 22 under the control of the control unit 21.

[0050] Then, the filtering processing unit 41 of the control unit 21 reads biosignal waveform data corresponding to the first position and biosignal waveform data corresponding to the second position from the storage unit 22, and extracts the subject's biosignal waveforms (heart sound waveform, cardiac pulse waveform, respiratory waveform) from the biosignal waveform (synthesized waveform) obtained based on each of the biosignal waveform data (step S13). Specifically, the subject's heart sound waveform, cardiac pulse waveform, and respiratory waveform are extracted from the biosignal waveform at the first position, the subject's cardiac pulse waveform is extracted from the biosignal waveform at the second position, and the extracted biosignal waveforms (see Figure 5) are stored in the storage unit 22.

[0051] Next, in the control unit 21, the posture prediction unit 42 reads the biological waveforms (heart sound waveform, cardiac pulse waveform, respiratory waveform) of the first position from the memory unit 22 and predicts the subject's posture using the posture prediction model (step S14). The posture prediction unit 42 then stores the subject's posture, which is the prediction result from the posture prediction model, in the memory unit 22.

[0052] Specifically, the trained posture prediction model takes the subject's bio-waveforms (heart sound waveform, cardiac pulse waveform, and respiratory waveform) as input and outputs the subject's posture (one of the following: right lateral decubitus, supine, left lateral decubitus, or sitting) corresponding to those bio-waveforms as the prediction result. This makes it possible to understand the transitions in posture (changes in posture) of a subject while bio-information (information such as heart sound, cardiac pulse, and respiration) is continuously detected, in a time-series manner. The posture prediction unit 42 then stores the posture prediction results in the memory unit 22, associating them with the cardiac pulse waveforms of the subject's first and second positions.

[0053] Next, in the control unit 21, the blood pressure estimation unit 43 reads the cardiac elasticity waveforms of the subject at a first position and a second position from the memory unit 22, and extracts only the cardiac elasticity waveforms of the first position and the second position corresponding to a predetermined specific posture from the read cardiac elasticity waveforms (step S15). In this embodiment, as an example, only the cardiac elasticity waveform when the subject is in a supine position is extracted. Figure 9 shows an example of a cardiac elasticity waveform when the subject is in a supine position. The waveform at the top of the figure is the cardiac elasticity waveform at the first position (chest position) when the subject is in a supine position, and the waveform at the bottom of the figure is the cardiac elasticity waveform at the second position (calf position) when the subject is in a supine position.

[0054] Next, the blood pressure estimation unit 43 calculates the pulse wave propagation time ΔT (see Figure 9), which is the time difference between the peak of the cardiac elasticity waveform at the first position (chest position) (referred to as the first peak) and the peak of the cardiac elasticity waveform at the second position (calf position) immediately following it (the second peak), for all combinations of the first and second peaks in the cardiac elasticity waveform extracted in step S15 (step S16).

[0055] Then, the blood pressure estimation unit 43 calculates the systolic blood pressure P based on the calculated pulse wave propagation time ΔT. max Estimate the systolic blood pressure P for all pulse wave propagation times ΔT calculated in step S16, according to equation (1) below. max The calculation is performed. These calculation results are then stored in the memory unit 22, and the systolic blood pressure P is calculated. max The transition is displayed on the display unit 25 (step S18). Systolic blood pressure P max As described in Non-Patent Document 1, it can be expressed as shown in equation (1) using the pulse wave propagation time ΔT and coefficients a and b.

[0056] Systolic blood pressure P max (mmHg) = a ln ΔT + b …(1)

[0057] In this embodiment, the sheet-type piezoelectric sensor 11a at the second position may detect the cardioresonance of both calves and take the average, or it may use the cardioresonance of either one of the calves.

[0058] Furthermore, although this embodiment describes a case in which biological information is detected using two sheet-type piezoelectric sensors 11a installed at a first position and a second position, it is not limited to this. For example, three or more sheet-type piezoelectric sensors 11a can be placed in various locations to detect biological information from each, and the systolic blood pressure P in a supine position can be determined based on the pulse wave propagation time ΔT between each sheet-type piezoelectric sensor 11a. max It may also be possible to estimate this. This will allow for a more accurate estimation of the subject's blood pressure.

[0059] <Effects and Others> As described above, the blood pressure estimation system 1 of this embodiment comprises a sensor device 11 having a plurality of sheet-type piezoelectric sensors 11a installed at a specific distance from each other to detect biological information including cardiac elasticity at each position, and a control device 11b that converts the biological information detected by these sheet-type piezoelectric sensors 11a into biological signal waveform data and transmits it, and a server device 12 having a control unit 21 that estimates blood pressure based on the biological signal waveform data received from the sensor device 11, and a display unit 25 that displays the blood pressure estimation result by the control unit 21.

[0060] The control unit 21 of the server device 12 includes a filtering processing unit 41 that acquires bio-waveforms including cardiac pulse waveforms from bio-signal waveform data, a posture prediction unit 42 that uses bio-waveforms corresponding to a predetermined number of subjects detected while changing their posture as input data, and uses the subjects' postures at the time of bio-information detection corresponding to these bio-waveforms as training data to input the bio-waveforms of the subject for blood pressure estimation into a trained posture prediction model, and predicts the posture of the subject by inputting the bio-waveforms of the subject for blood pressure estimation into the trained posture prediction model, and calculates the systolic blood pressure P for a specific posture based on the prediction result of the subject's posture and the pulse wave propagation time ΔT calculated from cardiac pulse waveforms at two locations on the subject's body. max It is characterized by having a blood pressure estimation unit 43 that estimates the blood pressure.

[0061] Therefore, according to the blood pressure estimation system 1 of this embodiment, blood pressure can be easily estimated in a non-invasive and non-restrictive manner without the need to wear measuring devices such as an electrocardiograph or a blood oxygen saturator. For example, if blood pressure can be estimated using a sheet-type sensor, blood pressure can be monitored over a long period of time without stress. Furthermore, as in this embodiment, if the posture of the subject can be predicted and blood pressure can be monitored at all times while maintaining a constant posture, the accuracy of disease analysis can be greatly improved.

[0062] In this embodiment, the blood pressure estimation process and the learning model generation process (posture prediction model generation process) are performed on the server device 12. However, the system is not limited to this, and each process may be performed on a separate device (such as a personal computer or smartphone).

[0063] Furthermore, in this embodiment, the blood pressure estimation process is realized by the control unit 21 of the server device 12 executing a blood pressure estimation program (processing by the filtering processing unit 41, posture prediction unit 42, and blood pressure estimation unit 43), but it is not limited to this. For example, this blood pressure estimation process may be performed on the sensor device 11 side. That is, the control device 11b of the sensor device 11 may be made to execute the processing (blood pressure estimation program) by the filtering processing unit 41, posture prediction unit 42, and blood pressure estimation unit 43, which are functions of the control unit 21. In this case, in the sensor device 11, the control unit (not shown) of the control device 11b, which receives biological information from the above-mentioned plurality of sheet-type piezoelectric sensors 11a, converts each piece of biological information into biological signal waveform data, estimates the blood pressure based on the biological signal waveform data, and transmits the estimation result to the server device 12. The server device 12, which receives the blood pressure estimation result, then displays the blood pressure estimation result received from the sensor device 11.

[0064] Furthermore, from the viewpoint of reducing and distributing the processing load, instead of having the control device 11b of the sensor device 11 execute all the processing (blood pressure estimation program) of the filtering processing unit 41, posture prediction unit 42, and blood pressure estimation unit 43, for example, the processing by the filtering processing unit 41 may be executed on the sensor device 11 side, and the processing by the posture prediction unit 42 and blood pressure estimation unit 43 may be executed on the server device 12 side. In this case, the control device 11b of the sensor device 11 transmits the filtered biological waveform data (heart sound waveform, cardiac pulse waveform, respiratory waveform) to the server device 12.

[0065] In other words, the blood pressure estimation process in this embodiment only needs to be in a state where the blood pressure estimation result can be displayed and confirmed, and the individual processes (each step) related to the blood pressure estimation process can be executed on either the sensor device 11 or the server device 12, as long as there is coordination between them.

[0066] Furthermore, the blood pressure estimation system 1 of this embodiment is useful as a medical examination system in various facilities such as hospitals and nursing homes. It can also be used as a monitoring system to inform doctors, staff, family members, etc., in different rooms about changes in blood pressure by connecting multiple personal computers to a server via a network such as cables or communication lines, and using the displays of the connected personal computers as display units. It can also be used as a telemedicine system to transmit the changes in blood pressure of remote patients receiving online medical consultations.

[0067] It should be noted that the present invention is not limited to the embodiments described above. The embodiments described above are illustrative, and any configuration that has substantially the same technical idea as described in the claims and produces similar effects is included within the technical scope of the present invention. [Explanation of symbols]

[0068] 1. Blood pressure estimation system 11 Sensor device 11a Sheet-type piezoelectric sensor 11b Control device 12 Server devices 21 Control Unit 22 Memory section 23 Input section 24 Interface (I / F) section 25 Display section 26 Communications Department 31,41 Filtering Processing Unit 32. Dataset Generation Unit 33. Machine Learning Department 42 Posture prediction unit 43 Blood pressure estimation unit

Claims

1. A sensor device comprising: multiple sheet-type sensors installed at a specific distance from each other to detect biological information including cardiac resilience at each position; and a control device that converts the biological information detected by these sheet-type sensors into biological signal waveform data and transmits it; A server device having a control unit that estimates blood pressure based on biosignal waveform data received from the sensor device, and a display unit that displays the blood pressure estimation result by the control unit, Equipped with, The control unit of the server device is A filtering processing unit that acquires bio-waveforms, including cardiac elasticity waveforms, from bio-signal waveform data, A posture prediction unit predicts the posture of a subject by inputting the bio-waveforms of a subject for blood pressure estimation into a trained posture prediction model that uses bio-waveforms corresponding to the bio-waveforms of a predetermined number of subjects detected while they change their posture as input data, and the postures of the subjects at the time of bio-waveform detection corresponding to these bio-waveforms as training data. A blood pressure estimation unit estimates the systolic blood pressure at a specific posture based on the predicted posture of the subject and the pulse wave propagation time calculated from the cardiac elasticity waveforms at two locations on the subject's body. Having, A blood pressure estimation system characterized by the following features.

2. A sensor device comprising: multiple sheet-type sensors installed at a specific distance from each other to detect biological information including cardiac resonance at each position; and a control device having a control unit that converts the biological information detected by these sheet-type sensors into biological signal waveform data, estimates blood pressure based on the biological signal waveform data, and transmits the estimation result; A server device that displays the estimated blood pressure results received from the aforementioned sensor device, Equipped with, The control unit of the control device in the aforementioned sensor device is A filtering processing unit that acquires bio-waveforms, including cardiac elasticity waveforms, from bio-signal waveform data, A posture prediction unit predicts the posture of a subject by inputting the bio-waveforms of a subject for blood pressure estimation into a trained posture prediction model that uses bio-waveforms corresponding to the bio-waveforms of a predetermined number of subjects detected while they change their posture as input data, and the postures of the subjects at the time of bio-waveform detection corresponding to these bio-waveforms as training data. A blood pressure estimation unit estimates the systolic blood pressure at a specific posture based on the predicted posture of the subject and the pulse wave propagation time calculated from the cardiac elasticity waveforms at two locations on the subject's body. Having, A blood pressure estimation system characterized by the following features.

3. The aforementioned specific posture is defined as the supine position. A blood pressure estimation system according to claim 1 or 2.

4. The sheet-type sensor is divided into two parts. A blood pressure estimation system according to claim 1 or 2.

5. The aforementioned biological information further includes heart sounds and / or respiration. A blood pressure estimation system according to claim 1 or 2.

6. A control unit receives biological information, including cardiac movement, detected by multiple sheet-type sensors installed at a specific distance, as biological signal waveform data, and estimates blood pressure based on the said biological signal waveform data. A display unit that displays the blood pressure estimation result from the control unit, It has, The control unit, A filtering processing unit that acquires bio-waveforms, including cardiac elasticity waveforms, from bio-signal waveform data, A posture prediction unit predicts the posture of a subject by inputting the bio-waveforms of a subject for blood pressure estimation into a trained posture prediction model that uses bio-waveforms corresponding to the bio-waveforms of a predetermined number of subjects detected while they change their posture as input data, and the postures of the subjects at the time of bio-waveform detection corresponding to these bio-waveforms as training data. A blood pressure estimation unit estimates the systolic blood pressure at a specific posture based on the predicted posture of the subject and the pulse wave propagation time calculated from the cardiac elasticity waveforms at two locations on the subject's body. Having, A blood pressure estimation device characterized by the following features.

7. Multiple sheet-type sensors are installed at a specific distance from each other and detect biometric information, including cardiac resonance, at each location. A control device having a control unit that converts biological information detected by the plurality of sheet-type sensors into biological signal waveform data, estimates blood pressure based on the biological signal waveform data, and transmits the estimation result, Equipped with, The control unit, A filtering processing unit that acquires bio-waveforms, including cardiac elasticity waveforms, from bio-signal waveform data, A posture prediction unit predicts the posture of a subject by inputting the bio-waveforms of a subject for blood pressure estimation into a trained posture prediction model that uses bio-waveforms corresponding to the bio-waveforms of a predetermined number of subjects detected while they change their posture as input data, and the postures of the subjects at the time of bio-waveform detection corresponding to these bio-waveforms as training data. A blood pressure estimation unit estimates the systolic blood pressure at a specific posture based on the predicted posture of the subject and the pulse wave propagation time calculated from the cardiac elasticity waveforms at two locations on the subject's body. Having, A blood pressure estimation device characterized by the following features.

8. A blood pressure estimation method by a control unit of a blood pressure estimation device that receives biological information, including cardiac resilience, detected by multiple sheet-type sensors installed at a specific distance from each other, as biological signal waveform data, The control unit, A filtering step to obtain a bio-waveform including cardiac elasticity waveform from bio-signal waveform data, A posture prediction step involves inputting the bio-waveforms of a subject for blood pressure estimation into a trained posture prediction model, which is created by machine learning using bio-waveforms corresponding to the bio-waveforms of a predetermined number of subjects detected while changing their posture as input data, and the postures of the subjects at the time of bio-waveform detection corresponding to these bio-waveforms as training data. A blood pressure estimation step that estimates the systolic blood pressure at a specific posture based on the predicted posture of the subject and the pulse wave propagation time calculated from the cardiac elasticity waveforms at two locations on the subject's body, Execute A method for estimating blood pressure characterized by the following features.

9. A blood pressure estimation program is provided for estimating the blood pressure of a subject, which is a control unit of a blood pressure estimation device that receives biological information, including cardiac elasticity, detected by a sheet-type sensor, as biological signal waveform data, The computer operating as the control unit for the blood pressure estimation device, To perform a series of processes according to each step described in claim 8, A blood pressure estimation program characterized by the following features.

Images