Heart parameter measuring method and system

Abstract

The invention relates to the technical field of ultrasonic measurement, in particular to a heart parameter measuring method and system. According to the heart parameter measurement method, the position of the heart is acquired, and the probe is finely positioned according to the position of the heart, so that the probe can accurately and quickly find and acquire the ultrasonic image at the standard section of the heart; then capturing a target standard section in the ultrasonic image by adopting a pre-trained self-supervised deep learning model; and finally, carrying out key frame extraction on the 2D echocardiogram video so as to measure a heart structure in a key frame to obtain heart structure parameters, cutting out a region of interest in the Doppler echocardiogram, and carrying out segmentation processing on the region of interest to obtain functional parameters of the heart. By means of the mode, various parameters of the heart can be accurately and efficiently obtained, full automation is achieved in the whole process, and the accuracy and consistency of various parameter test results of the heart are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of ultrasonic measurement, and in particular to a cardiac parameter measurement method and system. Background Art

[0002] Echocardiography is an important diagnostic aid for evaluating the morphology and function of the heart in cardiology. However, due to the obvious individual differences in image acquisition, measurement, analysis, and judgment of current echocardiography, which are greatly affected by the experience and ability of medical workers, it is difficult to ensure the accuracy and consistency of the test, which often brings great difficulties to clinical identification. Therefore, it is necessary to design a method that can realize the automatic measurement of cardiac parameters. Summary of the invention

[0003] The present invention provides a cardiac parameter measurement method and system to solve the problem of low precision and low accuracy in cardiac parameter measurement in the prior art.

[0004] According to the first aspect, an embodiment provides a method for measuring cardiac parameters, comprising:

[0005] Acquiring the position of the heart, and finely positioning the probe according to the position of the heart, wherein the probe acquires an ultrasound image including a target standard section after fine positioning;

[0006] Capturing target standard sections in the ultrasound image by using a pre-trained self-supervised deep learning model; the target standard sections include 2D echocardiogram video and Doppler echocardiogram;

[0007] Extracting key frames from the 2D echocardiography video, and measuring cardiac structural parameters in the key frames to obtain cardiac structural parameters;

[0008] The region of interest of the Doppler echocardiogram is cropped and segmented to obtain the functional parameters of the heart.

[0009] In one achievable manner, the obtaining the position of the heart and finely positioning the probe according to the position of the heart include:

[0010] Segmenting the human body position according to the visual information of the inspection object, and locating the approximate position of the heart from the human body position by using a preset human body physical model;

[0011] Obtain auscultatory information of the approximate position of the heart;

[0012] Extracting features from the auscultation information, and locating a target area of ​​the heart according to the extracted features of the auscultation information;

[0013] Moving the probe to the target area to obtain an ultrasonic image of the target area;

[0014] The target position in the target area is determined according to the ultrasonic image of the target area, and the target position is the position of the probe when scanning the target standard section, so that the probe can obtain the ultrasonic image of the target standard section.

[0015] In one achievable manner, capturing the target standard section in the ultrasound image by using a pre-trained self-supervised deep learning model includes:

[0016] Pre-build a single classifier in a self-supervised deep learning model;

[0017] Using the collected standard sections to train the single classifier, so that the single classifier is used to identify the target standard section in the ultrasound image;

[0018] The ultrasound image is subjected to abnormality detection by the trained single classifier. When the single classifier detects an abnormality, it is identified as a non-target standard section. When the single classifier detects a non-abnormality, it is identified as a target standard section.

[0019] In one achievable manner, extracting key frames from the 2D echocardiography video includes:

[0020] Measuring cardiac structural parameters of a plurality of image frames in the 2D echocardiography video to obtain a measurement curve of changes in cardiac structural parameters over time;

[0021] The measurement curves of all cardiac structural parameters are fused by weighted summation to obtain a cardiac motion curve reflecting the heart beating;

[0022] extracting a first electrocardiogram from the 2D echocardiogram video, and identifying an R-wave apex in an electrocardiogram signal of the first electrocardiogram;

[0023] Obtaining a maximum value in the heart motion curve, and comparing a time corresponding to the maximum value with a time corresponding to the R wave apex;

[0024] If the time difference between the moment corresponding to the maximum value and the moment corresponding to the R wave apex does not exceed the preset threshold, the maximum and minimum values ​​of the cardiac motion curve are extracted, and the image frames to which the maximum and minimum values ​​of the cardiac motion curve belong are used as key frames; if the time difference between the moment corresponding to the maximum value and the moment corresponding to the R wave apex exceeds the preset threshold, the R wave apex in the first electrocardiogram and the minimum value of the cardiac motion curve are extracted, and the image frames to which the R wave apex and the minimum value of the cardiac motion curve belong are used as key frames.

[0025] In one achievable manner, the measuring of cardiac structural parameters of a plurality of image frames in the 2D echocardiography video includes:

[0026] A five-layer convolutional neural network is used as an encoder to extract image features of multiple image frames in the 2D echocardiography video;

[0027] Performing sequential regression on the coordinate offset of the image features of each image frame through a fully connected layer of a graph convolutional network to obtain measurement points of the cardiac structure parameters in the multiple image frames;

[0028] The distances between the measurement points of the cardiac structural parameters are measured to obtain the cardiac structural parameters.

[0029] In one achievable manner, the cropping of the region of interest of the Doppler echocardiogram and segmentation of the region of interest to obtain the functional parameters of the heart include:

[0030] Acquiring a blood flow spectrum image of the Doppler echocardiogram, and extracting a second electrocardiogram from the blood flow spectrum image;

[0031] cutting out a region of interest from the Doppler echocardiogram according to the cycle information of the second electrocardiogram;

[0032] Segmenting the contour of the blood flow spectrum image within the region of interest by using a segmentation model;

[0033] Acquiring vertex position and area information from the segmented blood flow spectrum image;

[0034] The functional parameters of the heart are calculated based on the apex position and area information.

[0035] According to the second aspect, an embodiment provides a cardiac parameter measurement system, comprising:

[0036] A probe positioning module is used to obtain the position of the heart and finely position the probe according to the position of the heart; after fine positioning, the probe obtains an ultrasound image containing a target standard section;

[0037] A target standard section capture module, used to capture the target standard section in the ultrasound image through a pre-trained self-supervised deep learning model; the target standard section includes a 2D echocardiogram video and a Doppler echocardiogram;

[0038] A cardiac structural parameter measurement module, used for extracting key frames from the 2D echocardiography video, measuring the cardiac structural parameters in the key frames, and obtaining cardiac structural parameters;

[0039] The cardiac function parameter measurement module is used to crop the region of interest of the Doppler echocardiogram, segment the region of interest, and obtain the cardiac function parameters.

[0040] In one achievable manner, the probe positioning module includes:

[0041] A coarse positioning submodule, used to segment the human body position according to the visual information of the inspection object, and locate the approximate position of the heart from the human body position through a preset human body physical model;

[0042] An auscultation information acquisition submodule is used to obtain auscultation information of the approximate position of the heart;

[0043] A fine positioning submodule, used for extracting features from the auscultation information and locating a target area of ​​the heart according to the extracted features of the auscultation information;

[0044] An ultrasonic image acquisition submodule, used for moving the probe to the target area to acquire an ultrasonic image of the target area;

[0045] The fine positioning submodule is used to determine the target position in the target area according to the ultrasonic image of the target area, and the target position is the position of the probe when scanning the target standard section, so that the probe can obtain the ultrasonic image of the target standard section.

[0046] In one achievable manner, in the target standard section capture module, capturing the target standard section in the ultrasound image by using a pre-trained self-supervised deep learning model includes:

[0047] Pre-build a single classifier in a self-supervised deep learning model;

[0048] Using the collected standard sections to train the single classifier, so that the single classifier is used to identify the target standard section in the ultrasound image;

[0049] The ultrasound image is subjected to abnormality detection by the trained single classifier. When the single classifier detects an abnormality, it is identified as a non-target standard section. When the single classifier detects a non-abnormality, it is identified as a target standard section.

[0050] According to a third aspect, an embodiment provides a computer-readable storage medium, on which a program is stored, and the program can be executed by a processor to implement the above method.

[0051] According to the cardiac parameter measurement method / system of the above-mentioned embodiment, by obtaining the position of the heart, the probe is precisely positioned according to the position of the heart, so that the probe can accurately and quickly find and obtain the ultrasound image at the standard section of the heart, and then the pre-trained self-supervised deep learning model is used to capture the target standard section in the ultrasound image, wherein the target standard section includes a 2D echocardiogram video and a Doppler echocardiogram; finally, the 2D echocardiogram video is subjected to key frame extraction to facilitate the measurement of the cardiac structure in the key frame to obtain the cardiac structural parameters, and in addition, the region of interest in the Doppler echocardiogram is cropped out, and the region of interest is segmented to obtain the functional parameters of the heart. By adopting the above method, various parameters of the heart can be accurately and efficiently obtained, and the whole process is fully automated, avoiding the influence of the experience and ability of medical workers, and ensuring the accuracy and consistency of the various parameter test results of the heart. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A flowchart of a cardiac parameter measurement method provided in an embodiment of the present application;

[0053] Figure 2 A flowchart of fine positioning of a probe provided in an embodiment of the present application;

[0054] Figure 3 A flowchart of the target standard section capture process provided in the embodiment of the present application;

[0055] Figure 4 A flowchart of the cardiac structural parameter measurement process provided in the embodiment of the present application;

[0056] Figure 5 A flowchart of obtaining cardiac structural parameters provided in an embodiment of the present application

[0057] Figure 6 A flowchart of the cardiac function parameter measurement process provided in the embodiment of the present application;

[0058] Figure 7 A flow chart of the cardiac structural parameter measurement process provided in the embodiment of the present application;

[0059] Figure 8 A flow chart of obtaining cardiac structural parameters provided in an embodiment of the present application;

[0060] Fig. 9 A flow chart of the cardiac function parameter measurement process provided in the embodiment of the present application;

[0061] Fig.10 A structural block diagram of a cardiac parameter measurement system provided in an embodiment of the present application;

[0062] Fig.11 This is a structural block diagram of the probe positioning module provided in an embodiment of the present application.

[0063] Figure numerals: 10, probe positioning module; 20, target standard section capture module; 30, cardiac structure parameter measurement module; 40, cardiac function parameter measurement module; 11, coarse positioning submodule; 12, auscultation information acquisition submodule; 13, fine positioning submodule; 14, ultrasound image acquisition submodule; 15, fine positioning submodule; 100, measurement curve; 200, heart motion curve; 300, key frame result; 400, 2D echocardiography video; 500, convolution layer; 600, graph convolution network; 700, fully connected layer. DETAILED DESCRIPTION

[0064] The present invention is further described in detail below by specific embodiments in conjunction with the accompanying drawings. Wherein similar elements in different embodiments adopt associated similar element numbers. In the following embodiments, many detailed descriptions are for making the present application better understood. However, those skilled in the art can easily recognize that some features can be omitted in different situations, or can be replaced by other elements, materials, methods. In some cases, some operations related to the present application are not shown or described in the specification, this is to avoid the core part of the present application being overwhelmed by too much description, and for those skilled in the art, it is not necessary to describe these related operations in detail, and they can fully understand the related operations according to the description in the specification and the general technical knowledge in the art.

[0065] In addition, the features, operations or characteristics described in the specification can be combined in any appropriate manner to form various implementations. At the same time, the steps or actions in the method description can also be interchanged or adjusted in a manner that is obvious to those skilled in the art. Therefore, the various sequences in the specification and the drawings are only for the purpose of clearly describing a certain embodiment and are not meant to be a required sequence, unless otherwise specified that a certain sequence must be followed.

[0066] The serial numbers of the components in this document, such as "first", "second", etc., are only used to distinguish the objects described and do not have any order or technical meaning. The "connection" and "coupling" mentioned in this application, unless otherwise specified, include direct and indirect connections (couplings).

[0067] The present application proposes a cardiac parameter measurement method and system, which aims to realize the standardized acquisition of echocardiogram and the full measurement of cardiac quantitative parameters. The cardiac parameter measurement method and the cardiac parameter measurement system are described in detail below.

[0068] This embodiment provides a method for measuring cardiac parameters, such as Figure 1 As shown, the following steps are included:

[0069] Step 100: Acquire the position of the heart, and finely position the probe according to the position of the heart. After fine positioning, the probe acquires an ultrasound image including a target standard section.

[0070] This step is implemented by the probe positioning module 10 of the cardiac parameter measurement system. Specifically, by installing a sensor (i.e., a probe) at the end of the robotic arm, in actual operation, since the mechanical conventional body design of a conventional robotic arm is already a stereotyped design, its accuracy mainly depends on the controller accuracy and the motion planning algorithm. Therefore, the robotic arm body of this embodiment is intended to be based on a robotic arm that can be redeveloped, and a robotic arm that meets the accuracy is selected according to the task space and load. Next, the end effector is developed and designed based on the robotic arm that can be redeveloped to mount the required positioning sensor. Then, the movement of the robotic arm is controlled by the probe positioning module 10 to reach the required positioning point and angle. At the same time, a certain pressure is applied to the sensor probe to achieve clear acquisition of ultrasonic images. In the process of applying pressure, the flexible control of the robotic arm is achieved based on the force / torque sensor.

[0071] In this step, if Figure 2 As shown in FIG. 1 , the fine positioning of the probe specifically includes the following steps:

[0072] Step 110: The rough positioning submodule 11 segments the human body position according to the visual information of the inspection object, and locates the approximate position of the heart from the human body position through a preset human body physical model. Among them, by integrating the camera at the end of the robotic arm, it can be set next to the probe to collect visual information around the part to be inspected.

[0073] Step 120: Obtain auscultation information of the approximate position of the heart through the auscultation information acquisition submodule 12. Specifically, the probe is moved to the heart position, and then the force / torque / tactile sensor at the end of the robotic arm senses that the probe contacts the skin, and then detects the auscultation information of the heart position. In this embodiment, the auscultation information is collected by using one of the acoustic sensors, which is mainly used to collect acoustic information of heart sounds.

[0074] Step 130: The fine positioning submodule 13 extracts features from the auscultation information and locates the target area of ​​the heart according to the extracted features of the auscultation information.

[0075] Step 140: The ultrasound image acquisition submodule 14 moves the probe to the target area to acquire an ultrasound image of the target area.

[0076] Step 150: the fine positioning submodule 15 determines the target position in the target area according to the ultrasonic image of the target area. The target position is the position of the probe when scanning the target standard section, so that the probe can obtain the ultrasonic image of the target standard section.

[0077] As for the positioning of the heart, since the heart is inside the body, the positioning target is invisible, which undoubtedly increases the difficulty of positioning. Conventional single sensor information is difficult to accurately locate. Therefore, this application uses multiple sensors to collect information and fuses the multi-sensor information to control the movement of the robotic arm, thereby realizing the collection of ultrasound images of the area to be measured.

[0078] Specifically, for the precise positioning of the probe, a positioning scheme that controls multi-stage multimodal fusion is adopted. First, the bed sheet is set to a single color. Based on the visual information observed by the camera, the human body position is segmented using a traditional segmentation algorithm. The position of the heart is estimated a priori through the preset human body physical model. Then, the approximate position of the heart is located from the human body position based on experience to achieve the rough positioning of the probe. When the force / torque / tactile information senses that the probe touches the skin, it enters the fine positioning stage; in the fine positioning stage, the acoustic information features of the heart sound in the auscultation information are extracted to identify the five major auscultation areas of the heart, and the target area of ​​the heart is located in the five major auscultation areas to achieve fine positioning of the heart, and then the probe is moved to the target area of ​​the heart to obtain the ultrasound image of the target area. Finally, the existing deep learning force model is used to determine the target position based on the auscultation information and the ultrasound image of the target area. The target position is the position where the probe is located when scanning the target standard section, and then the probe is moved to this position for ultrasound image acquisition to obtain the ultrasound image of the target standard section.

[0079] Step 200: Capture target standard sections in ultrasound images through a pre-trained self-supervised deep learning model; the target standard sections include 2D echocardiogram video 400 and Doppler echocardiogram.

[0080] This step is implemented by the target standard section capture module 20 of the cardiac parameter measurement system. Figure 3 The target standard section capture module 20 performs the following steps:

[0081] Step 210: Pre-build a single classifier in a self-supervised deep learning model.

[0082] Step 220: Use the collected standard sections to train a single classifier, so that the single classifier is used to identify the target standard section in the ultrasound image.

[0083] Step 230: Perform abnormality detection on the ultrasound image through the trained single classifier. When the single classifier detects an abnormality, it is identified as a non-target standard section. When it detects a non-abnormality, it is identified as a target standard section.

[0084] In this step, first, a self-supervised deep learning model is selected, a single classifier is constructed in the model, and then the collected standard sections or standard sections obtained online are used to train the single classifier, so that the single classifier can detect the target standard section in the ultrasound image. Specifically, the ultrasound image is detected for abnormalities by the single classifier. When the single classifier detects an abnormality, it is identified as a non-target standard section, and when a non-abnormality is detected, it is identified as a target standard section. The captured target standard sections include 2D echocardiogram video 400 and Doppler echocardiogram. In addition, the method for abnormality detection in this embodiment is to perform targeted distinction extraction by detecting the clarity of the section, and to construct a self-supervised auxiliary task that is particularly sensitive to texture clarity through information entropy measurement to achieve abnormality detection of ultrasound images.

[0085] Step 300: extract key frames from the 2D echocardiography video 400 through the cardiac structure parameter measurement module 30, measure the cardiac structure parameters in the key frames, and obtain the cardiac structure parameters.

[0086] Step 400: The cardiac function parameter measurement module 40 cuts out the region of interest of the Doppler echocardiogram, performs segmentation processing on the region of interest, and obtains cardiac function parameters.

[0087] The cardiac parameters obtained clinically include cardiac structural parameters and cardiac functional parameters. For example, the cardiac structural parameters commonly used in clinical practice are: Ao-a aortic valve annulus inner diameter; Ao-s aortic sinus inner diameter; LA-ap left atrial anteroposterior diameter; MV-ap mitral valve annulus anteroposterior diameter; RVOT-Prox right ventricular outflow tract proximal inner diameter; LVID-ap left ventricular anteroposterior diameter; LA-l left atrial long diameter; LA-t left atrial transverse diameter; LVID base-t left ventricular basal segment transverse diameter; LVID middle-t left ventricular middle segment transverse diameter; LVID-l left ventricular long diameter; MV-t mitral valve annulus transverse diameter; RA-l right atrial long diameter; RA-t right atrial transverse diameter; TV-t tricuspid valve annulus transverse diameter; RV base-t right ventricular basal segment transverse diameter; RV middle-t right ventricular middle segment transverse diameter; RV-l right ventricular long diameter, etc. The functional parameters of the heart commonly used in clinical practice include: maximum blood flow velocity through the heart valve orifice, average blood flow velocity, integral of flow velocity, E peak value, A peak value, E / A ratio, etc.

[0088] In the actual operation, it is planned to be carried out based on 3 2D echocardiography videos 400 and 7 Doppler echocardiography images. First, the cardiac structural parameters of the 2D echocardiography video 400 are measured, the key frames of the video are extracted, and the values ​​of the parameters are determined based on the key frames, that is, the values ​​of the cardiac structural parameters are obtained; then, the region of interest of the Doppler echocardiography video is segmented, and the properties such as the vertices or area of ​​the segmented region are used to achieve the measurement of cardiac function parameters.

[0089] In this embodiment, by obtaining the position of the heart, the probe is precisely positioned according to the position of the heart, so that the probe can accurately and quickly find and obtain the ultrasound image at the standard section of the heart, and then the pre-trained self-supervised deep learning model is used to capture the target standard section in the ultrasound image, wherein the target standard section includes a 2D ultrasound video 400 and a Doppler ultrasound video; finally, the 2D ultrasound video 400 is subjected to key frame extraction, so as to measure the heart structure in the key frame, obtain the heart structure parameters, and then cut out the region of interest in the Doppler ultrasound cardiogram, and segment the region of interest to obtain the functional parameters of the heart. By adopting the above method, various parameters of the heart can be accurately and efficiently obtained, and the whole process is fully automated, avoiding the influence of the experience and ability of medical workers, and ensuring the accuracy and consistency of the various parameter test results of the heart.

[0090] like Figure 4 As shown, in this embodiment, key frame extraction is performed on the 2D echocardiography video 400, which specifically includes the following steps:

[0091] Step 310: measuring cardiac structural parameters of a plurality of image frames in the 2D echocardiography video 400 to obtain a measurement curve 100 of changes in cardiac structural parameters over time.

[0092] In actual operation, generally, it is necessary to measure the cardiac structural parameters of all image frames in the 2D echocardiography video 400 to obtain a measurement curve 100 of the cardiac structural parameters of each image frame changing over time. Alternatively, the cardiac structural parameters of multiple image frames close to the target area to be measured may be selected for measurement, which may be selected according to actual conditions.

[0093] Generally, the cardiac structural parameters we measure are the diameters of various cardiac structures. A diameter represents the position and scale of a parameter measurement, and the length of the diameter can be obtained by locating the two end points of the diameter. Therefore, it is necessary to locate the positions of the two end points (i.e., measurement points) of each measurement parameter, and then determine the value of the diameter of each cardiac structure. The specific measurement method is as follows.

[0094] like Figure 5As shown, in this step, the measurement of cardiac structural parameters of multiple image frames in the 2D echocardiography video 400 specifically includes the following steps:

[0095] Step 311: Use a five-layer convolutional neural network as an encoder to extract image features of multiple image frames in the 2D echocardiography video 400.

[0096] like Figure 8 As shown, the convolution layer 500 of a five-layer CNN (convolutional neural network) is used as an encoder to extract image features of multiple image frames in a 2D echocardiography video 400 to reduce the prediction error of each image frame, thereby reducing the positioning error of the measurement point and improving the accuracy of the measurement point.

[0097] Step 312: The coordinate offsets of the image features of each image frame are sequentially regressed through the fully connected layer 700 of the graph convolutional network 600 to obtain measurement points of the cardiac structure parameters in multiple image frames.

[0098] Then, a shared GCN (graph convolutional network 600) model is designed, and the fully connected layer 700 of the GCN (graph convolutional network 600) model is used to sequentially regress the coordinate offset of the image features of each image frame, and complete the positioning of all measurement parameter measurement points of all frame images in the dynamic video. Through this step, the accurate positioning of the measurement points of the cardiac structure parameters in multiple image frames is achieved.

[0099] Step 313: Measure the distances between the measurement points of the cardiac structural parameters to obtain the cardiac structural parameters.

[0100] Finally, according to the measurement points of the cardiac structural parameters in each image frame, the diameters of various cardiac structures are calculated, and then the cardiac structural parameters of each image frame are obtained.

[0101] After measuring the cardiac structural parameters of multiple image frames in the 2D echocardiography video 400, as shown in FIG. Figure 7 As shown, a measurement curve 100 of the changes of cardiac structural parameters over time in each image frame is obtained according to the positioning network. Since the parameters of the cardiac structure increase or decrease with the contraction and relaxation of the heart (a change in time), the diameters of all parameters can be obtained according to the predicted measurement point positions, and combined with the time information of the image, a measurement curve 100 of the changes of cardiac structural parameters over time in each image frame can be obtained according to the positioning network.

[0102] Step 320: All the measurement curves 100 of the cardiac structural parameters are fused by weighted summation to obtain a cardiac motion curve 200 reflecting the heart beating.

[0103] Step 330: Extract a first electrocardiogram from the 2D echocardiogram video 400, and identify the R-wave apex in the electrocardiogram signal of the first electrocardiogram.

[0104] Step 340: Obtain the maximum value in the heart motion curve 200, and compare the time corresponding to the maximum value with the time corresponding to the R wave apex.

[0105] Step 350: If the time difference between the moment corresponding to the maximum value and the moment corresponding to the R wave apex does not exceed the preset threshold, the maximum and minimum values ​​of the cardiac motion curve 200 are extracted, and the image frames to which the maximum and minimum values ​​of the cardiac motion curve 200 belong are used as key frames.

[0106] Step 360: If the time difference between the moment corresponding to the maximum value and the moment corresponding to the R wave apex exceeds a preset threshold, the R wave apex in the first electrocardiogram and the minimum value of the cardiac motion curve 200 are extracted, and the image frame to which the R wave apex and the minimum value of the cardiac motion curve 200 belong is used as a key frame.

[0107] like Figure 7 As shown, the first electrocardiogram is extracted from the 2D ultrasound video 400. Specifically, it can be connected to the electrocardiogram wire during ultrasound image scanning, and the electrocardiogram signal can be synchronously detected during image acquisition and synchronously displayed below the ultrasound image. Specifically, the image is converted from the RGB (color mode) color domain to the HSV (color model), and then separated from the original ultrasound image according to the color of the electrocardiogram, and the one-dimensional signal of the ECG is obtained by filtering the separated image and according to the position of each pixel on the ECG curve.

[0108] After the electrocardiogram signal is collected, the R wave apex is identified according to the electrocardiogram signal. After the measurement curve 100 of the change of the cardiac structural parameters over time in each image frame is obtained, the multiple measurement curves 100 are fused by weighted summation to obtain a cardiac motion curve 200, which reflects the beating of the heart.

[0109] Among them, the R wave is a very characteristic position on the electrocardiogram waveform, and its apex corresponds to the end of diastole; in the heart motion curve 200, the frame where the maximum value on the curve is located is the end of diastole frame, and the frame where the minimum value is located is the end of systole frame. In clinical medicine, what we want to obtain are the parameter values ​​of the end of systole and end of diastole in the cardiac cycle. Therefore, we use the end of systole and end of diastole in the cardiac cycle as key frames, at which time the volume of the heart is at its maximum or minimum.

[0110] Then, we need to extract the key frame to obtain the key frame result 300. Specifically, by obtaining the maximum and minimum values ​​in the cardiac motion curve 200, the time corresponding to the maximum value is compared with the time corresponding to the R wave apex, which can be specifically compared with the time corresponding to the maximum value of the same cardiac cycle and the time corresponding to the R wave apex. When the time difference between the time corresponding to the maximum value and the time corresponding to the R wave apex does not exceed the preset threshold, the maximum and minimum values ​​of the cardiac motion curve 200 are extracted, and the image frame to which the maximum and minimum values ​​of the cardiac motion curve 200 belong is used as the key frame; when the time difference between the time corresponding to the maximum value and the time corresponding to the R wave apex exceeds the preset threshold, the R wave apex in the first electrocardiogram and the minimum value of the cardiac motion curve 200 are extracted, and the image frame to which the R wave apex and the minimum value of the cardiac motion curve 200 belong is used as the key frame. After determining the key frame, according to the position of the key frame, the diameter of all the measurement parameters on the corresponding image frame is obtained. This result is the cardiac structure parameter we ultimately need in clinical practice.

[0111] The above is the measurement of cardiac structural parameters. In addition, it is also necessary to obtain cardiac functional parameters. The specific acquisition method is as follows.

[0112] like Figure 6 As shown, in this embodiment, the region of interest of the Doppler echocardiogram, that is, the region where the spectrum appears, is cropped, and the region of interest is segmented to obtain the functional parameters of the heart, which specifically includes the following steps:

[0113] Step 410: Obtain a blood flow spectrum image of a Doppler echocardiogram, and extract a second electrocardiogram from the blood flow spectrum image. Specifically, a Doppler echocardiogram is a static image, including a blood flow spectrum image (PW) and a tissue Doppler image (DTI), and extract a second electrocardiogram from the blood flow spectrum image.

[0114] Step 420: Cut out a region of interest from the Doppler echocardiogram according to the cycle information of the second electrocardiogram.

[0115] Step 430: Segment the contour of the blood flow spectrum image within the region of interest using a segmentation model.

[0116] Step 440: Acquire vertex position and area information from the segmented blood flow spectrum image.

[0117] Step 450: Calculate the functional parameters of the heart based on the apex position and area information.

[0118] like Fig. 9As shown, specifically, the second electrocardiogram is extracted from the blood flow spectrum image of the Doppler echocardiogram. Since the second electrocardiogram contains multiple cardiac cycles, the region of interest is cut out from the Doppler echocardiogram according to the cycle information of the second electrocardiogram, and then the contour of the blood flow spectrum image in the region of interest is segmented by the Unet model (Unet segmentation model) to segment the blood flow spectrum image in the region of interest, wherein the Unet model is a codec structure divided into two parts, the encoder on the left performs feature extraction, and the decoder on the right is responsible for feature recovery, thereby segmenting the contour of the blood flow spectrum image. Finally, the vertex position and area information are obtained from the blood flow spectrum image in the segmented region of interest, and the functional parameters of the heart are calculated according to the vertex position and area information. Among them, the average blood flow velocity in the time period can be obtained from the area information, the blood flow rate flowing through the valve orifice in the time period and other information can be calculated by area integration, and the maximum blood flow velocity and the average blood flow velocity through the valve orifice and other information can be extracted from the vertex information.

[0119] This embodiment provides a cardiac parameter measurement system, such as Fig.10 As shown, it includes a probe positioning module 10, a target standard section capture module 20, a heart structure parameter measurement module 30 and a heart function parameter measurement module 40. Specifically, the probe positioning module 10 is used to obtain the position of the heart, and finely position the probe according to the position of the heart; after fine positioning, the probe obtains an ultrasound image containing a target standard section. The target standard section capture module 20 is used to capture the target standard section in the ultrasound image through a pre-trained self-supervised deep learning model; the target standard section includes a 2D echocardiogram video 400 and a Doppler echocardiogram. The heart structure parameter measurement module 30 is used to extract key frames from the 2D echocardiogram video 400, measure the structural parameters of the heart in the key frames, and obtain the heart structure parameters. The heart function parameter measurement module 40 is used to crop the region of interest of the Doppler echocardiogram, segment the region of interest, and obtain the function parameters of the heart. Its specific implementation method has been described in detail in the above method embodiment. Please refer to the above method embodiment for details. This embodiment will not be repeated in detail here.

[0120] In this embodiment, reference Fig.11, the probe positioning module 10 includes a coarse positioning submodule 11, an auscultation information acquisition submodule 12, a fine positioning submodule 13, an ultrasonic image acquisition submodule 14 and a fine positioning submodule 15. Specifically, the coarse positioning submodule 11 is used to segment the human body position according to the visual information of the inspection object, and locate the approximate position of the heart from the human body position through a preset human body physical model. The auscultation information acquisition submodule 12 is used to obtain the auscultation information of the approximate position of the heart. The fine positioning submodule 13 is used to extract features from the auscultation information, and locate the target area of ​​the heart according to the extracted auscultation information features. The ultrasonic image acquisition submodule 14 is used to move the probe to the target area and obtain the ultrasonic image of the target area. The fine positioning submodule 15 is used to determine the target position in the target area according to the ultrasonic image of the target area. The target position is the position where the probe is located when scanning the target standard section, so that the probe can obtain the ultrasonic image of the target standard section. Its specific implementation method has been described in detail in the above method embodiment. Please refer to the above method embodiment for details. This embodiment will not be repeated in detail here.

[0121] In this embodiment, in the target standard section capturing module 20, the target standard section in the ultrasound image is captured by a pre-trained self-supervised deep learning model, including:

[0122] Pre-build a single classifier in a self-supervised deep learning model.

[0123] The collected standard sections are used to train a single classifier, so that the single classifier can be used to identify the target standard section in the ultrasound image.

[0124] The ultrasound image is detected for abnormalities through a trained single classifier. When the single classifier detects an abnormality, it is identified as a non-target standard section. When a non-abnormality is detected, it is identified as a target standard section.

[0125] In this embodiment, the specific implementation method has been described in detail in the above method embodiment. Please refer to the above method embodiment for details, and this embodiment will not be described in detail here.

[0126] This embodiment provides a computer-readable storage medium, on which a program is stored. The program can be executed by a processor to implement the method steps of the above embodiment, and this embodiment will not be described in detail here.

[0127] Those skilled in the art will appreciate that all or part of the functions of the various methods in the above-mentioned embodiments can be implemented by hardware or by computer programs. When all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, and the storage medium can include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the program is executed by a computer to implement the above-mentioned functions. For example, the program is stored in the memory of the device, and when the program in the memory is executed by the processor, all or part of the above-mentioned functions can be implemented. In addition, when all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can also be stored in a storage medium such as a server, another computer, disk, optical disk, flash disk or mobile hard disk, and can be downloaded or copied and saved in the memory of the local device, or the system of the local device is updated, and when the program in the memory is executed by the processor, all or part of the functions in the above-mentioned embodiments can be implemented.

[0128] The above specific examples are used to illustrate the present invention, which are only used to help understand the present invention and are not used to limit the present invention. For those skilled in the art to which the present invention belongs, some simple deductions, modifications or substitutions can be made according to the idea of ​​the present invention.

Claims

1. A cardiac parameter measurement method, characterized in that: include: Acquiring the position of the heart, and finely positioning the probe according to the position of the heart, wherein the probe acquires an ultrasound image including a target standard section after fine positioning; Capturing target standard sections in the ultrasound image by using a pre-trained self-supervised deep learning model; the target standard sections include 2D echocardiogram video and Doppler echocardiogram; Extracting key frames from the 2D echocardiography video, and measuring cardiac structural parameters in the key frames to obtain cardiac structural parameters; The region of interest of the Doppler echocardiogram is cropped and segmented to obtain the functional parameters of the heart.

2. The cardiac parameter measurement method according to claim 1, characterized in that: The step of obtaining the position of the heart and finely positioning the probe according to the position of the heart includes: Segmenting the human body position according to the visual information of the inspection object, and locating the approximate position of the heart from the human body position by using a preset human body physical model; Obtain auscultatory information of the approximate position of the heart; Extracting features from the auscultation information, and locating a target area of ​​the heart according to the extracted features of the auscultation information; Moving the probe to the target area to obtain an ultrasonic image of the target area; The target position in the target area is determined according to the ultrasonic image of the target area, and the target position is the position of the probe when scanning the target standard section, so that the probe can obtain the ultrasonic image of the target standard section.

3. The cardiac parameter measurement method according to claim 1, characterized in that: The capturing of the target standard section in the ultrasound image by using a pre-trained self-supervised deep learning model includes: Pre-build a single classifier in a self-supervised deep learning model; Using the collected standard sections to train the single classifier, so that the single classifier is used to identify the target standard section in the ultrasound image; The ultrasound image is subjected to abnormality detection by the trained single classifier. When the single classifier detects an abnormality, it is identified as a non-target standard section. When the single classifier detects a non-abnormality, it is identified as a target standard section.

4. The cardiac parameter measurement method according to claim 1, characterized in that: The step of extracting key frames from the 2D echocardiography video comprises: Measuring cardiac structural parameters of a plurality of image frames in the 2D echocardiography video to obtain a measurement curve of changes in cardiac structural parameters over time; The measurement curves of all cardiac structural parameters are fused by weighted summation to obtain a cardiac motion curve reflecting the heart beating; extracting a first electrocardiogram from the 2D echocardiogram video, and identifying an R-wave apex in an electrocardiogram signal of the first electrocardiogram; Obtaining a maximum value in the heart motion curve, and comparing a time corresponding to the maximum value with a time corresponding to the R wave apex; If the time difference between the moment corresponding to the maximum value and the moment corresponding to the R wave apex does not exceed the preset threshold, the maximum and minimum values ​​of the cardiac motion curve are extracted, and the image frames to which the maximum and minimum values ​​of the cardiac motion curve belong are used as key frames; if the time difference between the moment corresponding to the maximum value and the moment corresponding to the R wave apex exceeds the preset threshold, the R wave apex in the first electrocardiogram and the minimum value of the cardiac motion curve are extracted, and the image frames to which the R wave apex and the minimum value of the cardiac motion curve belong are used as key frames.

5. The cardiac parameter measurement method according to claim 4, characterized in that: The measuring of cardiac structural parameters of a plurality of image frames in the 2D echocardiography video comprises: A five-layer convolutional neural network is used as an encoder to extract image features of multiple image frames in the 2D echocardiography video; Performing sequential regression on the coordinate offset of the image features of each image frame through a fully connected layer of a graph convolutional network to obtain measurement points of the cardiac structure parameters in the multiple image frames; The distances between the measurement points of the cardiac structural parameters are measured to obtain the cardiac structural parameters.

6. The cardiac parameter measurement method according to claim 1, characterized in that: The step of cropping the region of interest of the Doppler echocardiogram and performing segmentation processing on the region of interest to obtain the functional parameters of the heart includes: Acquiring a blood flow spectrum image of the Doppler echocardiogram, and extracting a second electrocardiogram from the blood flow spectrum image; cutting out a region of interest from the Doppler echocardiogram according to the cycle information of the second electrocardiogram; Segmenting the contour of the blood flow spectrum image within the region of interest by using a segmentation model; Acquiring vertex position and area information from the segmented blood flow spectrum image; The functional parameters of the heart are calculated based on the apex position and area information.

7. A cardiac parameter measurement system, characterized in that: include: A probe positioning module is used to obtain the position of the heart and finely position the probe according to the position of the heart; after fine positioning, the probe obtains an ultrasound image containing a target standard section; A target standard section capture module, used to capture the target standard section in the ultrasound image through a pre-trained self-supervised deep learning model; the target standard section includes a 2D echocardiogram video and a Doppler echocardiogram; A cardiac structural parameter measurement module, used for extracting key frames from the 2D echocardiography video, measuring the cardiac structural parameters in the key frames, and obtaining cardiac structural parameters; The cardiac function parameter measurement module is used to crop the region of interest of the Doppler echocardiogram, segment the region of interest, and obtain the cardiac function parameters.

8. The cardiac parameter measurement system according to claim 7, characterized in that: The probe positioning module comprises: A coarse positioning submodule, used to segment the human body position according to the visual information of the inspection object, and locate the approximate position of the heart from the human body position through a preset human body physical model; An auscultation information acquisition submodule is used to obtain auscultation information of the approximate position of the heart; A fine positioning submodule, used for extracting features from the auscultation information and locating a target area of ​​the heart according to the extracted features of the auscultation information; An ultrasonic image acquisition submodule, used for moving the probe to the target area to acquire an ultrasonic image of the target area; The fine positioning submodule is used to determine the target position in the target area according to the ultrasonic image of the target area, and the target position is the position of the probe when scanning the target standard section, so that the probe can obtain the ultrasonic image of the target standard section.

9. The cardiac parameter measurement system according to claim 7, characterized in that: In the target standard section capture module, the target standard section in the ultrasound image is captured by a pre-trained self-supervised deep learning model, including: Pre-build a single classifier in a self-supervised deep learning model; Using the collected standard sections to train the single classifier, so that the single classifier is used to identify the target standard section in the ultrasound image; The ultrasound image is subjected to abnormality detection by the trained single classifier. When the single classifier detects an abnormality, it is identified as a non-target standard section. When the single classifier detects a non-abnormality, it is identified as a target standard section.

10. A computer-readable storage medium, characterized in that: The medium stores a program, which can be executed by a processor to implement the method according to any one of claims 1 to 6.

Images