Information processing program, information processing method, and information processing device

Abstract

This information processing program causes a computer to execute processing of: acquiring health state information including disease information of a patient; identifying a disease or a constitution on the basis of the health state information; acquiring a heart sound waveform of the patient; setting an extraction condition of a feature point of the heart sound waveform on the basis of the disease or the constitution; and extracting the feature point of the heart sound waveform on the basis of the set extraction condition.

Description

Information processing program, information processing method, and information processing device

[0001] The present invention relates to an information processing program, an information processing method, and an information processing device for extracting feature points of a cardiac waveform.

[0002] Methods for non-invasively estimating cardiac conditions have been proposed for the purpose of, for example, follow-up observation of patients with a history of cardiac disease such as heart failure. For example, Patent Literature 1 proposes a method for calculating left ventricular end-diastolic pressure (LVEDP) as a function of heart beat intervals, electrocardiogram (ECG), phonocardiogram, etc.

[0003] US Patent Application Publication No. 2023 / 131629

[0004] However, if a patient has a disease such as valvular disease, it may not be possible to accurately extract feature points from cardiac waveforms such as electrocardiograms and phonocardiograms. As a result, the accuracy of estimating left ventricular end-diastolic pressure may decrease. Patent Document 1 mentions the need to correct left ventricular end-diastolic pressure for patients with valvular disease, but does not disclose a specific method for correction or extraction of feature points.

[0005] The present invention has been made in view of the above circumstances, and its purpose is to provide an information processing program, an information processing method, and an information processing device that extract feature points of a cardiac waveform while taking into account the disease or constitution of a patient.

[0006] The present invention is an information processing program that causes a computer to execute the following steps: (1) acquire health condition information including disease information of a patient; identify the disease or constitution based on the health condition information; acquire the phonocardiogram of the patient; set extraction conditions for feature points of the phonocardiogram based on the disease or constitution; and extract feature points of the phonocardiogram based on the set extraction conditions.

[0007] Here, an embodiment of the present invention is as follows: (2) In the information processing program of (1) above, it is preferable that the extraction condition is to recognize that there are two or more peaks in the cardiac sound waveform and select the first peak, or to change the time range for extracting the feature points.

[0008] (3) In the information processing program of (1) or (2) above, the extraction condition preferably changes the order in which the feature points appear, or replaces the feature points with other feature points.

[0009] (4) In the information processing program described in any one of (1) to (3) above, it is preferable that the extraction condition changes an amplitude threshold for determining noise in the cardiac sound waveform.

[0010] (5) In the information processing program according to any one of (1) to (4) above, it is preferable that the feature points are discarded.

[0011] (6) In the information processing program described in any one of (1) to (5) above, it is preferable that the health condition information is obtained from an electronic medical record.

[0012] (7) In the information processing program according to any one of (1) to (6) above, the disease preferably includes valvular disease or arrhythmia.

[0013] (8) It is preferable that the information processing program described in any one of (1) to (7) above estimates intracardiac pressure based on the extracted feature points and outputs the estimated intracardiac pressure.

[0014] (9) It is preferable that the information processing program described in any one of (1) to (8) above acquires an electrocardiogram and a pulse waveform synchronized with a cardiac sound waveform, extracts feature points of the acquired electrocardiogram and pulse waveform, and estimates intracardiac pressure based on the feature points of the cardiac sound waveform and the feature points of the electrocardiogram and the pulse waveform.

[0015] (10) In the information processing program described in (9) above, the disease causes an abnormal change in the morphology of the cardiac sound waveform, and the abnormal change includes the splitting of one or two cardiac sounds in the cardiac sound waveform. Preferably, the extraction conditions are set so that the start points of the first sound and the second sound are extracted on the premise that two or more waveform peaks appear due to the splitting, or so that the start points of the first sound and the second sound are extracted from the R wave extracted from the electrocardiogram within a time range that takes the split into consideration.

[0016] (11) It is preferable that the information processing program described in (9) or (10) above estimates intracardiac pressure based on the characteristic points of the phonocardiogram waveform and the characteristic points of the electrocardiogram and pulse wave waveform that are extracted assuming that there is no disease or constitution that causes an abnormal change in the shape of the phonocardiogram waveform, and outputs the estimated intracardiac pressure.

[0017] (12) Preferably, the information processing method includes a computer executing a process of acquiring health condition information including disease information of a patient, identifying a disease or constitution based on the health condition information, acquiring a phonocardiogram waveform of the patient, setting extraction conditions for feature points of the phonocardiogram waveform based on the disease or constitution, and extracting feature points of the phonocardiogram waveform based on the set extraction conditions.

[0018] (13) It is preferable that the information processing device is an information processing device that includes a control unit, and the control unit acquires health condition information including disease information of a patient, identifies the disease or constitution based on the health condition information, acquires the phonocardiogram of the patient, sets extraction conditions for feature points of the phonocardiogram based on the disease or constitution, and executes a process of extracting feature points of the phonocardiogram based on the set extraction conditions.

[0019] In one aspect of the present application, it is possible to extract feature points of the cardiac waveform taking into consideration the illnesses and constitutions of the patient.

[0020] 1 is an explanatory diagram showing an example of the configuration of a monitoring system. FIG. 1 is a block diagram showing an example of the hardware configuration of a server. FIG. 1 is a block diagram showing an example of the hardware configuration of a measurement device. FIG. 1 is a block diagram showing an example of the hardware configuration of a communication terminal. FIG. 1 is a block diagram showing an example of the hardware configuration of a medical professional terminal. FIG. 1 is an explanatory diagram showing an example of a patient DB. FIG. 1 is an explanatory diagram showing an example of a comorbidity DB. FIG. 1 is an explanatory diagram showing an example of a waveform DB. FIG. 1 is an explanatory diagram showing an example of a result DB. Graphs showing an electrocardiogram, a cardiac sound waveform, and a pulse wave, as well as feature points of each waveform. FIG. 1 is an explanatory diagram showing an example of a training DB. FIG. 1 is an explanatory diagram showing a learning model. A flowchart showing an example of the procedure for generation processing. A flowchart showing an example of the procedure for monitoring. A flowchart showing an example of the procedure for measurement processing. A flowchart showing an example of the procedure for estimation processing. A flowchart showing an example of the procedure for intracardiac pressure estimation processing. A flowchart showing an example of the procedure for feature point extraction processing. A flowchart showing a correction method. A flowchart showing an example of the procedure for correction processing. A flowchart showing an example of the correction method. A diagram showing an example of the correction method. A flowchart showing an example of the procedure for result reference processing. A diagram showing an example of a result reference screen. A diagram showing another example of a learning model.

[0021] (Embodiment 1) An embodiment will be described below with reference to the drawings. FIG. 1 is an explanatory diagram showing an example of the configuration of a monitoring system 100. The monitoring system 100 is a system used for remote monitoring of patients diagnosed with heart failure. The monitoring system 100 includes a server 1, a measuring device 2, a communication terminal 3, and a medical professional terminal 4. The monitoring system 100 may transmit and receive patient-related data with an electronic medical record system 5. Although FIG. 1 shows two measuring devices 2 and two communication terminals 3, this is not limiting. There may be one measuring device 2 and one communication terminal 3, or three or more of each. Furthermore, although only one medical professional terminal 4 is shown in FIG. 1, there may be two or more medical professional terminals 4.

[0022] A patient who is a monitoring target using the monitoring system 100 performs measurements to evaluate cardiac function using the measuring device 2 at home, at work, or at a residence such as a nursing home. In this specification, the measuring device 2 synchronously measures an electrocardiogram, heart sounds, and pulse waves. When the patient visits a hospital or is hospitalized, the measuring device 2 installed in the medical institution may be used to measure the electrocardiogram, heart sounds, and pulse waves. It is desirable to measure the electrocardiogram, heart sounds, and pulse waves at least once a day.

[0023] The measurement data of the electrocardiogram, heart sounds, and pulse waves measured by the measurement device 2 is transmitted to and stored in the server 1 via the communication terminal 3 and the network N. The server 1 estimates the patient's intracardiac pressure based on the measurement data. The medical professional terminal 4 acquires and displays the patient's measurement data and estimated intracardiac pressure from the server 1. The communication terminal 3 may also estimate the patient's intracardiac pressure based on the measurement data.

[0024] Intracardiac pressure is an index of cardiac function. It is known that in heart failure, intracardiac pressure increases before the onset of subjective symptoms. Therefore, by monitoring intracardiac pressure, it is possible to detect the progression of heart failure at an earlier stage than the onset of subjective symptoms. Heart failure can be caused by myocardial infarction, angina pectoris, arteriosclerosis, hypertension, valvular disease, cardiomyopathy, arrhythmia, or the like. Examples of intracardiac pressure include, but are not limited to, left heart pressure, right heart pressure, left ventricular pressure waveform, right ventricular pressure waveform, left ventricular end-diastolic pressure (LVEDP), left atrium pressure (LAP), left ventricular pressure (LVP), pulmonary artery end-diastolic pressure (PAEDP, or pulmonary artery diastolic pressure (PADP)), right atrial pressure (RAP), right ventricular pressure (RVP), central venous pressure (CVP), pulmonary artery pressure (PAP), and pulmonary wedge pressure (PWP). Pulmonary artery occlusion pressure (PAOP) is also called pulmonary arterial wedge pressure (PAWP), pulmonary capillary wedge pressure (PCWP), or pulmonary artery occlusion pressure (PAOP). Intracardiac pressure varies periodically with each heartbeat, and therefore includes systolic pressure, diastolic pressure, and mean pressure.

[0025] The doctor refers to the measurement data and intracardiac pressure displayed on the medical professional terminal 4 to understand the patient's condition. If the doctor finds an out-of-hospital patient suspected of having worsening heart failure, he or she will recommend that the patient visit a doctor, change the medication prescription, or provide guidance on improving lifestyle habits, and will provide prompt treatment to outpatients or hospitalized patients. Paramedical staff under the doctor's instructions can check the measurement data and intracardiac pressure instead of the doctor and, if necessary, seek the doctor's judgment.

[0026] 2 is a block diagram showing an example of the hardware configuration of the server 1. The server 1 includes a control unit 11, a main memory unit 12, an auxiliary memory unit 13, a communication unit 14, and a reading unit 15. Each component is connected via a bus B. The server 1 is configured with a server computer, a workstation, a PC (Personal Computer), etc. The server 1 may also be configured with a multi-computer consisting of multiple computers, a virtual machine virtually constructed by software, or a quantum computer. Furthermore, the functions of the server 1 may be realized by a cloud service.

[0027] The control unit 11 has one or more arithmetic processing devices such as a central processing unit (CPU), a micro-processing unit (MPU), a graphics processing unit (GPU), etc. The control unit 11 reads and executes a control program 1P (program, program product) stored in the auxiliary storage unit 13, thereby performing various information processing, control processing, etc., and realizing various functional units.

[0028] The main memory unit 12 is a static random access memory (SRAM), a dynamic random access memory (DRAM), a flash memory, etc. The main memory unit 12 mainly temporarily stores data required for the control unit 11 to execute arithmetic processing.

[0029] The auxiliary storage unit 13 is a hard disk or a solid state drive (SSD), etc., and stores the control program 1P and various databases (DBs) required for the control unit 11 to execute processing. The auxiliary storage unit 13 stores a patient DB 131, a comorbidity DB 132, a waveform DB 133, and a result DB 134. The auxiliary storage unit 13 also stores a learning model M. The auxiliary storage unit 13 may be separate from the server 1 and may be an external storage device externally connected. The various DBs, etc. stored in the auxiliary storage unit 13 may be stored in a database server or cloud storage different from the server 1.

[0030] The communication unit 14 communicates with the communication terminal 3 and the medical professional terminal 4 via the network N. In addition, the control unit 11 may use the communication unit 14 to download the control program 1P from another computer via the network N or the like and store it in the auxiliary storage unit 13.

[0031] The reading unit 15 reads portable storage media 1a including CD (Compact Disc)-ROM and DVD (Digital Versatile Disc)-ROM. The control unit 11 may read the control program 1P from the portable storage medium 1a via the reading unit 15 and store it in the auxiliary storage unit 13. The control unit 11 may also read the control program 1P from the semiconductor memory 1b.

[0032] 3 is a block diagram showing an example of the hardware configuration of the measurement device 2. The measurement device 2 includes a sensor unit 21 and a processing unit 22. The sensor unit 21 includes an ECG (Electrocardiogram) sensor 211 that measures an electrocardiogram, a heart sound sensor 212 that measures heart sounds, and a pulse wave sensor 213 that measures a pulse wave.

[0033] The sensor unit 21 may, for example, house the ECG sensor 211, heart sound sensor 212, and pulse wave sensor 213 in one housing, or may house the ECG sensor 211 and pulse wave sensor 213 in one housing and the heart sound sensor 212 in another housing. By placing or fixing the sensor unit 21 near the heart on the patient's chest, it is possible to simultaneously measure an electrocardiogram, heart sounds, and pulse waves. If the sensor unit 21 does not have a housing, the patient places or fixes the ECG sensor 211, heart sound sensor 212, and pulse wave sensor 213 in pre-specified positions. The ECG sensor 211, heart sound sensor 212, and pulse wave sensor 213 are well known, so detailed description thereof will be omitted.

[0034] The processing unit 22 includes a control unit 221, a main memory unit 222, an auxiliary memory unit 223, and a communication unit 224. Each component is connected via a bus B. The control unit 221 has one or more arithmetic processing devices such as a central processing unit (CPU), a micro-processing unit (MPU), a graphics processing unit (GPU), etc. The control unit 221 reads and executes control programs stored in the auxiliary memory unit 223 to perform various information processing, control processing, etc., and realize various functional units.

[0035] The main memory unit 222 is an SRAM, a DRAM, a flash memory, etc. The main memory unit 222 mainly temporarily stores data necessary for the control unit 221 to execute arithmetic processing.

[0036] The auxiliary storage unit 223 is a hard disk or SSD, etc., and stores control programs and various data required for the control unit 221 to execute processing.

[0037] The communication unit 224 mainly communicates with the communication terminal 3. The communication unit 224 transmits measurement data to the communication terminal 3 via short-range wireless communication such as Bluetooth (registered trademark), WiFi ad-hoc mode, or ZigBee (registered trademark).

[0038] The measurement device 2 may be configured as an integrated unit with a health management device such as a blood pressure monitor or a weight scale. The measurement device 2 may be configured as an integrated unit with a so-called smart watch that is worn by the patient on a daily basis.

[0039] 4 is a block diagram showing an example of the hardware configuration of the communication terminal 3. The communication terminal 3 is configured by a smartphone, a smart display, a tablet computer, a notebook computer, etc. The communication terminal 3 includes a control unit 31, a main memory unit 32, an auxiliary memory unit 33, a communication unit 34, and a touch panel 35. Each component is connected by a bus B.

[0040] The control unit 31 has one or more arithmetic processing units such as a CPU, an MPU, a GPU, etc. The control unit 31 reads and executes control programs stored in the auxiliary storage unit 33 to perform various information processing, control processing, etc., and realize various functional units.

[0041] The main memory unit 32 is an SRAM, a DRAM, a flash memory, etc. The main memory unit 32 mainly temporarily stores data required for the control unit 31 to execute arithmetic processing.

[0042] The auxiliary storage unit 33 is a hard disk, an SSD, or the like, and stores control programs and various data required for the control unit 31 to execute processing.

[0043] The communication unit 34 communicates with the measurement device 2 and the server 1. The communication unit 34 communicates with the measurement device 2 by short-range wireless communication as described above. The communication unit 34 communicates with the server 1 via the network N.

[0044] The touch panel 35 includes a display unit 351 and an input unit 352. The display unit 351 includes a liquid crystal display panel or an organic EL (electro-luminescence) display panel. The input unit 352 is stacked on the display unit 351. The communication terminal 3 may include a keyboard and a mouse or a trackpad instead of or in addition to the touch panel 35.

[0045] Figure 5 is a block diagram showing an example of the hardware configuration of the medical professional terminal 4. The medical professional terminal 4 is configured with a notebook computer, tablet computer, smartphone, smart display, etc. The medical professional terminal 4 includes a control unit 41, a main memory unit 42, an auxiliary memory unit 43, a communication unit 44, a display unit 45, and an input unit 46. Each component is connected by a bus B.

[0046] The control unit 41 has one or more arithmetic processing units such as a CPU, an MPU, a GPU, etc. The control unit 41 provides various functions by reading and executing a control program 4P (program, program product) stored in the auxiliary storage unit 43.

[0047] The main memory unit 42 is an SRAM, a DRAM, a flash memory, etc. The main memory unit 42 mainly temporarily stores data necessary for the control unit 41 to execute arithmetic processing.

[0048] The auxiliary storage unit 43 is a hard disk or SSD, etc., and stores various data necessary for the control unit 41 to execute processing. The auxiliary storage unit 43 may be an external storage device that is separate from the medical professional terminal 4 and externally connected. The various DBs, etc. stored in the auxiliary storage unit 43 may be stored in a database server or cloud storage.

[0049] The control unit 41 communicates with the server 1 via the communication unit 44 and the network N. The control unit 41 may use the communication unit 44 to download the control program 4P from another computer via the network N or the like and store it in the auxiliary storage unit 43.

[0050] The display unit 45 includes a liquid crystal display panel or an organic EL display panel. The input unit 46 is a keyboard and a mouse. The display unit 45 displays the intracardiac pressure and other information output by the server 1. The display unit 45 and the input unit 46 may be integrated to form a touch panel display. The medical professional terminal 4 may display information on an external display device.

[0051] Next, the databases used by the monitoring system 100 will be described. FIG. 6 is an explanatory diagram showing an example of the patient DB 131. The patient DB 131 stores information about patients who are the subject of monitoring. The patient DB 131 includes a patient ID column, a name column, an attending physician column, a medical record ID column, a comorbidity column, and a constitution column. The patient ID column stores a patient ID that can uniquely identify a patient. The so-called My Number may be used as the patient ID. The My Number is a 12-digit personal number stipulated in the Act on the Use of Numbers to Identify Specific Individuals in Administrative Procedures in Japan. The name column stores the patient's name. The attending physician column stores the name of the patient's doctor. The medical record ID column stores a medical record ID that can identify the patient's medical record. The comorbidity column stores diseases (disease information) that the patient has other than heart failure. The constitution column stores the patient's constitution. FIG. 6 shows that a patient with patient ID P102 has aortic stenosis and hypertension as coexisting diseases and a frail constitution. While the names of diseases are stored in the coexisting diseases column in FIG. 6 , disease codes may be stored instead. Disease codes may be codes defined in, for example, the ICD-10 Standard Disease Name Master (Medical Information Systems Development Center, a general incorporated foundation) or the Illness and Injury Name Master for Electronic Receipt Processing (Social Insurance Medical Fee Payment Fund). Information regarding coexisting diseases and constitution is an example of health condition information. That is, in FIG. 6 , aortic stenosis, hypertension, and a frail constitution are examples of health condition information.

[0052] The information stored in the patient DB 131 may be obtained from the patient chart stored in the electronic medical chart system 5. Alternatively, the server 1 or the like may obtain patient information from the electronic medical chart system 5 without providing the patient DB 131.

[0053] FIG. 7 is an explanatory diagram showing an example of the comorbidity DB 132. The comorbidity DB 132 stores information about abnormal changes in the morphology of the cardiac sound waveform when a patient being monitored has a corresponding comorbidity. The comorbidity DB 132 includes a code string, a name string, a severity string, and a detection target string. The code string stores disease codes (in FIG. 7 , a master data set for disease name processing for medical receipts). The name string stores disease names. The severity string stores disease severity. The detection target string stores information about phenomena to be detected in the cardiac sound waveform, etc. FIG. 7 stores detection target codes assigned to phenomena to be detected. The detection target codes will be described later. Abnormal changes in the morphology of a patient's cardiac sound waveform are not limited to diseases; the patient's constitution may also be a cause. Information about constitutions that cause abnormal changes in the morphology of the cardiac sound waveform may also be stored in the comorbidity DB 132. When storing constitutions, the code column stores codes that are assigned in a system that does not overlap with disease codes. The name column stores the name of the constitution. The contents stored in the detection target column are the same as for diseases.

[0054] FIG. 8 is an explanatory diagram showing an example of the waveform DB 133. The waveform DB 133 stores waveform data from the measurement data obtained by the measurement device 2. The waveform data is, for example, waveform data of heart sounds, waveform data of electrocardiograms, and waveform data of pulse waves. The waveform DB 133 includes a measurement ID column, a patient ID column, a measurement date and time column, a type column, and a measurement data column. The measurement ID column stores a measurement ID that can identify each measurement. The measurement ID makes it possible to associate multiple measurement data measured simultaneously. The patient ID column stores the patient ID of the patient who was the subject of the measurement. The measurement date and time column stores the measurement date and time and the date and time the measurement started. In addition to the start date and time, the date and time the measurement ended may also be stored. The type column stores the type of measurement data. For example, HS indicates heart sounds, ECG indicates electrocardiograms, and PW indicates pulse waves. The measurement data column stores waveform data obtained by measurement. The waveform data is preferably in a general-purpose format such as CSV (Comma Separated Values) format or MFER (Medical Waveform Format Encoding Rules) format. Data for each point constituting the waveform data is associated with time information, enabling synchronization between multiple waveforms.

[0055] 9 is an explanatory diagram showing an example of the result DB 134. The result DB 134 stores feature points extracted from waveform data and intracardiac pressures estimated based on the feature points.

[0056] The feature points are points in the waveform data that should be noted in order to understand the movement of the heart. Fig. 10 is a graph showing an electrocardiogram, a phonocardiogram waveform, a pulse wave, and the feature points of each waveform. The feature points of the phonocardiogram waveform are the start point of the first sound, the peak point of the first sound, the start point of the second sound, and the second sound. A Point, and II P The first sound occurs during ventricular contraction and is mainly the mitral valve closing sound and the aortic opening sound. The second sound occurs at the beginning of ventricular diastole and is mainly the aortic valve closing sound (II sound). A ), pulmonary valve closure sound (II P In this specification, I represents the Roman numeral 1, and II represents the Roman numeral 2. The I syllable may be written as the first syllable, S1, and the II syllable as the second syllable, S2.

[0057] On an electrocardiogram, waveforms called P waves, QRS waves, T waves, and U waves are observed with each heartbeat. P waves represent atrial excitation (contraction). The first half of the P wave represents right atrial excitation, and the second half represents left atrial excitation; these combine to form the P wave. QRS waves represent ventricular excitation. The first downward waveform is called the Q wave, indicating the beginning of ventricular excitation. The following upward waveform is called the R wave, indicating the point at which the ventricles are most contracting. The final downward waveform is called the S wave, indicating the end of ventricular excitation. T waves represent the disappearance of ventricular excitation and a return to normal. T waves are characterized by their mountain-shaped appearance. U waves are small waveforms that appear after the T wave. U waves are sometimes not seen normally. The points of interest here are the Q point and the R point. The Q point indicates the beginning of the Q wave. The R point indicates the peak of the R wave.

[0058] The pulse wave is composed of components such as the ejection wave (PW: Percussion Wave) from the heart, the reflected wave (TW: Tidal Wave) from the blood vessel walls, etc., the dicrotic notch (DN), and the dicrotic wave (DW: Dicrotic Wave). Characteristic points on the pulse wave include the up-stroke (US) point, peak, and DN point. The US point is the point where the pulse wave rises. The peak is the point where the pulse wave is at its maximum. The DN point is the position of the notch.

[0059] The values ​​indicating the feature points of the cardiac sound waveform, the feature points of the electrocardiogram, and the feature points of the pulse wave are relative times indicating the elapsed time up to the feature point when a predetermined time point, for example, the start of the waveform, is set as time 0. If the time length between two feature points can be calculated, the value indicating each feature point may be an absolute time.

[0060] Returning to FIG. 9, the result DB 134 includes a measurement ID sequence, a beat number sequence, a feature point sequence, and an intracardiac pressure sequence. The measurement ID sequence stores the measurement ID. The beat number sequence stores the sequential number assigned to the data for one beat, which is obtained by dividing the measurement data into beats. In measurements using the measurement device 2, waveforms for multiple beats are acquired. Waveform data for one beat is used to estimate intracardiac pressure. Therefore, data for each beat is stored in the result DB 134. The feature point sequence stores information on feature points extracted from the waveform data. The feature point sequence includes the first sound start sequence, the first sound peak sequence, the second sound start sequence, the second sound peak sequence, the A tone sequence, II P These include the note sequence, Q sequence, R sequence, US sequence, peak sequence, and DN sequence. Each sequence stores the time when each feature point was measured. P The sound sequence stores three types of values: an uncorrected value, a corrected value, and a flag indicating whether or not the sound has been removed (removal flag). The intracardiac pressure sequence stores intracardiac pressure estimated based on feature points. The intracardiac pressure sequence includes a none sequence, a corrected sequence, and a removed sequence. In this embodiment, intracardiac pressure is estimated using three methods. The intracardiac pressure sequence stores values ​​estimated by each method. The none sequence stores intracardiac pressure values ​​estimated without performing feature point correction processing or removal processing. The corrected sequence stores intracardiac pressure values ​​estimated by performing only feature point correction processing. The removed sequence stores intracardiac pressure values ​​estimated by performing only feature point removal processing.

[0061] Next, the generation of the learning model M will be described. The learning model M is generated before the monitoring system 100 starts operation. FIG. 11 is an explanatory diagram showing an example of a training DB. The training DB stores training data used when generating the learning model M. The training DB includes a feature point sequence and an intracardiac pressure sequence. The feature point sequence stores the value of each feature point. The feature point sequence includes the first sound onset sequence, the first sound peak sequence, the second sound onset sequence, the second sound peak sequence, the first sound peak ... A tone sequence, II P These include a tone sequence, a Q sequence, an R sequence, a US sequence, a peak sequence, and a DN sequence. The values ​​stored in each sequence are the same as those stored in the result DB 134, so a description thereof will be omitted. The intracardiac pressure sequence stores the measured intracardiac pressure values ​​together with the values ​​of the feature points. The intracardiac pressure values ​​here are measured during cardiac catheterization.

[0062] A learning model M for estimating intracardiac pressure will be described. In this embodiment, the learning model M estimates intracardiac pressure using only feature points extracted from the phonocardiogram waveform. FIG. 12 is an explanatory diagram showing the learning model M. The learning model M receives feature points of the phonocardiogram waveform as input, and estimates and outputs an estimated value of intracardiac pressure based on the received feature points. The learning model M outputs, for example, an estimated representative value of intracardiac pressure in one heartbeat. The representative value is, for example, the diastolic intracardiac pressure, the systolic intracardiac pressure, and the average value of the intracardiac pressure. The diastolic intracardiac pressure is the minimum value of the intracardiac pressure. The systolic intracardiac pressure is the maximum value of the intracardiac pressure. A learning model M that outputs an estimated value of the diastolic intracardiac pressure, a learning model M that outputs an estimated value of the systolic intracardiac pressure, and a learning model M that outputs an estimated value of the average value of the intracardiac pressure may be provided. The learning model M may also estimate the time course of the intracardiac pressure and output it as a waveform.

[0063] FIG. 13 is a flowchart showing an example of the generation process. The generation process generates a learning model M using a training DB. As described above, since the generation of the learning model M is assumed to be performed before the start of operation, it is assumed that a computer other than the server 1 performs the generation process. Here, the description is given assuming that the generation process is performed by the generation computer. The generation computer acquires one record of training data from the training DB (step S1). The generation computer inputs the values ​​of feature points contained in the training data into the learning model M and acquires the intracardiac pressure output by the learning model M (step S2). The generation computer compares the intracardiac pressure contained in the training data with the intracardiac pressure output by the learning model M and adjusts parameters such as weights between neurons constituting the learning model M so as to minimize the difference between the two values ​​(step S3). A technique such as backpropagation is used to adjust the parameters. The generation computer determines whether there is unprocessed training data (step S4). If the generation computer determines that there is unprocessed training data (YES in step S4), the process returns to step S1 and processes the unprocessed training data. If the generation computer determines that there is no unprocessed training data (NO in step S4), it stores the adjusted parameters (step S5) and ends the process. The parameters characterizing the learning model M are stored in the auxiliary storage unit 13 of the server 1.

[0064] Next, patient monitoring using the monitoring system 100 will be described. FIG. 14 is a flowchart showing an example of the monitoring procedure. The patient uses the measuring device 2 to measure an electrocardiogram, heart sounds, and pulse waves (step S11). The waveform obtained as the measurement result is sent to the server 1 and stored in the waveform DB 133. The control unit 11 of the server 1 extracts feature points from the heart sound waveform (step S12). The control unit 11 estimates the intracardiac pressure (step S13). The control unit 11 inputs the extracted feature points into the learning model M and obtains an estimated value of intracardiac pressure output by the learning model M. The control unit 11 associates the estimated value with the feature points and stores it in the result DB 134. In response to a doctor's instruction, the medical professional terminal 4 displays the heart sound waveform, feature points, and estimated intracardiac pressure. The doctor refers to the displayed content and takes the necessary action (step S14). The control unit 11 stores the action content, etc. (step S15). Monitoring then ends. The doctor may take measures such as sending the patient a message recommending that they visit a doctor and undergo a detailed examination, or changing the prescription of the medicine that has been prescribed to the patient.

[0065] FIG. 15 is a flowchart showing an example of the procedure for the measurement process. The measurement process is initiated when a measurement command is issued after the sensor unit 21 of the measurement device 2 has been placed or fixed on the patient's body. The control unit 221 of the measurement device 2 synchronously measures an electrocardiogram, heart sounds, and pulse waves via the sensor unit 21 (step S21). The control unit 221 determines whether the measurement has ended normally (step S22). The control unit 221 determines that the measurement has not ended normally, for example, in the following cases: when the sensor unit 21 has become detached from the body during measurement, or when a predetermined percentage (e.g., 90%) or more of the measurement data has not been obtained. Note that, since the function of determining whether the measurement has ended normally is well known in medical measurement devices such as electrocardiographs, a detailed description of the determination method will be omitted.

[0066] If the control unit 221 determines that the measurement has ended normally (YES in step S22), it transmits the measurement data to the communication terminal 3 (step S23). Then, the control unit 221 terminates the process. The control unit 31 of the communication terminal 3 receives the measurement data (step S24). The control unit 31 adds information such as the patient ID and the measurement date and time to the measurement data and transmits it to the server 1 (step S25). The server 1 stores the received information in the waveform DB 133.

[0067] If the control unit 221 determines that the measurement has not ended normally (NO in step S22), it determines whether or not to perform remeasurement (step S26). For example, if the measurement has already been repeated a predetermined number of times or more, the control unit 221 determines not to perform remeasurement. If the control unit 221 determines that remeasurement should be performed (YES in step S26), it returns the process to step S21.

[0068] If the control unit 221 determines not to perform remeasurement (NO in step S26), it sends a notification to the communication terminal 3 that the measurement was not performed normally (step S27). Then, the control unit 221 ends the processing. The control unit 31 of the communication terminal 3 receives the notification (step S28). The control unit 31 displays the notification on the display unit 351 (step S29). Note that the control unit 221 of the measurement device 2 may determine whether the measurement is being performed normally for each heartbeat, and send a notification to the communication terminal 3 if it determines that the measurement is not being performed normally.

[0069] FIG. 16 is a flowchart showing an example of the procedure for the estimation process. The estimation process is a process for estimating intracardiac pressure from the measurement results of the measurement device 2. The control unit 11 of the server 1 acquires one unprocessed waveform data from the waveform DB 133 (step S41). The control unit 11 divides the waveform data into pieces for each heartbeat (step S42). The control unit 11 estimates intracardiac pressure using the data for each beat (step S43). The control unit 11 determines whether to end the process (step S44). If the control unit 11 determines not to end the process (NO in step S44), the process returns to step S41. If the control unit 11 determines to end the process (YES in step S44), the control unit 11 stores the estimated value of intracardiac pressure and the like in the result DB 134 (step S45) and ends the process.

[0070] FIG. 17 is a flowchart showing an example of the procedure for intracardiac pressure estimation. The intracardiac pressure estimation process corresponds to step S43 in FIG. 16. The control unit 11 of the server 1 determines whether there is a measurement abnormality in the waveforms of the electrocardiogram, heart sounds, and pulse wave for one heartbeat (step S61). Waveform measurement abnormalities refer to abnormalities caused by the measurement conditions. For example, the control unit 11 performs pattern matching between the actually measured waveform and each of the abnormal waveform samples measured when various abnormalities occur in the measurement, and determines that there is an abnormality if the waveform is similar to any of the abnormal waveform samples. If the control unit 11 determines that there is no abnormality in any of the waveforms (NO in step S61), it extracts feature points (step S62). The control unit 11 inputs the feature points into the learning model M and obtains the intracardiac pressure output from the learning model M (step S63). The control unit 11 stores the intracardiac pressure in the result DB 134 (step S64). If the control unit 11 determines that there is a measurement abnormality (YES in step S61), it stores the measurement abnormality in the result DB 134 (step S66). After step S64 or step S66 is completed, it determines whether or not processing of the data acquired in one measurement is completed (step S65). If the control unit 11 determines that processing is not completed (NO in step S65), it returns the processing to step S61. If the control unit 11 determines that processing is completed (YES in step S65), it returns the processing to the caller.

[0071] FIG. 18 is a flowchart showing an example of the procedure for the feature point extraction process. The feature point extraction process corresponds to step S62 in FIG. 17. The control unit 11 extracts feature points from the waveforms of the electrocardiogram, heart sounds, and pulse wave to be processed and stores the extracted feature point values ​​in the result DB 134 (step S81). The control unit 11 acquires information on comorbidities of the patient to be processed from the patient DB 131 (step S82). The control unit 11 compares the acquired comorbidity information with the comorbidity DB 132 to determine whether a phenomenon to be detected exists (step S83). If the control unit 11 determines that a phenomenon to be detected exists (YES in step S83), it selects a detection target stored in the detection target column of the comorbidity DB 132 (step S84). The control unit 11 determines whether correction is necessary (step S85). If the control unit 11 determines that correction is not necessary (NO in step S85), the process proceeds to step S88. If the control unit 11 determines that correction is necessary (YES in step S85), it performs the correction (step S86). The control unit 11 extracts feature points based on the correction results and stores the extraction results in the result DB 134 (step S87). The control unit 11 removes the feature points (step S88). The control unit 11 sets a flag in the result DB 134 for the feature points to be removed. The control unit 11 determines whether or not there are any unprocessed detection targets (step S89). If the control unit 11 determines that there are any unprocessed detection targets (YES in step S89), it returns the process to step S84. If the control unit 11 determines that there are no phenomena to be detected (NO in step S83) or that there are no unprocessed detection targets (NO in step S89), it returns the process to the caller.

[0072] Next, correction of extraction conditions for each detection target and removal of feature points will be described. The detection targets are the following eight types: accentuated I (1A), attenuated I (1B), ejected I (split) (1C), accentuated II (2A), attenuated II (2B), split II (2C), paradoxical split II (reversed) (2D), an outlier in the I-I interval or II-II interval, or accentuated I (3). The code combination of a number and an alphabetic character in parentheses at the end of each type, or the numeric code, corresponds to the detection target code stored in the detection target column of the comorbidity DB 132 shown in FIG. 7.

[0073] Correction of the extraction conditions for a split I sound ejection sound (1C) or a split II sound (2C) will be described. Correction is made to the onset of I sound in 1C, and to the onset of II sound in 2C. FIG. 19 is an explanatory diagram showing the correction method. Hereinafter, the area above the horizontal line in FIG. 19 will be referred to as the upper section, and the area below the line will be referred to as the lower section. The upper section of FIG. 19 shows an example of a normal cardiac sound waveform. The method for detecting the onset of I sound in a normal waveform is as follows: Step 1: Detect the R wave on the electrocardiogram. Step 2: Detect the peak of the cardiac sound waveform (I sound peak) at a predetermined time (for example, within x milliseconds) after the R wave is detected. Step 3: Count back in time from the peak and determine the start of the amplitude as the I sound onset. In FIG. 19, the position of the I sound onset is indicated by a solid line that straddles the waveform vertically, and the I sound peak is indicated by a dotted line that straddles the waveform vertically. The detection of the II sound onset is similar to the I sound onset.

[0074] The bottom row of Figure 19 shows a cardiac waveform indicating a split I sound ejection sound. Although three waveforms are shown, they are identical. The top waveform shows an example in which the start of the I sound is incorrectly determined when detecting the start of the I sound using the algorithm described above. Because the I sound is split, the start of the I sound is searched for from the latter peak that appears after the true I sound peak (the falsely detected I sound peak, indicated by the dotted line straddling the waveform above and below). Therefore, the amplitude start point between the true I sound peak and the latter peak is incorrectly detected as the start of the I sound (the falsely detected I sound start point, indicated by the solid line straddling the waveform above and below). The start of the II sound is also incorrectly detected. The middle waveform shows correction method 1. A search is performed under the assumption that two or more waveform peaks will appear due to splitting (a waveform peak (second peak) will appear in addition to the true first sound peak (first peak)), and the correct position can be selected by setting the amplitude start point before the first peak (true first sound peak) as the first sound start point (true first sound start point, indicated by the solid line straddling the waveform above and below). The waveform below shows correction method 2. By setting the time range for peak detection to y milliseconds, which is shorter than x milliseconds, it is possible to capture the first peak (true first sound peak) and correctly detect the first sound start point (true first sound start point, indicated by the solid line straddling the waveform above and below). The second sound start point can also be correctly extracted using a method similar to that used for the first sound start point.

[0075] The correction of the extraction conditions for the paradoxical II split (reversal) (2D) will be explained. In the paradoxical II split (reversal), II A and II P It is observed that the order in which the first sound and the second sound appear is reversed, and the start of the second sound is delayed during inspiration. This point is taken into consideration when making corrections. FIG. 20 is a flowchart showing an example of the procedure for correction processing. The control unit 11 of the server 1 analyzes the waveform of the pulse wave and determines whether or not a notch can be obtained (step S101). If the control unit 11 determines that a notch can be obtained (YES in step S101), it compares the notch and the second sound. A The control unit 11 determines whether or not the notch and II match (step S102). A If it is determined that they do not match (NO in step S102), AThe control unit 11 replaces the notch with II (step S103), and the process ends. A If it is determined that the two match (YES in step S102), the processing ends. If the control unit 11 determines that the notch cannot be acquired (NO in step S101), it determines whether or not respiration determination is possible (step S104). If the measurement device 2 is equipped with a respiration sensor and a respiration waveform can be acquired, or if respiratory fluctuations can be acquired by analyzing an electrocardiogram, the control unit 11 determines that respiration determination is possible. If the control unit 11 determines that respiration determination is possible (YES in step S104), it determines whether or not the start of the second sound is delayed by inspiration (step S105). If the control unit 11 determines that the start of the second sound is delayed by inspiration (YES in step S105), it determines whether or not the second sound is delayed by inspiration (YES in step S105). P and II A The control unit 11 reverses the order of appearance of the first sound and the second sound (step S106), and then ends the process. If the control unit 11 determines that breathing detection is impossible (NO in step S104), or if it determines that the start of the second sound is not delayed by just the amount of inspiration (NO in step S105), it ends the process.

[0076] Correction of extraction conditions for an accentuated I sound (1A) and an accentuated II sound (2A) will be described. FIG. 21 is an explanatory diagram showing an example of the correction method. In the case of an accentuated I sound and an accentuated II sound, the threshold for determining noise is key. In FIG. 21 , the phonocardiogram waveform shown in the upper row of the column labeled "normal" is an example of a waveform measured without noise or patient movement. The phonocardiogram waveform shown in the lower row is an example of a waveform with noise due to external disturbances or patient movement. The two dotted lines indicate the threshold for determining noise. The portion of the waveform whose amplitude exceeds the upper line is determined to be noise due to body movement, etc. The portion of the waveform whose amplitude exceeds the lower line is also determined to be noise. The middle column of FIG. 21 (the column labeled "accentuated I sound" and "accentuated II sound") is an example of a phonocardiogram waveform with an accentuated I sound and an accentuated II sound. In the case of an accentuated I sound and an accentuated II sound, even if there is no noise, the amplitude is large and the waveform will be determined to be noise. Therefore, as shown in the right column of FIG. 21 (the column labeled "correction method"), the threshold value for noise determination is set wide to prevent part of the waveform from being determined as noise, and then feature points are extracted.

[0077] Correction of extraction conditions for attenuated I sound (1B) and attenuated II sound (2B) will be described. FIG. 22 is an explanatory diagram showing an example of the correction method. In the cases of attenuated I sound and attenuated II sound, the threshold for determining noise is also key. In FIG. 22 , the phonocardiogram waveform shown in the upper row of the column labeled "normal" is an example of a waveform measured without baseline noise or environmental sound. The phonocardiogram waveform shown in the lower row is an example of a waveform containing baseline noise and environmental sound. The two dotted lines indicate the threshold for determining noise. A waveform whose amplitude is sandwiched between the two lines (falls between the two lines) is determined to be noise. The middle column of FIG. 22 (the column labeled "attenuated I sound" and "attenuated II sound") is an example of a phonocardiogram waveform in the case of attenuated I sound and attenuated II sound. In the case of attenuated I sound and attenuated II sound, the amplitude is small, so it is determined to be noise. Therefore, as shown in the right column of FIG. 22 (the column labeled "correction method"), the threshold width for noise determination is set narrow to prevent the waveform from being determined as noise, and then feature points are extracted.

[0078] If the I-I interval or II-II interval is an outlier, or if the I sound is accentuated (3), no correction is performed, and only feature points are removed. An I-I interval or II-II interval is an outlier means that if some of the beats being measured include a beat that is not in sinus rhythm due to arrhythmia and is out of time, that beat is compared with the intervals of other beats and detected as an outlier beat. Furthermore, for beats that are not in sinus rhythm due to arrhythmia, the I sound of the following beat may be accentuated.

[0079] The removal of feature points will be explained. In the cases of I sound accentuation (1A), I sound attenuation (1B), and I sound ejection (split) (1C), I sound-related feature points, i.e., I sound onset and I sound peak, are removed. In the cases of II sound accentuation (2A), II sound attenuation (2B), II sound split (2C), and II sound paradoxical split (reversed) (2D), II sound-related feature points, i.e., II sound onset and II sound peak, are removed. A Point, and II PRemove the feature points. If the I-I interval or the II-II interval is an outlier, or if the I sound is accentuated (3), remove the outliers of the I-I interval, the outliers of the II-II interval, and the accentuated I sound beat. Removal means that the feature points for which the removal flag is set in DB 134 are not used in the intracardiac pressure estimation process. Therefore, the values ​​of the removed feature points are not used as input to the learning model M in the intracardiac pressure estimation process.

[0080] Next, the process of result reference by medical professionals will be described. Figure 23 is a flowchart showing an example of the procedure for result reference processing. At a predetermined timing, the doctor operates the medical professional terminal 4 to begin referencing the estimated intracardiac pressure results. The control unit 41 of the medical professional terminal 4 sends a request for a list of patients to be monitored to the server 1 (step S121). The control unit 11 of the server 1 receives the request (step S122). The control unit 11 creates a patient list by referencing the patient DB 131 and sends it to the medical professional terminal 4 (step S123). The control unit 41 of the medical professional terminal 4 receives the patient list and displays it on the display unit 45 (step S124). The doctor selects a patient to reference from the patient list. The control unit 41 accepts the selection (step S125). The control unit 41 sends selection information, such as the patient ID of the selected patient, to the server 1 (step S126). The control unit 11 of the server 1 receives the selection information (step S127). The control unit 11 obtains the waveform of the measurement results for the selected patient from the waveform DB 133 and the characteristic points and estimated values ​​of intracardiac pressure from the result DB 134 (step S128). The control unit 11 creates a reference screen (step S129). The control unit 11 sends the created reference screen to the medical professional terminal 4 (step S130). The control unit 41 of the medical professional terminal 4 receives the reference screen (step S131). The control unit 41 displays the reference screen on the display unit 45 (step S132) and ends the process.

[0081] FIG. 24 is an explanatory diagram showing an example of a result reference screen. The result reference screen d01 includes patient information d011, phonocardiogram waveform d012, and intracardiac pressure d013. The patient information d011 includes the patient ID, patient name, comorbidities, and measurement date and time. The phonocardiogram waveform d012 and intracardiac pressure d013 each include three display patterns. While the phonocardiogram waveform d012 itself is the same for all patterns, the positions of the superimposed feature points and the display of feature points removed during intracardiac pressure estimation differ. The top row shows results without feature point correction or removal. The middle row shows results with feature point correction. The bottom row shows results with feature point removal. Although not shown, a recommended ranking of the three intracardiac pressure d013 patterns may also be displayed. The recommended ranking indicates the priority order of which estimated value of intracardiac pressure d013 is most likely to be closest to the true value. The recommended ranking is determined based on empirical rules from past data on which method tends to be closest to the true value for each disease or constitution. For example, in the case of aortic stenosis, if the estimated value with correction is often closest to the true value, the recommended value with correction will be displayed higher. The doctor refers to the three patterns of results on the result reference screen, checks the position of the feature points superimposed on the cardiac waveform, and determines which pattern is most likely, and then refers to the estimated intracardiac pressure to determine the patient's condition. Then, if necessary, the doctor takes action such as notifying the patient to visit a doctor or changing the medication prescription. The doctor may refer to the recommended rankings described above when determining which pattern is most likely.

[0082] This embodiment has the following advantages: When extracting feature points from the phonocardiogram waveform, the extraction conditions are changed, such as by correction or removal, based on the type of cause (comorbidity or constitution) that causes abnormal changes in the phonocardiogram waveform, enabling appropriate extraction.

[0083] Furthermore, when estimating intracardiac pressure using feature points, intracardiac pressure is estimated using feature points extracted without changing the extraction conditions (no correction or removal) and feature points extracted after changing the extraction conditions (correction and removal). The cardiac sound waveform, feature points, and intracardiac pressure are then displayed in association with each other for each case. This allows doctors to confirm the most likely of multiple cardiac sound waveform and intracardiac pressure estimation results, enabling them to accurately assess the patient's condition.

[0084] This allows appropriate treatment to be administered before the patient experiences any symptoms, preventing the onset of acute heart failure. Furthermore, because patients can receive medical advice from their doctor before they experience any symptoms, they can live their daily lives with peace of mind, improving their quality of life (QOL).

[0085] (Embodiment 2) In embodiment 1, in estimating intracardiac pressure, the input to the learning model M was limited to the feature points of the cardiac sound waveform. In this embodiment, indices obtained from the electrocardiogram and pulse waveform are added as input to the learning model M. In the following explanation, explanations of the same content as in embodiment 1 will be omitted, and differences from embodiment 1 will be mainly explained.

[0086] In this embodiment, the monitoring procedure shown in FIG. 14 is as follows: The patient uses the measurement device 2 to synchronously measure an electrocardiogram, heart sounds, and pulse waves (step S11). The measurement data of the electrocardiogram, heart sounds, and pulse waves measured by the measurement device 2 is transmitted to the server 1 via the communication terminal 3 and the network N and stored in the waveform DB 133. The control unit 11 of the server 1 extracts feature points (step S12). In this embodiment, in addition to the feature points of the phonocardiogram waveform shown in embodiment 1, the Q point of the electrocardiogram and the US point and DN point of the pulse wave are extracted. The control unit 11 estimates intracardiac pressure (step S13). The control unit 11 inputs the extracted feature points and index values ​​calculated from the feature points of the electrocardiogram and pulse wave into the learning model M, and obtains an estimated value of intracardiac pressure output by the learning model M. The index values ​​will be described later. The control unit 11 associates the estimated values ​​with the feature points and stores them in the result DB 134. The result DB 134 may also store the index values. The subsequent steps are the same as those in the first embodiment, and therefore the explanation will be omitted.

[0087] FIG. 25 is an explanatory diagram showing another example of the learning model M. The learning model M receives three indices, PEP, PTT, and STI, as input in addition to the feature points of the cardiac sound waveform. The learning model M estimates and outputs an estimated value of intracardiac pressure based on the received feature points and indices. Note that the indices used to estimate intracardiac pressure are not limited to those shown here. Other indices that can derive feature points of the electrocardiogram, cardiac sounds, and pulse wave may also be used.

[0088] The three indices are as shown in FIG. 10 . PEP (Pre-Ejection Period) is the elapsed time from point Q on the electrocardiogram to the peak of the first sound (S1) on the phonocardiogram waveform. PTT (Pulse Transition Time) is the elapsed time from point Q to point US. STI (Systolic Time) is the elapsed time from point US to point DN. The indices indicate the temporal relationship (time difference) between characteristic points of at least two of the electrocardiogram data, phonocardiogram data, and pulse wave data, or the temporal relationship (time difference) between characteristic points of any one of the electrocardiogram data, phonocardiogram data, and pulse wave data. Because all of these indices are commonly used in the field of cardiovascular medicine, their medical meanings will not be explained here.

[0089] The generation process of the learning model M used in this embodiment will be described with reference to FIG. 13 . The generation computer 11 acquires one record of training data from the training DB (step S1). The generation computer inputs the values ​​of the feature points included in the training data and the values ​​of the indicators PEP, PTT, and STI into the learning model M and acquires the intracardiac pressure output by the learning model M (step S2). The generation computer compares the intracardiac pressure included in the training data with the intracardiac pressure output by the learning model M and adjusts parameters such as weights between neurons constituting the learning model M so as to minimize the difference between the two values ​​(step S3). The generation computer determines whether there is unprocessed training data (step S4). If the generation computer determines that there is unprocessed training data (YES in step S4), it returns the process to step S1 and processes the unprocessed training data. If the generation computer determines that there is no unprocessed training data (NO in step S4), it stores the adjusted parameters (step S5) and ends the process. Although the values ​​of the indicators PEP, PTT, and STI are not stored in the training DB shown in FIG. 11, the values ​​may be calculated when the data is created and stored in the training DB.

[0090] The measurement process and estimation process in this embodiment are similar to those described with reference to FIGS. 15 and 16 in the first embodiment, and therefore will not be described again.

[0091] The intracardiac pressure estimation process in this embodiment will be described with reference to FIG. 17 . The control unit 11 of the server 1 determines whether there is a measurement abnormality in the synchronized electrocardiogram, heart sounds, and pulse wave waveforms for one heartbeat (step S61). If the control unit 11 determines that there is no abnormality in any of the waveforms (NO in step S61), it performs feature point extraction (step S62). The control unit 11 inputs the values ​​of the feature points of the heart sound waveform and the values ​​of the indices PEP, PTT, and STI into the learning model M to obtain the intracardiac pressure output from the learning model M (step S63). The control unit 11 stores the intracardiac pressure in the result DB 134 (step S64). If the control unit 11 determines that there is a measurement abnormality (YES in step S61), it stores the measurement abnormality in the result DB 134 (step S66). After step S64 or step S66 is completed, it determines whether processing of the data obtained in one measurement has been completed (step S65). If the control unit 11 determines that the process has not ended (NO in step S65), the process returns to step S61. If the control unit 11 determines that the process has ended (YES in step S65), the process returns to the caller.

[0092] Correction of extraction conditions and removal of feature points are the same as in embodiment 1, but in this embodiment, the following processing is added. If the value of the first sound peak point changes due to correction of the feature point extraction conditions, the value of PEP also changes. Furthermore, if the first sound peak point is removed during feature point removal, the value of PEP becomes indefinite and is not input to the learning model M.

[0093] In addition to the effects of the first embodiment, the present embodiment has the following effects: By using the indices PEP, PTT, and STI in addition to the feature points obtained from the cardiac sound waveform to estimate the intracardiac pressure, it is possible to improve the accuracy of estimating the intracardiac pressure.

[0094] In the above-described embodiment, the learning model M is provided in the server 1, but it may also be provided in the communication terminal 3 or the medical practitioner terminal 4. In this case, the communication terminal 3 or the medical practitioner terminal 4 reads out the waveform data and executes the estimation process. At this time, the estimation process may be incorporated into the result reference process, and the communication terminal 3 or the medical practitioner terminal 4 may perform the estimation process for a patient selected by a doctor and display the results on the screen.

[0095] The measuring device 2 may also be equipped with the learning model M. In this case, after measuring an electrocardiogram, heart sounds, and pulse, the measuring device 2 performs an estimation process. The measuring device 2 transmits the waveform data and the estimation result to the server 1 via the communication terminal 3. Alternatively, the communication terminal 3 may be equipped with the learning model M. In this case, after measuring an electrocardiogram, heart sounds, and pulse, the communication terminal 3 performs an estimation process. The communication terminal 3 transmits the waveform data and the estimation result to the server 1.

[0096] The technical features (constituent elements) described in each embodiment can be combined with each other, and by combining them, new technical features can be formed. The embodiments disclosed herein are illustrative in all respects and should be considered not to be limiting. The scope of the present invention is indicated by the claims, not by the meaning described above, and is intended to include all modifications within the meaning and scope of the claims. Furthermore, although the claims use a form in which a claim cites two or more other claims (multiple claim form), this is not limited to this. A form in which a multiple claim cites at least one other multiple claim (multi-multi claim) may also be used.

[0097] 100 Monitoring system 1 Server 11 Control unit 12 Main memory unit 13 Auxiliary memory unit 131 Patient DB 132 Comorbid disease DB 133 Waveform DB 134 Result DB M Learning model 14 Communication unit 15 Reading unit 1P Control program 1a Portable storage medium 1b Semiconductor memory 2 Measuring device 21 Sensor unit 211 ECG sensor 212 Heart sound sensor 213 Pulse wave sensor 22 Processing unit 221 Control unit 222 Main memory unit 223 Auxiliary memory unit 224 Communication unit 3 Communication terminal 31 Control unit 32 Main memory unit 33 Auxiliary memory unit 34 Communication unit 35 Touch panel 351 Display unit 352 Input unit 4 Medical professional terminal 41 Control unit 42 Main memory unit 43 Auxiliary memory unit 44 Communication unit 45 Display unit 46 Input unit 4P Control program 5 Electronic medical record system B Bus N Network

Claims

1. An information processing program that causes a computer to execute the following processes: acquiring health condition information including disease information of a patient; identifying the disease or constitution based on the health condition information; acquiring the phonocardiogram of the patient; setting extraction conditions for feature points of the phonocardiogram based on the disease or constitution; and extracting feature points of the phonocardiogram based on the set extraction conditions.

2. The information processing program according to claim 1, wherein the extraction condition is to recognize that there are two or more peaks in the cardiac sound waveform and select the first peak, or to change the time range for extracting the feature points.

3. The information processing program according to claim 1, wherein the extraction condition is to change the order in which the feature points appear or to replace the feature points with other feature points.

4. The information processing program according to claim 1, wherein the extraction condition changes the amplitude threshold for determining noise in the cardiac sound waveform.

5. The information processing program according to claim 1, wherein the feature points are discarded.

6. The information processing program according to any one of claims 1 to 5, wherein the health condition information is obtained from an electronic medical record.

7. The information processing program according to any one of claims 1 to 5, wherein the diseases include valvular disease or arrhythmia.

8. An information processing program according to any one of claims 1 to 5, which estimates an intracardiac pressure based on the extracted feature points, and outputs the estimated intracardiac pressure.

9. An information processing program as claimed in any one of claims 1 to 5, which acquires an electrocardiogram and pulse waveforms synchronized with a cardiac sound waveform, extracts feature points of the acquired electrocardiogram and pulse waveform, and estimates intracardiac pressure based on the feature points of the cardiac sound waveform and the feature points of the electrocardiogram and pulse waveform.

10. The information processing program of claim 9, wherein the disease causes an abnormal change in the shape of the cardiac sound waveform, the abnormal change including the splitting of one or two heart sounds in the cardiac sound waveform, and the extraction conditions are set to extract the start points of the first sound and the second sound on the premise that two or more waveform peaks appear due to the splitting, or to extract the start points of the first sound and the second sound from the R wave extracted from the electrocardiogram within a time range that takes into account the splitting.

11. An information processing program as described in claim 9, which estimates intracardiac pressure based on the characteristic points of the phonocardiogram waveform and the characteristic points of the electrocardiogram and pulse wave waveform extracted assuming that there is no disease or predisposition that causes abnormal changes in the morphology of the phonocardiogram waveform, and outputs the estimated intracardiac pressure.

12. An information processing method in which a computer executes the following processes: acquiring health condition information including disease information of a patient; identifying the disease or constitution based on the health condition information; acquiring the acoustocardiogram of the patient; setting extraction conditions for feature points of the acoustocardiogram based on the disease or constitution; and extracting feature points of the acoustocardiogram based on the set extraction conditions.

13. An information processing device having a control unit, wherein the control unit executes the following processes: acquire health condition information including disease information of a patient; identify a disease or constitution based on the health condition information; acquire the acoustocardiogram of the patient; set extraction conditions for feature points of the acoustocardiogram based on the disease or constitution; and extract feature points of the acoustocardiogram based on the set extraction conditions.

Images