Sphygmomanometer, blood pressure measurement method, blood pressure measurement program, learning model construction method, and learning model construction program

Abstract

To acquire a blood pressure with high accuracy even if a pulse wave includes noise.SOLUTION: A sphygmomanometer includes: a cuff that is wound around an arm of a person to be measured; a control unit for controlling a pressure applied on the arm by the cuff; a detection unit for detecting a cuff pressure that the cuff receives from the arm; a generation unit for generating a first waveform image that includes a first image of a first pulse waveform which is obtained from a time-series change in the cuff pressure and indicates a change in the pressure due to pulsation in a blood flow flowing in an artery of the person to be measured and indicates the cuff pressure by a feature amount of the first image; a calculation model constructed through machine learning using a blood pressure of a subject as an objective variable while using a second waveform image that includes a second image of a second pulse waveform that is obtained from a time-series change in the cuff pressure of the subject and indicates a change in the pressure due to pulsation of the blood flow flowing in the artery of the subject and indicates the cuff pressure by a feature amount of the second image as an explanatory variable; and an acquisition unit for inputting the first waveform image to the calculation model and acquiring the blood pressure of the person to be measured from the calculation model.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a sphygmomanometer, a blood pressure measurement method, a blood pressure measurement program, a learning model construction method, and a learning model construction program. [Background technology]

[0002] Health conditions are known through blood pressure measurement. Patent Document 1 describes a technique for storing a correlation between a subject's blood pressure measured using a cuff and the subject's blood pressure measured by auscultation, and correcting the subject's blood pressure using the stored correlation. Patent Document 2 describes a technique for measuring blood pressure using a photoelectric volume pulse wave. Patent Document 3 proposes a technique for measuring blood pressure using a feature image generated by applying wavelet transformation or the like to a cuff pressure pulse wave. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 5-309073 [Patent Document 2] Special Publication No. 2022-516820 [Patent Document 3] US Patent Application Publication No. 2020 / 0146568 Summary of the Invention [Problem to be solved by the invention]

[0004] The pulse wave and photoplethysmogram obtained from the cuff pressure may contain noise due to the subject's body movements, the way the cuff is wrapped, the position where the cuff is wrapped, etc. When blood pressure is measured using the oscillometric method, blood pressure is measured based on the waveform of the pulse wave, so if such noise is included in the pulse wave, there is a risk that the measurement accuracy of blood pressure will decrease.

[0005] One aspect of the disclosed technology aims to provide a blood pressure monitor, a blood pressure measurement method, a blood pressure measurement program, a method for constructing a learning model, and a program for constructing a learning model that can obtain blood pressure with high accuracy even if the pulse wave contains noise. [Means for solving the problem]

[0006] One aspect of the disclosed technology is exemplified by a blood pressure monitor as follows. This blood pressure monitor includes a cuff wrapped around the arm of a subject, a control unit that controls the pressure applied to the arm by the cuff, a detection unit that detects the cuff pressure that the cuff receives from the arm, a generation unit that includes a first image of a first pulse waveform that indicates a change in pressure due to pulsation of blood flowing in the artery of the subject, obtained from a time-series change in the cuff pressure, and generates a first waveform image that indicates the cuff pressure using a feature amount of the first image, a calculation model that includes a second image of a second pulse waveform that indicates a change in pressure due to pulsation of blood flowing in the artery of the subject, obtained from a time-series change in the cuff pressure for the subject, and is constructed by machine learning using a second waveform image that indicates the cuff pressure using a feature amount of the second image as an explanatory variable and the blood pressure of the subject as an objective variable, and an acquisition unit that inputs the first waveform image into the calculation model and acquires the blood pressure of the subject from the calculation model.

[0007] The time series change of the cuff pressure includes the time series change of the pressure caused by the pulsation of the blood flow in the artery of the arm, which is superimposed as minute vibrations on the time series change of the pressure caused by the pulsation of the blood flow in the artery of the arm. According to this blood pressure meter, the waveform image includes a waveform showing a global change of the pulse waveform, which shows minute vibrations, and a waveform showing local changes. The waveform image includes the shape of the pulse wave and also includes information on the cuff pressure. A calculation model constructed by machine learning using such a waveform image as a response variable can output the blood pressure of the subject with high accuracy even if the waveform image contains noise. In other words, this blood pressure monitor can achieve high-accuracy blood pressure measurement even if the pulse wave obtained from the cuff pressure contains noise. The arm of the subject also includes the wrist of the subject.

[0008] This blood pressure monitor may further include the following features: The generating unit draws the pulse waveform as a line image, and generates the first waveform image indicating the cuff pressure by a feature amount of a line forming the line image. Such a waveform image includes both global and local features of the pulse waveform by drawing the pulse waveform as a line image, and can also include information of the cuff pressure by the feature amount of the line. Here, the feature amount of the line may include a color of the line, and the cuff pressure may be indicated by the color of the line. Furthermore, the feature amount of the line may include a thickness of the line, The cuff pressure may be indicated by the thickness of the line. By using such a waveform image, the blood pressure monitor can perform blood pressure measurement with higher accuracy.

[0009] This blood pressure monitor may further include the following features. The generating unit draws the pulse waveform with a line image, and generates the first waveform image showing the cuff pressure by a feature amount of the background of the line image. Such a waveform image includes both global and local features of the pulse waveform by drawing the pulse waveform with a line image, and can also include information of the cuff pressure by the feature amount of the background. Here, the feature amount of the background may include a color of the background, and the cuff pressure may be indicated by the color of the background. By using such a waveform image, this blood pressure monitor can perform blood pressure measurement with higher accuracy.

[0010] The blood pressure monitor may further include the following feature: The calculation model is constructed by the machine learning using the diastolic blood pressure and the systolic blood pressure of the subject as the objective variables. With such a calculation model, it is possible to learn feature amounts effective for calculating the diastolic blood pressure and the systolic blood pressure.

[0011] The disclosed technology can also be understood from the aspects of a blood pressure measurement method, a blood pressure measurement program, a learning model construction method, and a learning model construction program. Effect of the Invention

[0012] According to the disclosed technology, blood pressure can be obtained with high accuracy even if the pulse wave contains noise. [Brief description of the drawings]

[0013] [Figure 1] FIG. 1 is a diagram illustrating an example of the configuration of a sphygmomanometer according to an embodiment. [Diagram 2] FIG. 2 is a schematic diagram showing a state in which the blood pressure monitor is attached to the wrist of a subject. [Diagram 3] FIG. 3 is a diagram illustrating variations of the waveform image. [Figure 4] FIG. 4 is a diagram illustrating an example of a verification result of blood pressure measurement accuracy when blood pressure is measured using a waveform image. [Diagram 5] FIG. 5 is a diagram illustrating an example of a blood pressure model construction process according to the embodiment. [Figure 6] FIG. 6 is a diagram illustrating an example of a process flow of the sphygmomanometer according to the embodiment. [Figure 7] FIG. 7 is a diagram illustrating an example of the configuration of a sphygmomanometer according to a comparative example. [Figure 8] FIG. 8 is a diagram illustrating the comparative verification results. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] <Application Examples> An application example of the present invention will be described. A blood pressure monitor 100 according to the application example is a device that measures the blood pressure of a person to be measured by wrapping a cuff 120 around the upper arm or wrist of the person to be measured. The unit 114 inputs at least one of the time series changes in the cuff pressure and the time series changes in the pulse wave to the blood pressure model 116, and obtains the blood pressure of the subject from the blood pressure model 116.

[0015] The acquiring unit 114 controls the pressure of the cuff 120 to acquire time series changes in the cuff pressure and time series changes in the pulse wave from the pressure sensor 113. The generating unit 115 generates a waveform image showing the time series changes in the cuff pressure and the pulse wave acquired by the acquiring unit 114. The waveform image is, for example, a line image showing the time series changes in the pulse wave with a line, to which information showing the time series changes in the cuff pressure is added using a feature amount of the line image.

[0016] The acquiring unit 114 inputs the waveform image generated by the generating unit 115 to the blood pressure model 116, and acquires the blood pressure of the subject from the blood pressure model 116. The acquiring unit 114 causes the output unit 117 to output the acquired blood pressure.

[0017] The blood pressure model 116 is a learning model that outputs the blood pressure of a subject when a waveform image is input. The blood pressure model 116 is constructed by machine learning that uses the waveform image of the subject as an explanatory variable and the blood pressure adopted as a true value as a target variable. Machine learning includes deep learning. The blood pressure adopted as a true value may be a blood pressure measured by auscultation.

[0018] According to this application example, the waveform image includes a waveform showing a global change in the time series change of the pulse wave and a waveform showing a local change. The blood pressure model 116 constructed by machine learning using such a waveform image as a response variable can output the blood pressure of the subject with high accuracy even if the waveform image contains noise. That is, according to this application example, even if the waveform image generated by the generation unit 115 contains noise, the blood pressure can be acquired with high accuracy.

[0019] <Embodiment> Hereinafter, an embodiment will be described with reference to the drawings. Fig. 1 is a diagram showing an example of the configuration of a blood pressure monitor 100 according to an embodiment. Fig. 2 is a diagram showing a schematic diagram of the blood pressure monitor 100 attached to the wrist H1 of a person to be measured. The blood pressure monitor 100 includes a main body 110 and a cuff 120. The blood pressure monitor 100 acquires the blood pressure of the person to be measured by wrapping the cuff 120 around the wrist H1 of the person to be measured, for example. Note that the position where the cuff 120 is wrapped is not limited to the wrist H1, and may be the upper arm of the person to be measured.

[0020] The cuff 120 is formed in a belt shape and has a bag into which air is supplied. When air is supplied to the bag inside the cuff 120 while the cuff 120 is wrapped around the wrist H1 of the subject, the pressure by the cuff on the wrist H1 increases. When air is inhaled from the bag inside the cuff 120 while the cuff 120 is wrapped around the wrist H1 of the subject, the pressure by the cuff on the wrist H1 decreases.

[0021] The main body 110 includes, for example, a processor and a storage unit, and the processor executes a program stored in the storage unit to realize each processing unit, such as a start switch 111, a control unit 112, a pressure sensor 113, an acquisition unit 114, a blood pressure model 116, and an output unit 117. The acquisition unit 114 includes a generation unit 115. In other words, the sphygmomanometer 100 can be regarded as a computer having a cuff 120. Note that each processing unit, such as the start switch 111, the control unit 112, the pressure sensor 113, the acquisition unit 114, the blood pressure model 116, and the output unit 117, may be a dedicated hardware circuit. In the sphygmomanometer 100, when the start switch 111 is turned on, the sphygmomanometer 100 is started and blood pressure measurement is started.

[0022] The control unit 112 controls the pressure applied to the wrist H1 of the subject by the cuff 120 by supplying air to the cuff 120 and inhaling air from the cuff 120. The control unit 112 starts controlling the pressure of the cuff 120 when the start switch 111 is turned on.

[0023] The pressure sensor 113 is a sensor that detects the pressure (cuff pressure) by the cuff 120. The pressure sensor 113 outputs the detected cuff pressure to the acquisition unit 114. Such a time series change in the cuff pressure includes, for example, a time series change in pressure (pressure vibration) due to pulsation of the blood flow in the artery of the wrist H1 superimposed as a minute vibration on the time series change in the pressure of the cuff 120 pressing against the wrist H1.

[0024] The acquiring unit 114 acquires the blood pressure of the subject using the blood pressure model 116. The acquiring unit 114 acquires from the pressure sensor 113 a time series change in the cuff pressure of the cuff 120, which is increased to a pressure higher than the systolic blood pressure and then gradually reduced to a pressure lower than the diastolic blood pressure. The acquiring unit 114 also extracts a time series change in pressure due to pulsation of blood flowing in an artery from the time series change in the cuff pressure acquired from the pressure sensor 113, and acquires a waveform indicating a time series change in the pulse wave (hereinafter also referred to as a pulse waveform).

[0025] The generating unit 115 generates a single waveform image showing the extracted time series change of the cuff pressure and the pulse waveform. As explained in the application example, the waveform image may be a waveform image in which the pulse waveform is drawn with a line and information showing the time series change of the cuff pressure is added using the feature amount of the drawn line image. It is preferable that the waveform image is generated in a non-compressed image file format. The image size of the waveform image is preferably an image size that allows the pulse waveform to be visually observed. The image size of the waveform image is, for example, 448 pixels vertical x 448 pixels horizontal.

[0026] Here, the waveform image will be described. FIG. 3 is a diagram illustrating variations of the waveform image. In each of FIGS. 3A, 3B, 3C, and 3D, the vertical axis illustrates pressure (mmHg) and the horizontal axis illustrates time (seconds). FIG. 3A illustrates a waveform image generated by filling in the pulse waveform. FIG. 3B illustrates a waveform image in which the pulse waveform is drawn with a line. FIG. 3C illustrates a waveform image in which the waveform of the time series change of the cuff pressure (hereinafter also referred to as the cuff pressure waveform) is overwritten with a line on the image generated by filling in the pulse waveform. FIG. 3D illustrates a waveform image in which the pulse waveform is drawn with a line and the time series change of the cuff pressure is added by changing the color of the line according to the cuff pressure.

[0027] The waveform image of FIG. 3A includes the variation of the global waveform shape such as the envelope shape of the pulse waveform. However, since the waveform image of FIG. 3A is filled in, the information of the local variation of the pulse waveform is not included. Since the waveform image of FIG. 3B is drawn with a line, the global waveform shape and the local waveform shape of the pulse waveform are included. However, neither of the waveform images of FIG. 3A nor 3B includes the information of the time series change of the cuff pressure. Moreover, although the waveform image of FIG. 3C includes the information of the time series change of the cuff pressure, it does not include the information of the local variation of the pulse waveform, as in FIG. 3A. Therefore, in order to realize high-precision blood pressure measurement, it is considered that it is not preferable to use the waveform images exemplified in FIG. 3A, 3B, and 3C for blood pressure measurement.

[0028] In the waveform image of Fig. 3D, the pulse waveform is drawn with a line, so that the overall waveform shape and the local waveform shape of the pulse waveform are included.Furthermore, in the waveform image of Fig. 3D, the information showing the time series change of the cuff pressure is shown by the color of the line that forms the pulse waveform of the pulse wave.That is, Fig. 3D includes both the information showing the time series change of the pulse waveform and the information showing the time series change of the cuff pressure.

[0029] Here, the accuracy of blood pressure measurement using the waveform images exemplified in each of Figs. 3A, 3B, 3C, and 3D will be verified. In this verification, systolic blood pressure (SBP) was measured using the waveform images exemplified in each of Figs. 3A, 3B, 3C, and 3D. In this verification, blood pressure was measured multiple times for each of multiple subjects to obtain measurement results. In this verification, the obtained measurement results are evaluated using Standard Deviation of Error (SDE).

[0030] Figure 4 shows an example of the verification result of blood pressure measurement accuracy when blood pressure is measured using a waveform image. 4 is a diagram showing the waveform image shown in FIG. 3A as a "silhouette image," the waveform image shown in FIG. 3B as a "line image," the waveform image shown in FIG. 3C as an "overlay image," and the waveform image shown in FIG. 3C as a "colored image." Referring to FIG. 4, the highest accuracy was observed when blood pressure was measured using the "colored image." This is considered to be because the "colored image" contains information showing the global waveform shape of the pulse wave, the local waveform shape, and the time series change in cuff pressure, and therefore has a larger amount of information than other waveform images. In addition, it is considered that the regularity of the relationship between the color information (RGB information) and the cuff pressure, which indicates the cuff pressure by the color of the line, has a favorable effect on the accuracy of blood pressure measurement.

[0031] As shown in FIG. 3D, the generating unit 115 generates a waveform image so that the information indicating the global waveform shape of the pulse waveform, the local waveform shape, and the time series change of the cuff pressure are included in one waveform image. For example, the generating unit 115 generates a waveform image showing the pulse waveform as a line image and showing information related to the cuff pressure as the characteristics of the line. Examples of the characteristics of the line showing the information related to the cuff pressure include the thickness of the line, the color of the line, the type of line (single line, one-dot chain line, two-dot chain line, double line, etc.), the brightness of the line, etc. In addition, the generating unit 115 may generate a waveform image showing the pulse waveform as a line image and showing information related to the cuff pressure as the characteristics of the background of the pulse waveform. Examples of the characteristics of the background showing the information related to the cuff pressure include the color of the background, the brightness of the background, etc.

[0032] The acquiring unit 114 inputs the waveform image generated by the generating unit 115 to the blood pressure model 116. The acquiring unit 114 acquires the diastolic blood pressure and systolic blood pressure of the subject as outputs from the blood pressure model 116 to which the waveform image has been input.

[0033] The blood pressure model 116 is a learning model that outputs a diastolic blood pressure and a systolic blood pressure when a waveform image is input. The blood pressure model 116 is stored in, for example, a storage unit contained in the main body 110. The blood pressure model 116 includes, for example, a Deformable Conv Model.

[0034] The output unit 117 outputs the diastolic blood pressure and the systolic blood pressure acquired by the acquisition unit 114. The output unit 117 is, for example, a Liquid Crystal Display (LCD), a Plasma Display Panel (PDP), an inorganic electroluminescence (EL) panel, or an organic EL panel. The output unit 117 may be a printer or a speaker.

[0035] <Construction of Blood Pressure Model 116> 5 is a diagram showing an example of a process for constructing the blood pressure model 116 according to the embodiment. An example of a process for constructing the blood pressure model 116 will be described with reference to FIG.

[0036] In step S1, multiple subjects are prepared. In this embodiment, multiple subjects are prepared, but the number of subjects prepared may be one. Cuffs 120 are attached to the multiple subjects prepared, and cuff pressure waveforms and pulse waveforms are collected. In addition, diastolic blood pressures and systolic blood pressures measured by auscultation for the multiple subjects prepared are collected.

[0037] In step S2, the generator 115 generates a waveform image based on each of the cuff pressure waveform and the pulse waveform collected in step S1. In step S3, learning is performed to construct a blood pressure model 116 by machine learning using the waveform image generated in step S2 as an explanatory variable and the diastolic blood pressure and the systolic blood pressure measured in step S1 as objective variables.

[0038] In step S4, the blood pressure model 116 learned in step S3 is evaluated. In the evaluation, for example, a waveform image is generated for a certain subject, and the diastolic blood pressure and systolic blood pressure are measured by auscultation. The generated waveform image is input to the blood pressure model 116. The diastolic blood pressure and the systolic blood pressure are output from the blood pressure model 116. The diastolic blood pressure and the systolic blood pressure output by the blood pressure model 116 are compared with the diastolic blood pressure and the systolic blood pressure measured by the auscultation method, thereby evaluating the blood pressure model 116.

[0039] If the evaluation result of step S4 indicates that the blood pressure model 116 satisfies the desired accuracy (YES in step S5), the process proceeds to step S6. If the blood pressure model 116 does not satisfy the predetermined accuracy (NO in step S5), the process proceeds to step S3. In step S6, the generated blood pressure model 116 is stored in the storage unit of the sphygmomanometer 100.

[0040] <Processing flow> 6 is a diagram showing an example of a processing flow of the sphygmomanometer 100 according to the embodiment. Hereinafter, an example of a processing flow of the sphygmomanometer 100 will be described with reference to FIG.

[0041] In step S11, the start switch 111 is turned on to start the blood pressure monitor 100. In step S12, the control unit 112 pressurizes the wrist H1 of the subject by supplying air to the cuff 120. After the control unit 112 pressurizes the wrist H1 of the subject to a pressure higher than the systolic blood pressure using the cuff 120, the control unit 112 gradually reduces the pressure in the cuff 120.

[0042] In step S13, the acquisition unit 114 acquires a cuff pressure waveform and a pulse waveform based on the cuff pressure detected by the pressure sensor 113. Furthermore, the generation unit 115 generates a waveform image based on the waveform acquired in step S13. In step S14, the acquisition unit 114 inputs the waveform image generated in step S13 to the blood pressure model 116 to acquire the diastolic blood pressure and systolic blood pressure of the subject.

[0043] In step S15, the acquisition unit 114 determines whether or not the blood pressure measurement of the subject is completed. The acquisition unit 114 may determine that blood pressure acquisition is completed when, for example, both the diastolic blood pressure and the systolic blood pressure of the subject are acquired from the blood pressure model 116. If completed (YES in step S15), the process proceeds to step S16. If not completed (NO in step S15), the process proceeds to step S12.

[0044] In step S16, the control unit 112 deflates the air from the cuff 120. In step S17, the output unit 117 outputs the blood pressure acquired in step S14.

[0045] <Comparative Example> In the embodiment, the blood pressure model 116 is constructed to output the diastolic blood pressure and the systolic blood pressure. In the comparative example, a configuration in which the blood pressure model is divided into a blood pressure model that outputs the diastolic blood pressure and a blood pressure model that outputs the systolic blood pressure will be described.

[0046] 7 is a diagram showing an example of the configuration of a blood pressure monitor 500 according to a comparative example. The blood pressure monitor 500 includes a main body 510, a start switch 511, a control unit 512, a pressure sensor 513, an acquisition unit 514, a generation unit 515, a blood pressure model 516A, a blood pressure model 516B, an output unit 517, and a cuff 520. The main body 510, the start switch 511, the control unit 512, the pressure sensor 513, the acquisition unit 514, the generation unit 515, and the cuff 520 are similar to the main body 110, the start switch 111, the control unit 112, the pressure sensor 113, the acquisition unit 114, the generation unit 115, and the cuff 120 of the blood pressure monitor 100 according to the embodiment, and therefore description thereof will be omitted.

[0047] The blood pressure model 516A is a learning model that outputs the systolic blood pressure when a waveform image is input. The blood pressure model 516A is constructed by machine learning with the waveform image as an explanatory variable and the systolic blood pressure adopted as a true value as a target variable. The blood pressure model 516B is a learning model that outputs the diastolic blood pressure when a waveform image is input. The blood pressure model 516B is constructed by machine learning with the waveform image as an explanatory variable and the systolic blood pressure adopted as a true value as a target variable. It is constructed by machine learning with diastolic blood pressure, which is taken as the true value and the mean variable, as the objective variable.

[0048] <Comparative verification> A comparative verification of blood pressure measurement accuracy between the blood pressure monitor 100 according to the embodiment and the blood pressure monitor 500 according to the comparative example will be described below. In this comparative verification, the systolic blood pressure (SBP) was measured for each of the blood pressure monitors 100 and 500. In this comparative verification, blood pressure was measured multiple times for each of multiple subjects to obtain measurement results. In this comparative verification, the obtained measurement results are evaluated by SDE.

[0049] FIG. 8 is a diagram illustrating the results of the comparative verification. Referring to FIG. 8, the highest accuracy was observed when blood pressure was measured using the "embodiment". From this, it is considered that higher measurement accuracy can be obtained by learning systolic blood pressure and diastolic blood pressure simultaneously when constructing a blood pressure model.

[0050] <Effects of the embodiment> In this embodiment, the waveform image generated by the generating unit 115 includes a waveform showing a global change in the time series change of the pulse wave and a waveform showing a local change. The blood pressure model 116 constructed by machine learning using such a waveform image as a response variable can output the blood pressure of the subject with high accuracy even if the waveform image contains noise. That is, according to this application example, high accuracy blood pressure measurement can be realized even if the waveform image generated by the generating unit 115 contains noise.

[0051] In this embodiment, the pulse waveform is drawn as a line image, and a waveform image showing the cuff pressure is generated according to the characteristic amount of the line forming the line image. Such a waveform image includes both the global characteristic and the local characteristic of the pulse waveform by drawing the pulse waveform as a line image, and can include the information of the cuff pressure according to the characteristic amount of the line. By using such a waveform image, the blood pressure monitor 100 can perform blood pressure measurement with higher accuracy.

[0052] In this embodiment, the pulse waveform may be drawn as a line image, and a waveform image showing the cuff pressure may be generated based on the feature amount of the background of the line image. Such a waveform image may include both the global feature and the local feature of the pulse waveform by drawing the pulse waveform as a line image, and may include the information of the cuff pressure based on the feature amount of the background. By using such a waveform image, the blood pressure monitor 100 can perform blood pressure measurement with higher accuracy.

[0053] In this embodiment, the blood pressure model 116 simultaneously learns the systolic blood pressure and the diastolic blood pressure. Therefore, the blood pressure model 116 can learn features that are effective in predicting both the systolic blood pressure and the diastolic blood pressure. Furthermore, according to this embodiment, the measurement of the systolic blood pressure and the diastolic blood pressure using the blood pressure model 116 can be made more accurate.

[0054] <Modification> In the above-described embodiment, the waveform image is added to the line image showing the pulse waveform by a line, and the information showing the time series change of the cuff pressure is added to the line image by using the feature amount of the line image.However, the waveform image is not limited to this type of image.For example, the waveform image may be added to the line image showing the cuff pressure waveform by a line, and the information showing the time series change of the pulse wave is added to the line image.

[0055] The embodiments and modifications disclosed above can be combined with each other.

[0056] <Computer-readable recording medium> A computer or other machine or device (hereinafter referred to as a computer, etc.) that performs any of the above functions. An information processing program for realizing the functions of the present invention can be recorded on a recording medium that can be read by a computer, etc. Then, the computer, etc. can provide the functions by reading and executing the program from the recording medium.

[0057] Here, a computer-readable recording medium refers to a recording medium that stores information such as data and programs electrically, magnetically, optically, mechanically, or chemically and can be read from a computer. Among such recording media, those that can be removed from a computer include, for example, flexible disks, magneto-optical disks, Compact Disc Read Only Memory (CD-ROM), Compact Disc-Recordable (CD-R), Compact Disc-ReWriterable (CD-RW), Digital Versatile Disc (DVD), Blu-ray Disc (BD), Digital Audio Tape (DAT), 8mm tape, flash memory, external hard disk drives, and Solid State Drives (SSD). In addition, recording media fixed to a computer include built-in hard disk drives, SSDs, ROMs, and the like.

[0058] <Appendix 1> A cuff (120) to be wrapped around the subject's arm; a control unit (112) for controlling the pressure applied to the arm (H1) by the cuff (120); a detection unit (113) for detecting a cuff pressure received by the cuff (120) from the arm (H1); a generating unit (115) that generates a first waveform image indicating the cuff pressure based on a feature amount of the first image, the first image including a first pulse waveform indicating a pressure change due to pulsation of the blood flow in the artery of the subject, the first image being obtained from a time-series change of the cuff pressure; A calculation model (116) constructed by machine learning using a second image of a second pulse waveform showing a change in pressure due to pulsation of blood flow in the artery of the subject, the second image being obtained from a time-series change in the cuff pressure of the subject, the second waveform image showing the cuff pressure being an explanatory variable based on a feature amount of the second image, and the blood pressure of the subject being a target variable; an acquisition unit (114) that inputs the first waveform image into the calculation model (116) and acquires the blood pressure of the subject from the calculation model (116), Blood pressure monitor(100). <Appendix 2> The generation unit (115) The pulse waveform is drawn as a line image, and the first waveform image showing the cuff pressure is generated based on a feature amount of the line forming the line image. 2. The blood pressure monitor (100) of claim 1. <Appendix 3> the feature amount of the line includes a color of the line, The cuff pressure is indicated by the color of the line. 3. The blood pressure monitor (100) of claim 2. <Appendix 4> the line feature amount includes a thickness of the line, The cuff pressure is indicated by the thickness of the line. 4. The blood pressure monitor (100) according to claim 2 or 3. <Appendix 5> The generation unit (115) The pulse waveform is drawn as a line image, and the first waveform image showing the cuff pressure is generated based on a feature amount of the background of the line image. 2. The blood pressure monitor (100) of claim 1. <Appendix 6> the feature amount of the background includes a color of the background, The cuff pressure is indicated by the color of the background. 6. The blood pressure monitor (100) according to claim 5. <Appendix 7> The calculation model (116) is constructed by the machine learning using the diastolic blood pressure and the systolic blood pressure of the subject as the objective variables. 7. A blood pressure monitor (100) according to any one of claims 1 to 6. <Appendix 8> a detection step of detecting a cuff pressure received from the arm (H1) of a subject by a cuff (120) wrapped around the arm (H1); A generating step of generating a first waveform image showing the cuff pressure according to a feature amount of the first image, the first image including a first image showing a pulse waveform obtained from a time series change of the cuff pressure; and an acquisition step of inputting the first waveform image into a calculation model (116) constructed by machine learning using a second image showing a pulse waveform acquired from a time-series change in the cuff pressure of the subject, the second waveform image showing the cuff pressure being an explanatory variable based on a feature amount of the second image, and the blood pressure of the subject being an objective variable, and acquiring the blood pressure of the subject from the calculation model (116), the computer (100) executes. How to measure blood pressure. <Appendix 9> a detection step of detecting a cuff pressure received from the arm (H1) of a subject by a cuff (120) wrapped around the arm (H1); A generating step of generating a first waveform image showing the cuff pressure according to a feature amount of the first image, the first image including a first image showing a pulse waveform obtained from a time series change of the cuff pressure; and inputting the first waveform image into a calculation model (116) constructed by machine learning using a second image showing a pulse waveform obtained from a time-series change in the cuff pressure of the subject, the second waveform image showing the cuff pressure being an explanatory variable according to a feature amount of the second image, and the blood pressure of the subject being a target variable, to obtain the blood pressure of the subject from the calculation model (116). Blood pressure measurement program. <Appendix 10> A detection step of detecting a cuff pressure received from a cuff (120) wrapped around the arm of a subject; A generating step of generating a waveform image showing the cuff pressure according to a feature amount of the image, the waveform image including an image showing a pulse waveform obtained from a time series change of the cuff pressure; A construction step of constructing a learning model by machine learning using the waveform image as an explanatory variable and the blood pressure of the subject as a target variable, executed by a computer (100). How to build a learning model. <Appendix 11> A detection step of detecting a cuff pressure received from a cuff (120) wrapped around the arm of a subject; A generating step of generating a waveform image showing the cuff pressure according to a feature amount of the image, the waveform image including an image showing a pulse waveform obtained from a waveform of the time series change of the cuff pressure; A construction step of constructing a learning model by machine learning using the waveform image as an explanatory variable and the blood pressure of the subject as a target variable, A program for building learning models. [Explanation of symbols]

[0059] 100 Blood Pressure Monitor 110··Main unit 111··Start switch 112 Control section 113 Pressure Sensor 114...Acquisition Department 115...Generation part 116 Blood Pressure Model 117··Output section 120··Cuff 110··Main unit 500··Blood pressure monitor 510··Main unit 511··Start switch 512 Control unit 513··Pressure Sensor 514...Acquisition Department 515...Generation part 516A Blood Pressure Model 516B Blood Pressure Model 517··Output section 520··Cuff H1 Wrist

Claims

1. A cuff to be wrapped around the subject's arm; a control unit for controlling the pressure applied to the arm by the cuff; A detection unit that detects a cuff pressure received by the cuff from the arm; A generator that includes a first image of a first pulse waveform that indicates a pressure change due to pulsation of blood flowing in the artery of the subject, the first image being obtained from a time-series change in the cuff pressure, and generates a first waveform image that indicates the cuff pressure based on a feature amount of the first image. A calculation model constructed by machine learning using a second image of a second pulse waveform showing a change in pressure due to pulsation of blood flow in the artery of the subject, the second image being obtained from a time series change in the cuff pressure of the subject, the second waveform image showing the cuff pressure being an explanatory variable based on a feature amount of the second image, and the blood pressure of the subject being a target variable; an acquisition unit that inputs the first waveform image into the calculation model and acquires the blood pressure of the subject from the calculation model. Blood pressure monitor.

2. The generation unit is The pulse waveform is drawn as a line image, and the first waveform image showing the cuff pressure is generated based on a feature amount of the line forming the line image.

2. The blood pressure monitor according to claim 1.

3. the feature amount of the line includes a color of the line, The cuff pressure is indicated by the color of the line. The blood pressure monitor according to claim 2.

4. the line feature amount includes a thickness of the line, The cuff pressure is indicated by the thickness of the line. The blood pressure monitor according to claim 2.

5. The generation unit is The pulse waveform is drawn as a line image, and the first waveform image showing the cuff pressure is generated based on a feature amount of the background of the line image.

2. The blood pressure monitor according to claim 1.

6. the feature amount of the background includes a color of the background, The cuff pressure is indicated by the color of the background.

6. The blood pressure monitor according to claim 5.

7. The calculation model is constructed by the machine learning using the diastolic blood pressure and the systolic blood pressure of the subject as the objective variables. The blood pressure monitor according to any one of claims 1 to 6.

8. a detection step of detecting a cuff pressure received from a subject's arm by a cuff wrapped around the subject's arm; a generating step of generating a first waveform image showing a change in pressure due to pulsation of blood flow in the artery of the subject, the first image being obtained from a time-series change in the cuff pressure, the first waveform image showing the cuff pressure being generated based on a feature amount of the first image; The second image of a second pulse waveform showing a change in pressure due to pulsation of blood flow in an artery of the subject is obtained from a time-series change in the cuff pressure of the subject, and the first waveform image is input to a calculation model constructed by machine learning using a feature amount of the second image as an explanatory variable and the blood pressure of the subject as a target variable. and acquiring the blood pressure of the subject from the calculation model, How to measure blood pressure.

9. a detection step of detecting a cuff pressure received from a subject's arm by a cuff wrapped around the subject's arm; a generating step of generating a first waveform image showing a change in pressure due to pulsation of blood flow in the artery of the subject, the first image being obtained from a time-series change in the cuff pressure, the first waveform image showing the cuff pressure being generated based on a feature amount of the first image; and acquiring the blood pressure of the subject from the calculation model, the calculation model including a second image of a second pulse waveform showing a change in pressure due to pulsation of blood flowing in the artery of the subject, the second image being acquired from a time-series change in the cuff pressure of the subject, the second waveform image showing the cuff pressure being an explanatory variable based on a feature amount of the second image, and the first waveform image being input into a calculation model constructed by machine learning using the blood pressure of the subject as a target variable. Blood pressure measurement program.

10. a detection step of detecting a cuff pressure received from a subject's arm by a cuff wrapped around the subject's arm; a generating step of generating a waveform image showing the cuff pressure based on a feature amount of the image, the waveform image including a pulse waveform image showing a pressure change due to pulsation of the blood flow in the artery of the subject, the pulse waveform image being obtained from the time series change of the cuff pressure; a construction step of constructing a learning model by machine learning using the waveform image as an explanatory variable and the blood pressure of the subject as a target variable, the computer executing the construction step. How to build a learning model.

11. a detection step of detecting a cuff pressure received from a subject's arm by a cuff wrapped around the subject's arm; a generating step of generating a waveform image showing the cuff pressure based on a feature amount of the image, the waveform image including a pulse waveform image showing a pressure change due to pulsation of the blood flow in the artery of the subject, the pulse waveform image being obtained from the time series change of the cuff pressure; A construction step of constructing a learning model by machine learning using the waveform image as an explanatory variable and the blood pressure of the subject as a target variable. A program for building learning models.

Images