Premature contraction detection device, premature contraction detection method, premature contraction detection program

Abstract

The system accurately detects the period during which premature contractions occur from images of the heart. [Solution] The premature contraction detection device includes: an acquisition unit (401) that acquires a target signal waveform that shows the time-series change in the area of ​​at least one of the regions of the left atrium, left ventricle, right atrium, and right ventricle, generated by analyzing an image of the subject's heart, and includes a plurality of target partial waveforms corresponding to one or more predetermined number of heartbeats of the subject's heart; a similarity calculation unit (403) that calculates a first similarity based on comparison information including the result of comparing the shapes of the preceding number of target partial waveforms included in the target signal waveform; and a premature contraction detection unit (404) that, if there is a particular partial waveform whose first similarity is below a predetermined standard, detects the period corresponding to the particular partial waveform in the target signal waveform as a period of interest during which premature contractions are occurring in the subject's heart.

Description

[Technical Field]

[0001] The present invention relates to a technique for detecting extrasystole. [Background Art]

[0002] Techniques for grasping the relationship between the left and right sides of the heart and the relationship between the atria and ventricles from ultrasonic diagnostic moving images of the heart are widely used. For example, Patent Document 1 discloses an ultrasonic diagnostic apparatus that identifies boundary position groups in a plurality of heart chambers in an ultrasonic moving image, and acquires boundary positions of the plurality of heart chambers over a section of at least one heartbeat or more based on the tracking result of the boundary position groups. [Prior Art Documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Unexamined Patent Application Publication No. 2022-149097 [Summary of the Invention] [Problem to be Solved by the Invention]

[0004] Conventionally, it has been difficult to detect a period during which extrasystole occurs. Extrasystole is a type of arrhythmia in which abnormal stimulation occurs in the heart, and beats caused by additional contraction are added in addition to normal beats.

[0005] An object of one aspect of the present invention is to accurately detect a period in which extrasystole occurs from an image obtained by imaging a heart. [Means for Solving the Problem]

[0006] An out-of-body contraction detection device according to one aspect of the present invention includes: an acquisition unit that acquires a target signal waveform that shows the time-series change in the area of ​​at least one of the regions of the left atrium, left ventricle, right atrium, and right ventricle, generated by analyzing an image of the heart of a subject, and includes a plurality of target partial waveforms corresponding to one or a predetermined number of heartbeats of the subject's heart; a similarity calculation unit that calculates a first similarity based on comparison information including the result of comparing the shapes of each of the plurality of target partial waveforms included in the target signal waveform; and an out-of-body contraction detection unit that, if there is a particular partial waveform whose first similarity is below a predetermined standard, detects the period corresponding to the particular partial waveform in the target signal waveform as a period of interest during which out-of-body contractions occur in the subject's heart.

[0007] A premature contraction detection method according to one aspect of the present invention is a premature contraction detection method performed by one or more information processing devices, comprising: an acquisition step of acquiring a target signal waveform that shows the time-series change in the area of ​​at least one of the regions of the left atrium, left ventricle, right atrium, and right ventricle, generated by analyzing an image of the heart of a subject, and includes a plurality of target partial waveforms corresponding to one or more predetermined number of heartbeats of the heart of the subject; a similarity calculation step of calculating a first similarity based on comparison information including the result of comparing the shapes of each of the plurality of target partial waveforms included in the target signal waveform; and a premature contraction detection step of detecting, if there is a particular partial waveform whose first similarity is below a predetermined standard, the period corresponding to the particular partial waveform in the target signal waveform as a period of interest during which premature contractions occur in the heart of the subject.

[0008] Each aspect of the present invention may be implemented by a computer. In this case, a control program for the premature contraction detection device, which enables the computer to implement the premature contraction detection device by operating the computer as each part (software element) of the premature contraction detection device, and a computer-readable recording medium on which the program is recorded, also fall within the scope of the present invention. [Effects of the Invention]

[0009] According to one aspect of the present invention, the period during which premature contractions occur can be accurately detected from images of the heart. [Brief explanation of the drawing]

[0010] [Figure 1] This figure shows an example configuration of an arrhythmia type determination system according to one embodiment of the present invention. [Figure 2] This figure shows an example of detecting the left atrium, left ventricle, right atrium, and right ventricle regions from a cardiac ultrasound image. [Figure 3] This figure shows an example of detecting the left atrium, left ventricle, right atrium, and right ventricle regions from ultrasound images of a subject who does not suffer from arrhythmia. [Figure 4] This figure shows an example of a target signal waveform that includes the period during which ventricular premature contractions (PVCs) occur. [Figure 5] This figure shows an example of a target signal waveform that includes the period during which supraventricular premature contractions (PACs) occur. [Figure 6] This is a block diagram showing an example of the main components of an early contraction detection device according to Embodiment 1 of the present invention. [Figure 7] This diagram illustrates the process of obtaining a target single-beat waveform from a target signal waveform as a target partial waveform, and comparing the shape of the target partial waveform contained within the target signal waveform with the shape of the target signal waveform. [Figure 8] This diagram illustrates the process of obtaining a target single-beat waveform from a target signal waveform as a target partial waveform, and comparing the shape of the target partial waveform contained within the target signal waveform with the shape of the target signal waveform. [Figure 9] This figure shows an example of a similarity graph created based on comparison information that includes the results of comparing the shape of the target partial waveform with the shape of the target signal waveform. [Figure 10] This figure shows an example of a similarity map created for a pair of two target subwaveforms selected from multiple target subwaveforms contained within a target signal waveform. [Figure 11]It is a diagram for describing an example of a method of comparing shapes of two target partial waveforms selected from a target signal waveform. [Figure 12] It is a flowchart showing an example of a processing flow performed by an extrasystole detection device. [Figure 13] It is a diagram for describing another example of a target signal waveform acquired from a target signal waveform. [Figure 14] It is a diagram for describing still another example of a target signal waveform acquired from a target signal waveform. DESCRIPTION OF EMBODIMENTS

[0011] Embodiment 1 Hereinafter, one embodiment of the present invention will be described in detail.

[0012] (Overview of Extrasystole Detection Device 4) The extrasystole detection device 4 according to one embodiment of the present invention detects, from a target signal waveform generated by analyzing an image of the heart of a target subject, a period during which extrasystole occurs in the heart of the target subject (hereinafter referred to as a period of interest). Here, the target signal waveform is a waveform indicating a time-series change in area of at least one region among the left atrium, left ventricle, right atrium, and right ventricle, which is generated by analyzing an image of the heart of the target subject.

[0013] Hereinafter, the term "atrial waveform" intends a signal waveform indicating a time-series change in area of an atrial region, without particularly limiting whether it is the left atrium or the right atrium. Also hereinafter, the term "ventricular waveform" intends a signal waveform indicating a time-series change in area of a ventricular region, without particularly limiting whether it is the left ventricle or the right ventricle.

[0014] Extrasystole is an arrhythmia in which the heart contracts earlier outside its original cycle due to an abnormal electrical stimulus generated in the heart, referring to a condition where occasional early beats intersperse among normal and regular pulses. The main types of extrasystole include premature ventricular contraction (PVC) and premature atrial contraction (PAC). PVC is an extrasystole in which an abnormal electrical stimulus generated in the ventricle activates the ventricle before a normal pulsation occurs, thereby causing a contraction that deviates from the normal contraction rhythm. On the other hand, PAC is an extrasystole in which an abnormal electrical stimulus generated at an abnormal site activates the atrium before a normal pulsation occurs, thereby causing a contraction that deviates from the normal contraction rhythm.

[0015] The inventors have found that a target period can be detected with high accuracy based on a signal waveform indicating time-series changes in the area of at least any one of the left atrium, left ventricle, right atrium, and right ventricle regions included in images obtained by imaging the heart. The extrasystole detection device 4 can detect a target period with high accuracy from images obtained by imaging the heart of a target subject.

[0016] The target subject may be an animal having a heart and a fetus thereof. For example, the target subject may be a human fetus. In this case, the area of each atrium and each ventricle may be calculated based on ultrasonic images including a plurality of frame images obtained by imaging the fetus. The fetus's heart is shown in the ultrasonic images obtained by imaging the fetus, and the area of each region of the fetus's heart can be calculated from the ultrasonic images. The process of calculating the area of each region of the heart from ultrasonic images will be described later with specific examples.

[0017] (Configuration of Arrhythmia Type Determination System 100) First, the configuration of an arrhythmia type determination system 100 including the extrasystole detection device 4 according to an embodiment of the present invention will be described with reference to FIG. 1. FIG. 1 is a diagram showing a configuration example of the arrhythmia type determination system 100. The arrhythmia type determination system 100 may include an imaging device 2, an image analysis device 3, the extrasystole detection device 4, an arrhythmia type determination device 1, and a display device 6.

[0018] [Imaging Device 2] The imaging device 2 may be an ultrasound imaging device capable of non-invasively imaging the inside of the subject's body. That is, the input image captured by the imaging device 2 may be an ultrasound image such as an echo image for tomographic imaging. The imaging device 2 may be connected to the image analysis device 3 in a communicative manner, as shown in the figure, in which case the image analysis device 3 may directly acquire the input image from the imaging device 2. The input image may be stored in an image management device (not shown) in association with subject information for each subject (e.g., electronic medical record information), in which case the arrhythmia type determination device 1 only needs to acquire the input image from the image management device.

[0019] [Image analysis device 3] Image analysis device 3 detects the left atrium, left ventricle, right atrium, and right ventricle regions of the subject's heart as captured in the input image. Image analysis device 3 also calculates the area of ​​each detected region and generates a target signal waveform that shows the time-series change in the area of ​​each region.

[0020] The image analysis device 3 may be installed in the facility where the imaging device 2 is installed (for example, a medical facility), or it may be installed in a remote location. If installed in a remote location, the image analysis device 3 can acquire input images via communication over a communication network such as the Internet.

[0021] Here, the process of detecting the regions corresponding to each atrium and ventricle of the heart from the input image will be explained based on Figure 2. Figure 2 shows an example of detecting the regions of the left atrium, left ventricle, right atrium, and right ventricle from an ultrasound image of the heart. Image A1 in Figure 2 is an ultrasound image of the heart, and image A2 shows the detection results for each region R1 to R4 superimposed on the ultrasound image.

[0022] In image A1, the heart is visible in the area slightly below the center, and its external shape and the division of its interior into multiple compartments can be seen. Thus, in ultrasound images, each of the two ventricles and each of the two atria of the heart can be seen as a closed region. Therefore, by analyzing ultrasound images of the heart, it is possible to detect the left atrium, left ventricle, right atrium, and the respective regions R1-R4 of the right ventricle.

[0023] For example, machine learning can be performed using training data in which regions corresponding to the left atrium, left ventricle, right atrium, and right ventricle of the heart, as seen in cardiac ultrasound images, are labeled as ground truth data. These labels can also be called annotations. Through such machine learning, it is possible to construct a learning model capable of detecting the regions R1-R4 of the left atrium, left ventricle, right atrium, and right ventricle from cardiac ultrasound images. For example, by constructing a learning model using a convolutional neural network and using that model, highly accurate region detection becomes possible.

[0024] Image A2 in Figure 2 shows the results of detection using this learning model. The detected regions R1 to R4 are the right ventricle, left ventricle, left atrium, and right atrium, respectively. By detecting each region R1 to R4, it becomes possible to calculate the area of ​​each region R1 to R4.

[0025] For example, the area of ​​each detected region R1 to R4 can be represented by the number of pixels contained in each region R1 to R4. By calculating the area of ​​each region R1 to R4 of the left atrium, left ventricle, right atrium, and right ventricle of the subject's heart in each frame of the ultrasound image, a target signal waveform showing the time-series change of the area of ​​each region R1 to R4 can be obtained. The target signal waveform includes multiple target partial waveforms corresponding to one or a predetermined number of heartbeats of the subject's heart. Here, the target partial waveform only needs to be a waveform corresponding to a length of time or longer than the time corresponding to one heartbeat. That is, the target partial waveform may be a waveform corresponding to one heartbeat of the subject's heart (hereinafter referred to as the target single-beat waveform), or a waveform corresponding to two heartbeats of the subject's heart. Alternatively, the target partial waveform may be a waveform corresponding to a predetermined time (for example, 1 second) in the target signal waveform.

[0026] Figure 3 shows an example of detecting the left atrium, left ventricle, right atrium, and right ventricle regions from ultrasound images of a subject without arrhythmias. In a heart without arrhythmias, as shown in Figure 3, the area of ​​each atrium and ventricle increases and decreases in a nearly constant cycle. Furthermore, the heart rhythm is reproduced in which the ventricular area decreases when the area of ​​the atrial region increases, and the ventricular area decreases when the area of ​​the atrial region decreases.

[0027] [Premature contraction detection device 4] Figure 4 shows an example of a target signal waveform including the period of interest in which ventricular premature contractions (PVCs) occur, with PVCs occurring in periods P1 and P2. On the other hand, Figure 5 shows an example of a target signal waveform including the period of interest in which supraventricular premature contractions (PACs) occur, with PACs occurring in periods P3 to P6. Here, a PVC is a premature contraction in which the ventricle is activated before a normal heartbeat occurs due to abnormal electrical stimulation occurring in the ventricle, resulting in a contraction that deviates from the normal contractile rhythm. A PAC, on the other hand, is a premature contraction in which the atrium is activated before a normal heartbeat occurs due to electrical stimulation occurring in an abnormal location, resulting in a contraction that deviates from the normal contractile rhythm.

[0028] The target signal waveforms during periods P1 and P2 when premature ventricular contractions (PVCs) occur in the target subjects' hearts, and the target signal waveforms during periods P3 to P6 when premature ventricular contractions (PACs) occur in the target subjects' hearts, exhibit characteristics that differ from normal (i.e., premature ventricular contraction) target signal waveforms.

[0029] For example, in the target signal waveform shown in Figure 3, the shapes of the target partial waveforms corresponding to one or more heartbeats in each region of the left atrium, right atrium, left ventricle, and right ventricle are very similar to each other. In contrast, the duration of one heartbeat in each of periods P1 and P2 of the target signal waveform shown in Figure 4 is different from the duration of one heartbeat in periods other than P1 and P2. Therefore, the similarity between the shapes of the target partial waveforms in each of periods P1 and P2 and the shapes of the target partial waveforms in periods other than P1 and P2 is low. The target partial waveforms in each of periods P1 and P2 are the partial waveforms of interest. For example, the shapes of the two target single-beat waveforms in each of periods P1 and P2 of the target signal waveform shown in Figure 4 are both narrower in width than the shapes of the target single-beat waveforms in periods other than P1 and P2.

[0030] Furthermore, the duration of one heartbeat in each of the periods P3 to P6 of the target signal waveform shown in Figure 5 is different from the duration of one heartbeat in periods other than P3 to P6. Therefore, the similarity between the shape of the target partial waveform in each of the periods P3 to P6 and the shape of the target partial waveform in periods other than P3 to P6 is low. The target partial waveform in each of the periods P2 to P6 is a partial waveform of interest. For example, the shape of one target single beat waveform in each of the periods P3 to P6 of the target signal waveform shown in Figure 5 is wider than the shape of the target single beat waveform in periods other than P3 to P6.

[0031] The inventors discovered that by focusing on the shape of the target partial waveform included in the target signal waveform, they can accurately detect the target partial waveform and period of interest when premature contractions occur in the heart of the subject. The premature contraction detection device 4 can detect the period of interest in which premature contractions such as PVCs and PACs occur in the target signal waveform based on comparison information including the results of comparing the shapes of each of the multiple target partial waveforms included in the target signal waveform.

[0032] The premature contraction detection device 4 detects a target partial waveform from among multiple target partial waveforms included in the target signal waveform acquired from the image analysis device 3, which is the waveform when premature contractions such as PVCs and PACs occur in the heart of the subject, and detects a period of interest corresponding to the target partial waveform. The period of interest only needs to include the period corresponding to the target waveform detected in the target signal waveform. Here, the length of the period of interest may be the period corresponding to the target waveform itself, or it may be the period corresponding to a predetermined number (e.g., 3) consecutive target partial waveforms that include the target waveform. Alternatively, the period of interest may be a period of a predetermined time length that includes the period corresponding to the target waveform in the target signal waveform. In this case, the period of interest may be 3 seconds, 5 seconds, or 10 seconds.

[0033] The premature contraction detection device 4 may be configured to detect the period of interest from the entire target signal waveform acquired from the image analysis device 3. The premature contraction detection device 4 may first divide the target signal waveform acquired from the image analysis device 3 into segments of a certain time interval, and then detect the period of interest from each of the segmented target signal waveforms. Alternatively, the premature contraction detection device 4 may first divide the entire target signal waveform acquired from the image analysis device 3 into segmented waveforms containing a predetermined number of partial waveforms, and then detect the period of interest from each of the segmented waveforms. In other words, the "target signal waveform" that the premature contraction detection device 4 analyzes may be the entire target signal waveform acquired from the image analysis device, or it may be one or more segmented target signal waveforms obtained by dividing the target signal waveform.

[0034] If a period of interest is detected in the target signal waveform acquired from the image analysis device 3, the premature contraction detection device 4 may output period of interest information, indicating the position of the period of interest in the target signal waveform, along with the target signal waveform, to the arrhythmia type determination device 1, which will be described later. In this case, the arrhythmia type determination device 1 determines the type of premature contraction the subject is suffering from based on the waveform in the period of interest identified by the period of interest information in the target signal waveform acquired from the premature contraction detection device 4.

[0035] Alternatively, if the premature contraction detection device 4 detects a particular waveform in the target signal waveform acquired from the image analysis device 3, it may extract the waveform corresponding to the period of interest that includes the particular waveform of interest (i.e., the target signal waveform during the period of interest). The premature contraction detection device 4 may output the extracted waveform to the arrhythmia type determination device 1, which will be described later. The extracted waveform includes one or more periods that contain the particular waveform of interest that was the basis for detecting that a premature contraction was occurring. In this case, the arrhythmia type determination device 1 determines the type of premature contraction the subject is suffering from based on the waveform extracted by the premature contraction detection device 4. The process by which the premature contraction detection device 4 detects a particular waveform of interest will be explained later with specific examples.

[0036] The premature contraction detection device 4 may be connected to the image analysis device 3 in a communication manner, as shown in the figure. In this case, the premature contraction detection device 4 may directly acquire the target signal waveform from the image analysis device 3. The target signal waveform generated by the image analysis device 3 may be stored, for example, on a portable recording medium. In this case, the premature contraction detection device 4 can read the target signal waveform from the recording medium. Alternatively, the target signal waveform generated by the image analysis device 3 may be stored in an arbitrary storage device (not shown) in association with the target subject information for each target subject. In this case, the premature contraction detection device 4 can acquire the target signal waveform from the storage device.

[0037] The premature contraction detection device 4 may be installed in the same facility as the image analysis device 3 (for example, a medical facility), or it may be installed in a remote location. If installed in a remote location, the premature contraction detection device 4 can acquire the target signal waveform generated by the image analysis device 3 via communication over a communication network such as the Internet.

[0038] The configuration may include having another computer perform some of the processing performed by the premature contraction detection device 4. In other words, the processing performed by the premature contraction detection device 4 may be performed by one or more information processing devices. Alternatively, the premature contraction detection device 4 may also have the functions of the image analysis device 3. For example, if the premature contraction detection device 4 also has the functions of the image analysis device 3, the image analysis device 3 is omitted from the components of the arrhythmia type determination system 100.

[0039] [Arrhythmia Type Determination Device 1] The arrhythmia type determination device 1 determines the type of premature contraction occurring in the subject's heart based on either (i) or (ii) below, obtained from the premature contraction detection device 4. (i) Information on the period of interest and the target signal waveform. (ii) The target signal waveform extracted from the target signal waveform, corresponding to the period of interest. The target signal waveform acquired by the arrhythmia type determination device 1 includes a pair of atrial and ventricular waveforms, for example, at least one of the following pairs (1) and (2). (1) A pair of left atrial waveforms and left ventricular waveforms. (2) A pair of right atrial waveforms and right ventricular waveforms.

[0040] The arrhythmia type determination device 1 determines the type of premature contraction that occurred in the subject's heart by utilizing the fact that there is a clear difference in the target signal waveform between cases in which PVCs occur in the subject's heart and cases in which PACs occur in the subject's heart.

[0041] For example, in the target signal waveform shown in Figure 3, during periods when the area shown in the atrial waveform is increasing, the area shown in the ventricular waveform is always decreasing, and during periods when the area shown in the atrial waveform is decreasing, the area shown in the atrial waveform is always increasing. In contrast, in the target signal waveform shown in Figure 4 (especially during periods P1 and P2), there are periods when both the area shown in the atrial waveform and the area shown in the ventricular waveform increase (hereinafter referred to as the increasing period), and periods when both the area shown in the atrial waveform and the area shown in the ventricular waveform decrease (hereinafter referred to as the decreasing period). On the other hand, in the target signal waveform shown in Figure 5, there are no increasing or decreasing periods.

[0042] Furthermore, in the target signal waveform shown in Figure 3, the similarity between the shape of the atrial waveform corresponding to one heartbeat in the atrial waveform and the shape of the ventricular waveform corresponding to the same heartbeat in the ventricular waveform is always high, and the similarity never falls below a predetermined standard value. In contrast, in the target signal waveform shown in Figure 4 (especially in periods P1 and P2), the similarity between the shape of the atrial waveform and the shape of the ventricular waveform often falls below a predetermined standard value. On the other hand, in the target signal waveform shown in Figure 5, even in periods P3 to P6, the similarity between the shape of the atrial waveform and the shape of the ventricular waveform rarely falls below a predetermined standard value.

[0043] Furthermore, in the target signal waveform shown in Figure 3, when the atria contract, the ventricles always contract afterward, and when the ventricles contract, the atria always contract afterward. In contrast, in the target signal waveform shown in Figure 4 (especially during periods P1 and P2), there is a loss of ventricular contraction corresponding to atrial contraction, and a loss of atrial contraction corresponding to ventricular contraction. On the other hand, in the target signal waveform shown in Figure 5, even during periods P3 to P6, there is no loss of ventricular contraction corresponding to atrial contraction, and no loss of atrial contraction corresponding to ventricular contraction.

[0044] Therefore, the arrhythmia type determination device 1 determines the type of premature contraction the subject is experiencing based on the target signal waveform during the period of focus. Compared to determining the type of premature contraction using an electrocardiogram, the arrhythmia type determination device 1 can output a more accurate determination result.

[0045] The arrhythmia type determination device 1 may be connected to the premature contraction detection device 4 in a communicative manner, as shown in the figure. In this case, the arrhythmia type determination device 1 may directly acquire the target signal waveform from the premature contraction detection device 4. Alternatively, the arrhythmia type determination device 1 may be connected to the image analysis device 3 and the premature contraction detection device 4 in a communicative manner. In this case, the arrhythmia type determination device 1 may acquire the target signal waveform from the image analysis device 3 and acquire the period of interest information from the premature contraction detection device 4.

[0046] The target signal waveform may be stored, for example, on a portable recording medium. In this case, the arrhythmia type determination device 1 simply reads the target signal waveform from the recording medium. Alternatively, the target signal waveform may be stored in a memory device (not shown) in association with the subject information for each subject. In this case, the arrhythmia type determination device 1 simply acquires the target signal waveform from the memory device.

[0047] The arrhythmia type determination device 1 may be installed in a facility where the premature contraction detection device 4 is installed (for example, a medical facility), or it may be installed in a remote location. If installed in a remote location, the arrhythmia type determination device 1 can acquire the target signal waveform from the premature contraction detection device 4 via communication over a communication network such as the Internet.

[0048] The arrhythmia type determination device 1 may be configured to have a part of its processing performed by another computer. In other words, the processing performed by the arrhythmia type determination device 1 may be performed by one or more information processing devices.

[0049] The premature contraction detection device 4 may also have the function of the arrhythmia type determination device 1. For example, if the premature contraction detection device 4 also has the function of the arrhythmia type determination device 1, the arrhythmia type determination device 1 is omitted from the components of the arrhythmia type determination system 100. Alternatively, the premature contraction detection device 4 may also have the function of the image analysis device 3 and the function of the arrhythmia type determination device 1. For example, if the premature contraction detection device 4 has the function of the image analysis device 3 and the function of the arrhythmia type determination device 1, the image analysis device 3 and the arrhythmia type determination device 1 are omitted from the components of the arrhythmia type determination system 100.

[0050] [Display device 6] The display device 6 is a device capable of displaying various information output from the arrhythmia type determination device 1. In addition to the various information output from the arrhythmia type determination device 1, the display device 6 may also display input images output from the imaging device 2, target signal waveforms output from the premature contraction detection device 4, etc.

[0051] (Configuration of premature contraction detection device 4) Next, the configuration of the premature contraction detection device 4 according to one embodiment of the present invention will be described with reference to Figure 6. Figure 6 is a block diagram showing an example of the main components of the premature contraction detection device 4. In the following description, the premature contraction detection device 4, which has a function to detect a portion waveform of interest from a target signal waveform acquired from the image analysis device 3, will be used as an example.

[0052] As shown in the figure, the premature contraction detection device 4 comprises a processor 40, memory 41, and storage device 42. The premature contraction detection device 4 may be a personal computer, server, or workstation. The processor 40 functions as each of the units from the acquisition unit 401 to the output control unit 405, which will be described later, by loading the premature contraction detection program 421 stored in the storage device 42 into the memory 41 and executing it.

[0053] The processor 40 can be implemented by logic circuits (hardware) formed on an integrated circuit (IC chip), or by software. When implemented by software, the processor 40 may be composed of, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or a combination of these. In this case, the software is stored in the storage device 42. The processor 40 then reads the software into memory 41 and executes it.

[0054] Memory 41 and storage device 42 are both storage devices that store various data used by the premature contraction detection device 4. Memory 41 is a storage device that can write and read data at a higher speed than storage device 42. Storage device 42 has a larger data storage capacity than memory 41. For memory 41, a high-speed access memory such as SDRAM (Synchronous Dynamic Random-Access Memory) can be applied. For storage device 42, for example, an HDD (Hard Disk Drive), SSD (Solid-State Drive), SD (Secure Digital) card, or eMMC (embedded Multi-Media Controller) can be applied.

[0055] Furthermore, the premature contraction detection device 4 is equipped with an input IF unit 43 and an output IF unit 44 as interfaces (IFs) with external devices. The input IF unit 43 is an interface for receiving input signals from input devices such as keyboards and mice, and for acquiring various information and data from external devices. For example, the input IF unit 43 can be connected to the premature contraction detection device 4, and the input IF unit 43 can be used as an interface for acquiring target signal waveforms, etc., from the premature contraction detection device 4. The output IF unit 44 is an interface for outputting the determination result of the type of premature contraction, etc., to an external device. For example, the output IF unit 44 can be connected to a display device, and the determination result of the type of premature contraction, etc., can be displayed on the display device.

[0056] The processor 40 functions as an acquisition unit 401, a partial waveform acquisition unit 402, a similarity calculation unit 403, a premature contraction detection unit 404, and an output control unit 405, by executing the premature contraction detection program 421 to acquire the target signal waveform from the image analysis device 3. In the following explanation, an example will be given in which the target single beat waveform in the target signal waveform is used as the target partial waveform.

[0057] The partial waveform acquisition unit 402 acquires each of several target partial waveforms from the target signal waveform. The similarity calculation unit 403, which will be described later, may also be capable of performing the processing that the partial waveform acquisition unit 402 performs, in which case the partial waveform acquisition unit 402 is omitted in the processor 40.

[0058] The similarity calculation unit 403 calculates a first similarity score based on comparison information, which includes the results of comparing the shapes of multiple target partial waveforms contained in the target signal waveform. Specific examples of this comparison information will be explained later.

[0059] The similarity calculation unit 403 performs the following first selection process and first calculation process. (First selection process) Select one target waveform from multiple target waveforms as the comparison waveform. (First calculation process) Each time the comparison portion waveform is moved along the time axis of the target signal waveform, the shape of the comparison portion waveform is compared with the shape of the target signal waveform at the time corresponding to the position of the comparison portion waveform to calculate the second similarity. Then, the first similarity is calculated using multiple maximum values ​​of the second similarity.

[0060] When moving the comparison portion waveform along the time axis direction of the target signal waveform, the similarity calculation unit 403 may move it at time intervals corresponding to the frame intervals in the multiple frames included in the input image used to generate the target signal waveform. Alternatively, the similarity calculation unit 403 may move the comparison portion waveform along the time axis direction of the target signal waveform at predetermined time intervals (e.g., 0.1 seconds).

[0061] As a second similarity measure, for example, cosine similarity can be applied. However, it is not limited to this, and other known similarity calculation methods may also be applied.

[0062] In the first selection process, the similarity calculation unit 403 first detects the point in the target signal waveform when the region of the target subject's heart transitions from systole to diastole, or when the region of the target subject's heart transitions from diastole to systole. This allows the similarity calculation unit 403 to identify individual target signal waveforms included in the target signal waveform and select one of them as the comparison sub-waveform.

[0063] Here, the similarity calculation unit 403 may perform a process to remove low-frequency components from each of the target signal waveform, the selected comparison signal waveform, and the multiple target sub-waveforms. Known filtering processes such as moving average filters, Fast Fourier Transform (FFT) filters, and Gaussian filters can be applied to remove low-frequency components.

[0064] Figures 7 and 8 illustrate the process of acquiring a target single-beat waveform as a target partial waveform from the target signal waveform, and comparing the shape of the target partial waveform contained in the target signal waveform with the shape of the target signal waveform. In Figures 7 and 8, the points indicated by black circles are the points in the target signal waveform where the region of the subject's heart transitions from systole to diastole. The target signal waveform shown in Figures 7 and 8 includes target partial waveforms Q1 to Q16.

[0065] Figure 7 illustrates a case in which the similarity calculation unit 403 selects the target partial waveform Q3 as the comparison partial waveform in the first selection process. In the first calculation process, the similarity calculation unit 403 moves the comparison partial waveform along the time axis from a position where the time point Pa of the comparison partial waveform coincides with the time point PA of the target signal waveform, to a position where the time point Pb of the comparison partial waveform coincides with the time point PZ of the target signal waveform. At this time, each time the comparison partial waveform is moved along the time axis, the similarity calculation unit 403 compares the shape of the comparison partial waveform with the shape of the target signal waveform at the time corresponding to the position of the comparison partial waveform and calculates the second similarity.

[0066] The second similarity score is generally maximized when the peak position of the comparison waveform coincides with the peak positions of each of the target waveforms Q1 to Q16 of the target signal waveform. The second similarity score calculated when comparing the comparison waveform with the target waveform selected as the comparison waveform is maximized at a value of "1". The second similarity score is minimized when the peak position of the comparison waveform coincides with the boundary positions of each of the target waveforms Q1 to Q16 of the target signal waveform.

[0067] In the example shown in Figure 7, the shape of the target partial waveform Q3, which was selected as the comparison partial waveform, has a high degree of similarity to the shapes of each of the target partial waveforms Q1-Q2 and Q4-Q16. In contrast, the shape of the target partial waveform Q12 is wider than the shapes of each of the other target partial waveforms Q1-Q11 and Q13-Q16, so the shape of the target partial waveform Q3, which was selected as the comparison partial waveform, has a low degree of similarity to the shape of the target partial waveform Q12.

[0068] Figure 8 shows how the similarity calculation unit 403 selects the target partial waveform Q12 as the comparison partial waveform in the first selection process. In the first calculation process, the similarity calculation unit 403 moves the comparison partial waveform along the time axis from a position where the time point Pc of the comparison partial waveform coincides with the time point PA of the target signal waveform to a position where the time point Pd of the comparison partial waveform coincides with the time point PZ of the target signal waveform. At this time, each time the comparison partial waveform is moved along the time axis, the similarity calculation unit 403 compares the shape of the comparison partial waveform with the shape of the target signal waveform at the time corresponding to the position of the comparison partial waveform and calculates the second similarity.

[0069] In the example shown in Figure 8, the shape of the target partial waveform Q12, which was selected as the comparison partial waveform, is wider than the shapes of the other target partial waveforms Q1-Q11 and Q13-Q16, and the degree of similarity to the shapes of the other target partial waveforms Q1-Q11 and Q13-Q16 is low in all cases.

[0070] The similarity calculation unit 403 then calculates the first similarity using multiple local maximums of the second similarity. Here, the first similarity may be the median or mean of the multiple local maximums of the calculated second similarity. The first similarity calculated when the target partial waveform Q12 is selected as the comparison partial waveform is lower than the first similarity calculated when the target partial waveform Q3 is selected as the comparison partial waveform.

[0071] The similarity calculation unit 403 may repeatedly perform the first selection process until it selects each of the multiple target partial waveforms as a comparison partial waveform, and may perform the first calculation process each time a comparison partial waveform is selected. This allows it to check whether each of the target partial waveforms included in the target signal waveform is a partial waveform of interest. This allows the premature contraction detection unit 404 to search in more detail whether or not the target signal waveform contains a partial waveform of interest.

[0072] The similarity calculation unit 403 may create a similarity graph plotting the relationship between the calculated second similarity and the position of the comparison waveform at the time each second similarity was calculated. In this case, the similarity calculation unit 403 may calculate the first similarity using the coordinate values ​​of multiple maximum points where the second similarity is maximized in the created similarity graph.

[0073] Figure 9 shows an example of a similarity graph created based on comparison information including the results of comparing the shape of the target partial waveform with the shape of the target signal waveform. Figure 9 shows 16 similarity graphs plotted by the similarity calculation unit 403 when each of the target partial waveforms Q1 to Q16 included in the target signal waveform shown in Figures 7 and 8 is selected as a comparison partial waveform. Among the similarity graphs shown in Figure 9, the median or mean of the maximum value of the second similarity in the similarity graph when target partial waveform Q12 is selected as a comparison partial waveform is lower than that of the other similarity graphs. Therefore, the first similarity calculated using the coordinate values ​​of multiple maximum points where the second similarity in the similarity graph for target partial waveform Q12 is maximized is low. If the calculated first similarity is below a predetermined standard, the premature contraction detection unit 404 may select target partial waveform Q12 as the focus partial waveform and detect the period corresponding to target partial waveform Q12 in the target signal waveform as the focus period.

[0074] Returning to Figure 6, the premature contraction detection unit 404 detects the period corresponding to the premature contraction waveform in the target signal waveform as the period of interest if there is a premature contraction waveform of interest whose first similarity is below a predetermined standard. For example, if the first similarity calculated for the target premature contraction waveform Q12 selected as the comparison premature contraction waveform is below a predetermined standard, the premature contraction detection unit 404 designates the target premature contraction waveform Q12 as the premature contraction waveform and detects the period corresponding to the target premature contraction waveform Q12 in the target signal waveform as the period of interest.

[0075] If the partial waveform acquisition unit 402 is configured to acquire each of a plurality of target partial waveforms from the target signal waveform, the similarity calculation unit 403 may select a pair consisting of two target partial waveforms from the acquired plurality of target partial waveforms (second selection process). The similarity calculation unit 403 may then compare the shapes of the two target partial waveforms included in the selected pair and calculate the first similarity (second calculation process).

[0076] In this case, the similarity calculation unit 403 may repeatedly perform the second selection process until it has selected all pairs consisting of two target partial waveforms from a plurality of target partial waveforms, and may perform the second calculation process each time a pair is selected. This allows the premature contraction detection unit 404 to check whether all target partial waveforms included in the target signal waveform are of interest.

[0077] The similarity calculation unit 403 may calculate a first similarity for each of the sets selected from a plurality of target partial waveforms. In that case, the similarity calculation unit 403 may create a similarity map by mapping the calculated first similarity to a two-dimensional coordinate system defined by a first axis and a second axis that intersect each other, with the following elements placed on each of the first and second axes. • First axis: One of the target partial waveforms included in each of the pairs selected from multiple target partial waveforms. • Second axis: The other target waveform included in each of the sets selected from multiple target waveforms.

[0078] Figure 10 shows an example of a similarity map created for pairs of two target subwaveforms selected from multiple target subwaveforms contained in a target signal waveform. The first axis (e.g., vertical axis) of the similarity map shown in Figure 10 is represented by one of the target subwaveforms Q1 to 16 included in each pair selected from the target signal waveforms shown in Figures 7 and 8. The second axis (e.g., horizontal axis) is represented by the other target subwaveform Q1 to 16 included in each of the same pair.

[0079] In the similarity map, the first similarity calculated for each set selected from multiple target partial waveforms is displayed by the intensity of the color. For example, in Figure 10, a gray band extending upward from the cell where (vertical axis, horizontal axis) is (Q12, Q12) and another gray band extending to the right from the same cell can be seen. These gray bands indicate that the first similarity calculated by comparing the shape of target partial waveform Q12 with the shapes of the other target partial waveforms Q1-Q11 and Q13-Q16 is below a predetermined standard. In this way, the similarity map can intuitively and visually show whether or not there are sets among the sets selected from multiple target partial waveforms included in the target signal waveform whose first similarity is below a predetermined standard.

[0080] The premature contraction detection unit 404 may, if there are sets whose first similarity is below a predetermined standard, identify the target partial waveform included in the set as the partial waveform of interest, and detect the period corresponding to the partial waveform of interest in the target signal waveform as the period of interest. For example, the premature contraction detection unit 404 may detect the partial waveform of interest in the target signal waveform based on whether or not an upward-extending gray band and a rightward-extending gray band are detected in the created similarity map, and detect the period corresponding to the partial waveform of interest as the period of interest.

[0081] Here, the process by which the similarity calculation unit 403 compares the shapes of two target partial waveforms will be explained using Figure 11. Figure 11 is a diagram illustrating an example of a method for comparing the shapes of two target partial waveforms selected from the target signal waveform. In Figure 11, the similarity calculation unit 403 selects target partial waveform A as the waveform between time points Pe and Pf, and selects target partial waveform B as the waveform between time points Pg and Ph. Alternatively, the similarity calculation unit 403 may normalize the amplitude of each selected waveform and calculate the similarity by comparing the shapes of the normalized waveforms.

[0082] The similarity calculation unit 403 moves the target partial waveform B along the time axis from a position where time point Pg of the target partial waveform B coincides with time point Pe of the target partial waveform A, to a position where time point Ph of the target partial waveform B coincides with time point Pf of the target partial waveform A. At this time, each time the target partial waveform B is moved along the time axis, the similarity calculation unit 403 calculates the similarity by comparing the shape of the target partial waveform B with the shape of the target partial waveform A at the time corresponding to the position of the target partial waveform B.

[0083] Returning to Figure 6, the output control unit 405 outputs the detection result from the premature contraction detection unit 404 to the arrhythmia type determination device 1. The output control unit 405 may also output the detection result from the premature contraction detection unit 404 to various output devices. For example, if a display device 6 is connected via the output IF unit 44, the output control unit 405 may display the determination result on the display device 6. The manner in which the detection result is output is arbitrary, and the output control unit 405 may output the detection result by display output, audio output, print output, or a combination thereof.

[0084] (Processing performed by arrhythmia type determination device 1) Next, the processing performed by the premature contraction detection device 4 (premature contraction detection method) will be explained using Figure 12. Figure 12 is a flowchart showing an example of the processing flow performed by the premature contraction detection device 4.

[0085] First, the acquisition unit 401 acquires the target signal waveform (Step S1: Acquisition step). Next, the similarity calculation unit 403 calculates a first similarity based on comparison information including the results of comparing the shapes of each of the multiple target partial waveforms included in the target signal waveform (Step S2: Similarity calculation step).

[0086] Then, if there is a portion waveform of interest with a first similarity below a predetermined standard, the premature contraction detection unit 404 detects the period corresponding to the portion waveform of interest in the target signal waveform as the period of interest during which premature contractions occur in the heart of the subject (Step S3: Premature Contraction Detection Step). The output control unit 405 outputs the detection result to the arrhythmia type determination device 1 (Step S4: Output Step). Alternatively, in Step S4, the output control unit 405 may display the detection result on a display device (for example, the display device 6 shown in Figure 1).

[0087] With the above configuration, the premature contraction detection device 4 can accurately detect the period of interest during which premature contractions occur in the subject's heart from images taken of the subject's heart.

[0088] [Embodiment 2] Other embodiments of the present invention are described below. For the sake of clarity, components having the same function as those described in the above embodiments will be denoted by the same reference numerals, and their descriptions will not be repeated.

[0089] In Embodiment 1, the premature contraction detection device 4 was described using the case where the target single-beat waveform in the target signal waveform is used as the target partial waveform, but the configuration is not limited to this. For example, the partial waveform acquisition unit 402 and the similarity calculation unit 403 may use waveforms corresponding to multiple heartbeats in the target signal waveform as the target partial waveforms.

[0090] Figure 13 illustrates another example of a target signal waveform obtained from a target signal waveform. Figure 13 illustrates a process in which waveforms for two heartbeats are obtained from the target signal waveform as target partial waveforms, and the shape of the target partial waveforms included in the target signal waveform is compared with the shape of the target signal waveform. In Figure 13, the points indicated by black circles are the points in the target signal waveform where the heart region of the subject changes from diastole to systole. Figure 13 illustrates the case in the first selection process where the similarity calculation unit 403 selects waveforms for two heartbeats that enclose the waveform indicated by "Q3" as comparison partial waveforms.

[0091] Alternatively, the partial waveform acquisition unit 402 and the similarity calculation unit 403 may use a wavelength for a predetermined time in the target signal waveform as the target partial waveform. Figure 14 is a diagram illustrating another example of a target signal waveform acquired from the target signal waveform. Figure 14 is a diagram illustrating a process in which a waveform for 1 second is acquired from the target signal waveform as a target partial waveform, and the shape of the target partial waveform included in the target signal waveform is compared with the shape of the target signal waveform. In Figure 14, the time points indicated by black circles are spaced at 1-second intervals from each other. Figure 14 illustrates a case in the first selection process in which the similarity calculation unit 403 selects a waveform corresponding to 1 second that includes the waveform indicated by "Q1" and the waveform indicated by "Q2" as the comparison partial waveform.

[0092] The premature contraction detection device 4 does not require the target partial waveform used to detect the period of interest to be the target beat waveform included in the target signal waveform. However, if the time width (length) of the target partial waveform increases, the sensitivity and accuracy of detecting premature contractions will decrease accordingly. Therefore, in order to ensure the sensitivity and accuracy of detecting the period of interest, it is desirable to use a waveform corresponding to a period of one to three heartbeats as the target partial waveform.

[0093] [Examples of implementation using software] The function of the premature contraction detection device 4 (hereinafter referred to as "the device") is a program that causes a computer to function as the device, and can be realized by a program that causes a computer to function as each control block of the device (in particular each part included in the processor 40).

[0094] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., memory) as hardware for executing the program. By executing the program using this control device and storage device, the functions described in each of the embodiments are realized.

[0095] The above program may be recorded on one or more computer-readable recording media, not temporary ones. These recording media may or may not be provided by the above device. In the latter case, the program may be supplied to the above device via any wired or wireless transmission medium.

[0096] Furthermore, some or all of the functions of each of the above control blocks can also be realized by logic circuits. For example, an integrated circuit in which logic circuits functioning as each of the above control blocks are formed is also included in the scope of the present invention. In addition, it is also possible to realize the functions of each of the above control blocks by, for example, a quantum computer.

[0097] Furthermore, each process described in the above embodiments may be performed by AI (Artificial Intelligence). In this case, the AI ​​may operate on the control device described above, or it may operate on other devices (for example, an edge computer or a cloud server).

[0098] The present invention is not limited to the embodiments described above, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0099] 〔summary〕 The premature contraction detection device according to embodiment 1 of the present invention comprises: an acquisition unit that acquires a target signal waveform that shows the time-series change in the area of ​​at least one of the regions of the left atrium, left ventricle, right atrium, and right ventricle, generated by analyzing an image of the heart of a subject, and includes a plurality of target partial waveforms corresponding to one or a predetermined number of heartbeats of the heart of the subject; a similarity calculation unit that calculates a first similarity based on comparison information including the result of comparing the shapes of each of the plurality of target partial waveforms included in the target signal waveform; and a premature contraction detection unit that, if there is a particular partial waveform whose first similarity is below a predetermined standard, detects the period corresponding to the particular partial waveform in the target signal waveform as a period of interest during which premature contractions occur in the heart of the subject.

[0100] In the premature contraction detection device according to embodiment 2 of the present invention, the similarity calculation unit may perform a first selection process to select one of the plurality of target partial waveforms as a comparison partial waveform, and a first calculation process to calculate a second similarity by comparing the shape of the comparison partial waveform with the shape of the target signal waveform at a time corresponding to the position of the comparison partial waveform each time the comparison partial waveform is moved along the time axis direction in the target signal waveform, and to calculate the first similarity using a plurality of maximum values ​​of the second similarity.

[0101] In the premature contraction detection device according to embodiment 3 of the present invention, in embodiment 2, the similarity calculation unit may repeatedly perform the first selection process until each of the plurality of target partial waveforms is selected as the comparison partial waveform, and the first calculation process may be performed each time the comparison partial waveform is selected.

[0102] In the premature contraction detection device according to embodiment 4 of the present invention, in embodiment 3, the similarity calculation unit may create a similarity graph plotting the relationship between the second similarity calculated each time the comparison partial waveform is moved along the time axis and the position of the comparison partial waveform when each of the second similarities was calculated, and calculate the first similarity using the coordinate values ​​of a plurality of maximum points where the second similarity is maximized in the similarity graph.

[0103] The premature contraction detection device according to embodiment 5 of the present invention further comprises a partial waveform acquisition unit that acquires each of the plurality of target partial waveforms from the target signal waveform, in any of embodiments 1 to 4 above, and the similarity calculation unit may perform a second selection process that selects a pair consisting of two of the target partial waveforms from the acquired plurality of target partial waveforms, and a second calculation process that calculates the first similarity by comparing the shapes of the two target partial waveforms included in the selected pair.

[0104] In the premature contraction detection device according to embodiment 6 of the present invention, in embodiment 5, the similarity calculation unit may repeatedly perform the second selection process until all pairs formed by combining two of the target partial waveforms are selected from the plurality of target partial waveforms, and the second calculation process may be performed each time a pair is selected.

[0105] In the premature contraction detection device according to embodiment 7 of the present invention, in embodiment 5 or 6 above, the similarity calculation unit may create a similarity map by mapping the first similarity calculated for each of the sets selected from the plurality of target partial waveforms to a two-dimensional coordinate system defined by a first axis and a second axis that intersect each other, wherein one of the target partial waveforms included in each of the sets is placed on the first axis and the other of the target partial waveforms included in each of the sets is placed on the second axis.

[0106] In the premature contraction detection device according to embodiment 8 of the present invention, in embodiment 7, if there is a set whose first similarity is less than or equal to the predetermined standard, the premature contraction detection unit may identify the target partial waveform included in the set as the focus partial waveform, and detect the period in the target signal waveform corresponding to the focus partial waveform as the focus period.

[0107] A premature contraction detection method according to aspect 9 of the present invention is a premature contraction detection method performed by one or more information processing devices, comprising: an acquisition step of acquiring a target signal waveform that shows the time-series change in the area of ​​at least one of the regions of the left atrium, left ventricle, right atrium, and right ventricle, generated by analyzing an image of the heart of a subject, and includes a plurality of target partial waveforms corresponding to one or more predetermined number of heartbeats of the heart of the subject; a similarity calculation step of calculating a first similarity based on comparison information including the result of comparing the shapes of each of the plurality of target partial waveforms included in the target signal waveform; and a premature contraction detection step of detecting, if there is a particular partial waveform whose first similarity is below a predetermined standard, the period corresponding to the particular partial waveform in the target signal waveform as a period of interest during which premature contractions occur in the heart of the subject.

[0108] The premature contraction detection program according to aspect 10 of the present invention is a premature contraction detection program for causing a computer to function as a premature contraction detection device according to any of aspects 1 to 8 above, and is a premature contraction detection program for causing a computer to function as the acquisition unit, the similarity calculation unit, and the premature contraction detection unit. [Explanation of Symbols]

[0109] 4. Premature contraction detection device 40 processors 401 Acquisition Department 402 Partial waveform acquisition section 403 Similarity calculation unit 404 Premature contraction detection unit 421 Premature Contraction Detection Program

Claims

1. An acquisition unit that acquires a target signal waveform that shows the time-series change in the area of ​​at least one of the regions of the left atrium, left ventricle, right atrium, and right ventricle, generated by analyzing images of the target subject's heart, and includes multiple target partial waveforms corresponding to one or more predetermined number of heartbeats of the target subject's heart, A similarity calculation unit calculates a first similarity based on comparison information including the result of comparing the shapes of each of the plurality of target partial waveforms included in the target signal waveform, If there is a portion waveform of interest whose first similarity is below a predetermined standard, the premature contraction detection unit detects the period corresponding to the portion waveform of interest in the target signal waveform as the period of interest during which premature contractions occur in the heart of the subject, A premature contraction detection device equipped with the following features.

2. The similarity calculation unit, A first selection process in which one of the multiple target partial waveforms is selected as a comparison partial waveform, A first calculation process that, each time the comparison portion waveform is moved along the time axis direction in the target signal waveform, compares the shape of the comparison portion waveform with the shape of the target signal waveform at the time corresponding to the position of the comparison portion waveform to calculate a second similarity, and calculates the first similarity using a plurality of maximum values ​​of the second similarity, Execute The premature contraction detection device according to claim 1.

3. The similarity calculation unit, The first selection process is repeatedly executed until each of the multiple target partial waveforms is selected as the comparison partial waveform. The first calculation process is executed each time the aforementioned comparison partial waveform is selected. The premature contraction detection device according to claim 2.

4. The similarity calculation unit, A similarity graph is created by plotting the relationship between the second similarity calculated each time the comparison portion waveform is moved along the time axis and the position of the comparison portion waveform at the time each of the second similarity values ​​was calculated. The first similarity is calculated using the coordinate values ​​of a plurality of local maximum points in the similarity graph where the second similarity is maximized. The premature contraction detection device according to claim 3.

5. The system further includes a partial waveform acquisition unit that acquires each of the plurality of target partial waveforms from the target signal waveform, The similarity calculation unit, A second selection process for selecting a set consisting of two target partial waveforms from the acquired plurality of target partial waveforms, A second calculation process that calculates the first similarity by comparing the shapes of the two target partial waveforms included in the selected set, Execute The premature contraction detection device according to claim 1.

6. The similarity calculation unit, The second selection process is repeatedly executed until all pairs consisting of two of the target partial waveforms are selected from the plurality of target partial waveforms. Each time the aforementioned pair is selected, the second calculation process is executed. The premature contraction detection device according to claim 5.

7. The similarity calculation unit, A similarity map is created by mapping the first similarity calculated for each of the sets selected from the plurality of target partial waveforms onto a two-dimensional coordinate system defined by a first axis and a second axis that intersect each other, wherein one of the target partial waveforms included in each set is placed on the first axis and the other of the target partial waveforms included in each set is placed on the second axis. The premature contraction detection device according to claim 5.

8. The aforementioned premature contraction detection unit is If there is a set whose first similarity is less than or equal to the predetermined standard, the target partial waveform included in the set is identified as the partial waveform of interest. The period corresponding to the portion waveform of interest in the target signal waveform is detected as the period of interest. The premature contraction detection device according to claim 7.

9. A method for detecting premature contractions, which is performed by one or more information processing devices, Acquisition step: Obtain a target signal waveform that shows the time-series change in the area of ​​at least one of the regions of the left atrium, left ventricle, right atrium, and right ventricle, generated by analyzing images of the target subject's heart, and includes multiple target partial waveforms corresponding to one or more predetermined number of heartbeats of the target subject's heart. A similarity calculation step that calculates a first similarity based on comparison information including the result of comparing the shapes of each of the plurality of target partial waveforms included in the target signal waveform, If there is a portion waveform of interest whose first similarity is below a predetermined standard, the premature contraction detection step includes detecting the period corresponding to the portion waveform of interest in the target signal waveform as the period of interest during which premature contractions occur in the heart of the subject, A method for detecting premature contractions, including the method described above.

10. An early contraction detection program for causing a computer to function as an early contraction detection device according to claim 1, wherein the computer functions as the acquisition unit, the similarity calculation unit, and the early contraction detection unit.

Images