Automatic measurement and analysis method, system and medium for heart ultrasound image

Abstract

This invention discloses an automatic measurement and analysis method, system, and medium for cardiac ultrasound images. The method includes the following steps: acquiring cardiac ultrasound images; classifying the cardiac ultrasound images into ultrasound sections based on a ResNet deep learning model; segmenting the classified ultrasound sections using a UNet segmentation model; identifying the mitral valve state in each segmented section and selecting an end-diastolic target frame based on the mitral valve state corresponding to each frame; calculating cardiac parameters in the end-diastolic target frame and generating an ultrasound report based on the cardiac parameters in multiple segmented sections using an ultrasound report generation model. Therefore, this invention can automatically segment and analyze cardiac ultrasound images and automatically generate measurement and analysis reports.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to an automatic measurement and analysis method, system and medium for cardiac ultrasound images. Background Technology

[0002] Cardiac ultrasound imaging is an important tool for the diagnosis of cardiovascular diseases. However, current ultrasound image analysis mainly relies on human experience, which has the following problems: image analysis depends on the doctor's experience and is easily affected by personal subjective factors, resulting in strong subjectivity; doctors need to manually mark cardiac structures for measurement, which is inefficient, has large errors, and is therefore cumbersome and prone to errors.

[0003] To address the aforementioned issues, an automated solution is needed to automatically segment and analyze cardiac ultrasound images and generate measurement and analysis reports. Summary of the Invention

[0004] The present invention provides an automatic measurement and analysis method, system and medium for cardiac ultrasound images, which can automatically segment and analyze cardiac ultrasound images and automatically generate measurement and analysis reports.

[0005] Firstly, an automated measurement and analysis method for cardiac ultrasound images is provided, comprising the following steps:

[0006] Acquire cardiac ultrasound images;

[0007] The cardiac ultrasound images were classified using an ultrasound cross-section based on a ResNet deep learning model.

[0008] Image segmentation of classified ultrasonic sections is performed based on the UNet segmentation model;

[0009] Identify the mitral valve state in each frame segmentation section, and select the end-diastolic target frame according to the mitral valve state corresponding to each frame;

[0010] Calculate cardiac parameters in the target frame at end-diastole;

[0011] An ultrasound report generation model is used to generate ultrasound reports based on cardiac parameters in multiple segmented sections.

[0012] According to the first aspect, in a first possible implementation of the first aspect, the step of identifying the mitral valve state in each frame segmentation plane and selecting the end-diastolic target frame according to the mitral valve state corresponding to each frame includes:

[0013] Extract all mitral lobe contours in each frame segmentation section based on a contour extraction algorithm;

[0014] When the number of mitral valve contours in a single frame segmentation plane is greater than or equal to a preset threshold, the two mitral valve contours with the largest area are selected, and the mitral valve state is determined based on the Euclidean distance between the centroids of the two selected mitral valve contours.

[0015] When the number of mitral valve contours in a single frame segmentation plane is less than a preset threshold, the mitral valve state is determined based on the aspect ratio of the circumscribed rectangle of the mitral valve contour.

[0016] Among all the segmented sections that indicate the mitral valve is closed, the frame with the largest left ventricular region is selected as the end-diastolic target frame.

[0017] According to the first possible implementation of the first aspect, in the second possible implementation of the first aspect, the step of determining the mitral valve state based on the Euclidean distance between the centroids of the two selected mitral valve profiles includes:

[0018] If the Euclidean distance between the centroids of the two mitral valve profiles is less than a preset distance threshold, then the mitral valve state is determined to be valve closure.

[0019] The step of determining the mitral valve condition based on the aspect ratio of the circumscribed rectangle of the mitral valve profile includes:

[0020] If the aspect ratio of the circumscribed rectangle of the mitral valve profile is within a preset range, the mitral valve state is determined to be valve closure.

[0021] According to the first aspect, in a third possible implementation of the first aspect, the step of calculating cardiac parameters in the end-diastolic target frame includes:

[0022] The right ventricular outflow tract contour is extracted from the target frame at the end of diastole, and the maximum and minimum values ​​of the ordinate of all points in the right ventricular outflow tract contour are calculated.

[0023] The pixel distance between the front and rear diameters of the RVOT is obtained by subtracting the maximum and minimum values ​​of the ordinate, and then the pixel distance is converted into the actual distance between the front and rear diameters of the RVOT.

[0024] According to the first aspect, in a fourth possible implementation of the first aspect, the step of calculating cardiac parameters in the end-diastolic target frame includes:

[0025] Extract the aortic contour point set from the target frame at the end of diastole;

[0026] Define a center point, divide the aortic contour point set into four quadrants based on the center point, and find the target point farthest from the center point in each quadrant;

[0027] The longitudinal distance between the two target points furthest from the center point is taken as the diameter of the ascending aorta.

[0028] According to the first aspect, in the fifth possible implementation of the first aspect, the functional expression of the ResNet deep learning model is as follows:

[0029] ;

[0030] In the formula, X is the input feature map; F is the residual mapping function; y represents the trainable convolutional weights in the network; y represents the output feature map.

[0031] According to the first aspect, in the sixth possible implementation of the first aspect, the functional expression of the UNet segmentation model is as follows:

[0032] ;

[0033] The loss function L of the UNet segmentation model Dice as follows:

[0034] ;

[0035] In the formula, X is the input image; It is a nonlinear mapping function; Y is the predicted mask; θ is the actual mask; and θ is the model parameter.

[0036] According to the first aspect, in the seventh possible implementation of the first aspect, the functional expression of the ultrasound report generation model is as follows:

[0037] ;

[0038] In the formula, X is the input image; It is a nonlinear mapping function; The mask is for prediction; θ represents the model parameters.

[0039] The reward loss function of the ultrasound report generation model as follows:

[0040] ;

[0041] In the formula, For prompts and reply With parameters The reward value under the reward model; To indicate preferred response results; To mark the responses you don't like; The number of responses displayed for each prompt.

[0042] In a second aspect, the present invention provides an automatic measurement and analysis system for cardiac ultrasound images, comprising:

[0043] Image acquisition module, used to acquire cardiac ultrasound images;

[0044] The classification module is communicatively connected to the image acquisition module and is used to perform ultrasound section classification on the cardiac ultrasound images based on the ResNet deep learning model.

[0045] The segmentation module is communicatively connected to the classification module and is used to segment the classified ultrasound cross-section based on the UNet segmentation model.

[0046] The target frame selection module is communicatively connected to the segmentation module and is used to identify the mitral valve state in each frame segmentation section and select the end-diastolic target frame according to the mitral valve state corresponding to each frame.

[0047] The parameter calculation module, communicatively connected to the target frame selection module, is used to calculate cardiac parameters in the end-diastolic target frame; and,

[0048] The report generation module is communicatively connected to the parameter calculation module and is used to generate an ultrasound report based on cardiac parameters in multiple segmented sections using the ultrasound report generation model.

[0049] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the automatic measurement and analysis method for cardiac ultrasound images as described above.

[0050] Fourthly, the present invention provides an electronic device, including a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that the processor, when running the computer program, implements the automatic measurement and analysis method for cardiac ultrasound images as described above.

[0051] Compared with existing technologies, the advantages of this invention are as follows: It classifies echocardiogram images into sections based on a ResNet deep learning model; then segments the classified sections using a UNet segmentation model; next, it identifies the mitral valve state in each segmented section and selects an end-diastolic target frame based on the corresponding mitral valve state; then, it calculates cardiac parameters in the end-diastolic target frame, and finally uses an ultrasound report generation model to generate an ultrasound report based on the cardiac parameters from multiple segmented sections. Therefore, this invention can automatically segment and analyze echocardiogram images and automatically generate measurement and analysis reports. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating an embodiment of an automatic measurement and analysis method for cardiac ultrasound images according to the present invention;

[0053] Figure 2 This is a flowchart illustrating another embodiment of the automatic measurement and analysis method for cardiac ultrasound images according to the present invention.

[0054] Figure 3 This is a schematic diagram of the structure of an automatic measurement and analysis system for cardiac ultrasound images according to the present invention. Detailed Implementation

[0055] Referring now to specific embodiments of the invention, examples of which are illustrated in the accompanying drawings. Although the invention will be described in conjunction with specific embodiments, it will be understood that it is not intended to limit the invention to the described embodiments. Rather, it is intended to cover variations, modifications, and equivalents included within the spirit and scope of the invention as defined by the appended claims. It should be noted that the method steps described herein can be implemented by any functional block or functional arrangement, and any functional block or functional arrangement can be implemented as a physical entity or a logical entity, or a combination of both.

[0056] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] Note: The examples described below are merely specific examples and are not intended to limit the embodiments of the present invention to the specific steps, values, conditions, data, order, etc. Those skilled in the art can utilize the concept of the present invention to construct more embodiments not mentioned herein by reading this specification.

[0058] See Figure 1 As shown, this embodiment of the invention provides an automatic measurement and analysis method for cardiac ultrasound images, including the following steps:

[0059] S100, acquires cardiac ultrasound images;

[0060] S200, Ultrasonic section classification of the cardiac ultrasound images is performed based on the ResNet deep learning model;

[0061] S300, based on the UNet segmentation model, performs image segmentation on the classified ultrasonic cross-section;

[0062] S400, identify the mitral valve state in each frame segmentation section, and select the end-diastolic target frame according to the mitral valve state corresponding to each frame;

[0063] S500, calculate cardiac parameters in the end-diastolic target frame;

[0064] The S600 uses an ultrasound report generation model to generate ultrasound reports based on cardiac parameters in multiple segmented sections.

[0065] Specifically, in this embodiment, current ultrasound image analysis mainly relies on human experience, which presents the following problems: image analysis depends on the doctor's experience, is easily affected by personal subjective factors, and is highly subjective; doctors need to manually annotate cardiac structures for measurement, which is inefficient, has large errors, and is therefore cumbersome and prone to errors. Therefore, to address these problems, this invention designs an automatic measurement and analysis method for cardiac ultrasound images. This method classifies cardiac ultrasound images into ultrasound sections based on a ResNet deep learning model; then segments the classified ultrasound sections using a UNet segmentation model; next, it identifies the mitral valve state in each segmented section and selects an end-diastolic target frame based on the corresponding mitral valve state; then, it calculates cardiac parameters in the end-diastolic target frame; and finally, it uses an ultrasound report generation model to generate an ultrasound report based on the cardiac parameters from multiple segmented sections. Therefore, this invention can automatically segment and analyze cardiac ultrasound images and automatically generate measurement and analysis reports.

[0066] Preferably, in another embodiment of this application, the residual network function expression of the ResNet deep learning model is as follows:

[0067] ;

[0068] In the formula, X is the input feature map, which is directly added to the result of the residual function through identity mapping; F is the residual mapping function, which is usually composed of several convolutional layers, activation functions, etc. y represents the trainable convolutional weights in the network; y represents the output feature map.

[0069] This residual network effectively solves the gradient vanishing problem in deep network training and improves the accuracy of ultrasound image section classification. Using the ResNet deep learning model for automatic classification of ultrasound cardiac images, it can identify 15 common sections, such as the parasternal long axis, apical 4C two-dimensional, apical 2C two-dimensional, parasternal mitral valve short axis, and parasternal papillary muscle short axis.

[0070] The ResNet deep learning model was also used to classify ultrasound images into three categories: excellent, qualified, and unqualified. Only images with qualified quality were analyzed later.

[0071] Preferably, in another embodiment of this application, the present invention employs a deep neural network model based on U-Net to automatically segment cardiac anatomical structures in ultrasound images. This model extracts image features through an encoder and combines a decoder and skip connections to reconstruct spatial information, ultimately outputting a structural mask with the same size as the original image. The nonlinear mapping function expression of the UNet segmentation model is as follows:

[0072] ;

[0073] The loss function L of the UNet segmentation model Dice as follows:

[0074] ;

[0075] In the formula, X is the input image; It is a nonlinear mapping function; Y is the predicted mask; θ is the actual mask; and θ is the model parameter.

[0076] Preferably, in another embodiment of this application, step S400, which identifies the mitral valve state in each frame segmentation plane and selects the end-diastolic target frame according to the mitral valve state corresponding to each frame, includes:

[0077] S410, based on the contour extraction algorithm, extracts all mitral lobe contours in each frame segmentation plane;

[0078] S420, when the number of mitral valve contours in a single frame segmentation plane is greater than or equal to a preset threshold, the two mitral valve contours with the largest area are selected, and the mitral valve state is determined based on the Euclidean distance between the centroids of the two selected mitral valve contours.

[0079] S430, when the number of mitral valve contours in a single frame segmentation plane is less than a preset threshold, the mitral valve valve state is determined based on the aspect ratio of the circumscribed rectangle of the mitral valve contour.

[0080] S440, among all the segmented sections that determine the mitral valve state as valve closure, select the frame with the largest left ventricular region as the end-diastolic target frame.

[0081] Preferably, in another embodiment of this application, the step of determining the mitral valve state based on the Euclidean distance between the centroids of the two selected mitral valve profiles includes:

[0082] If the Euclidean distance between the centroids of the two mitral valve profiles is less than a preset distance threshold, then the mitral valve state is determined to be valve closure.

[0083] The step of determining the mitral valve condition based on the aspect ratio of the circumscribed rectangle of the mitral valve profile includes:

[0084] If the aspect ratio of the circumscribed rectangle of the mitral valve profile is within a preset range, the mitral valve state is determined to be valve closure.

[0085] Specifically, in this embodiment, for the mitral valve region in each frame of segmented image, the contour extraction algorithm OpenCV.findContours() is used to obtain all contours, and the number of contours is denoted as N; different strategies are used to determine the valve state according to different N.

[0086] Case 1: If N>= 2, it means that the mitral valve is in an open state (the anterior and posterior leaflets are separated).

[0087] Extract the two contours C1 and C2 with the largest areas, and calculate the centroids P1 and P2 of the two contours C1 and C2 respectively;

[0088] Calculate the Euclidean distance between the two centroids, D = P1 - P2;

[0089] If D is greater than the preset distance threshold T, the mitral valve state can be determined to be open; otherwise, the mitral valve state can be determined to be closed.

[0090] Case 2: N = 1, indicating that the valve closure or overlap is a single structure;

[0091] For a unique contour C, calculate the aspect ratio r = w / h of its bounding rectangle;

[0092] If r satisfies the geometric characteristics of a closed valve (such as an approximate slender shape), that is, the aspect ratio of the circumscribed rectangle of the mitral valve profile is within a preset range, then the mitral valve state is determined to be valve closure.

[0093] All frames are obtained within a cardiac cycle. Then, all frames are judged to obtain all frames judged as valve closure. Among all the segmented sections that indicate the mitral valve is closed, the marked left ventricular region is calculated, and the frame with the largest left ventricular region is judged as the end-diastolic target frame.

[0094] Preferably, in another embodiment of this application, step S500, calculating cardiac parameters in the end-diastolic target frame, includes:

[0095] The right ventricular outflow tract contour is extracted from the target frame at the end of diastole, and the maximum and minimum values ​​of the ordinate of all points in the right ventricular outflow tract contour are calculated; specifically as follows:

[0096] ;

[0097] The pixel distance between the front and rear diameters of the RVOT is obtained by subtracting the maximum and minimum values ​​of the ordinate, as shown in the following formula:

[0098] ;

[0099] The pixel distance is then converted into the actual distance of the front and rear diameters of the RVOT. That is, if the image has physical calibration information (such as 1 mm = s pixels), it is converted into the actual distance.

[0100] For example, 1mm = 36 pixels, Ymax − Ymin = 72 pixels;

[0101] Therefore, 72 * 1 / 36 = 2mm, so the actual length of the RVOT front and rear diameters is 2mm.

[0102] Preferably, in another embodiment of this application, step S500, calculating cardiac parameters in the end-diastolic target frame, includes:

[0103] Extract the aortic contour point set from the target frame at the end of diastole;

[0104] Define a center point, divide the aortic contour point set into four quadrants based on the center point, and find the target point farthest from the center point in each quadrant;

[0105] The longitudinal distance between the two target points furthest from the center point is taken as the diameter of the ascending aorta.

[0106] Specifically, in this embodiment, the aortic mask contour point set is set as {P_i = (x_i, y_i)};

[0107] Define a center point (x_c, y_c), which is the geometric center of the contour points. Average the x and y coordinates of all contour points to obtain a 'center point'.

[0108] Based on the center point, the outline points are divided into four quadrants;

[0109]

[0110] Find the point farthest from the center point in each quadrant:

[0111] A1: Northwest point (top left)

[0112] A2: Northeast (top right)

[0113] B1: Southwest point (bottom left)

[0114] B2: Southeast point (bottom right)

[0115] The anteroposterior diameter of the ascending aorta is then calculated as follows:

[0116] If the two points furthest from the center point are A2 (upper right) and B2 (lower right), then the longitudinal distance between the two points is taken as the diameter of the ascending aorta.

[0117] .

[0118] The functional expression of the ultrasound report generation model is as follows:

[0119] ;

[0120] In the formula, X is the input image; It is a nonlinear mapping function; θ represents the prediction mask; θ represents the model parameters.

[0121] Efficient fine-tuning of indicators is performed using reinforcement learning to further align (or align) the model's output with the ultrasound scene. Specifically, a reward loss function is used for model fine-tuning. The training process of the reward model is achieved by fitting the doctor's tendency to respond to different responses. First, a supervised model fine-tuned based on indicator data is used, sampling the same indicator. Each reply is displayed in pairs to the annotator, with each prompt showing a total of [number] replies. One reply result; then... The responses constitute the training samples for the reward model. Instead of direct scoring, labelers provide biased ranking labels, manually assigning biased responses lower in the ranking to encourage the model to avoid generating content that doctors dislike. Finally, the reward model calculates the bias probability by fitting the doctor labels to the difference in reward values ​​between two responses in the sample, completing the training of the reward model. The loss function of the reward model-ultrasound report generation model can be expressed as:

[0122] ;

[0123] In the formula, For prompts and reply With parameters The reward value under the reward model; To indicate preferred response results; To mark the responses you don't like; The number of responses displayed for each prompt. The goal of this loss function is to maximize the difference between responses that annotators prefer and those they dislike.

[0124] See also Figure 3 As shown in the figure, an automatic measurement and analysis system for cardiac ultrasound images provided in this embodiment of the invention includes:

[0125] Image acquisition module, used to acquire cardiac ultrasound images;

[0126] The classification module is communicatively connected to the image acquisition module and is used to perform ultrasound section classification on the cardiac ultrasound images based on the ResNet deep learning model.

[0127] The segmentation module is communicatively connected to the classification module and is used to segment the classified ultrasound cross-section based on the UNet segmentation model.

[0128] The target frame selection module is communicatively connected to the segmentation module and is used to identify the mitral valve state in each frame segmentation section and select the end-diastolic target frame according to the mitral valve state corresponding to each frame.

[0129] The parameter calculation module is communicatively connected to the target frame selection module and is used to calculate cardiac parameters in the end-diastolic target frame.

[0130] The report generation module is communicatively connected to the parameter calculation module and is used to generate an ultrasound report based on cardiac parameters in multiple segmented sections using the ultrasound report generation model.

[0131] In summary, the automatic measurement and analysis system for cardiac ultrasound images designed in this invention classifies cardiac ultrasound images into ultrasound sections based on a ResNet deep learning model; then, it segments the classified ultrasound sections using a UNet segmentation model; next, it identifies the mitral valve state in each segmented section and selects the end-diastolic target frame based on the corresponding mitral valve state; then, it calculates cardiac parameters in the end-diastolic target frame; and finally, it generates an ultrasound report based on the cardiac parameters from multiple segmented sections using an ultrasound report generation model. Therefore, this invention can automatically segment and analyze cardiac ultrasound images and automatically generate measurement and analysis reports.

[0132] Specifically, this embodiment corresponds one-to-one with the above method embodiments. The functions of each module have been described in detail in the corresponding method embodiments, so they will not be repeated here.

[0133] Based on the same inventive concept, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements all or part of the method steps of the above method.

[0134] The present invention can implement all or part of the processes in the above methods, or it can be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0135] Based on the same inventive concept, embodiments of this application also provide an electronic device, including a memory and a processor. The memory stores a computer program that runs on the processor. When the processor executes the computer program, it implements all or part of the method steps described above.

[0136] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the computer device, connecting all parts of the computer device through various interfaces and lines.

[0137] Memory can be used to store computer programs and / or modules. The processor performs various functions of the computer device by running or executing the computer programs and / or modules stored in the memory, and by accessing data stored in the memory. Memory can primarily include a program storage area and a data storage area. The program storage area can store the operating system and at least one application program required for a function (e.g., sound playback, image playback, etc.); the data storage area can store data created based on the use of the mobile phone (e.g., audio data, video data, etc.). Furthermore, memory can include high-speed random access memory, and can also include non-volatile memory, such as hard disks, RAM, plug-in hard disks, SmartMedia Cards (SMC), Secure Digital (SD) cards, Flash Cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0138] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, servers, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0139] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), servers, and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0140] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0141] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0142] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for automatic measurement and analysis of cardiac ultrasound images, characterized in that, Includes the following steps: Acquire cardiac ultrasound images; The cardiac ultrasound images were classified using an ultrasound cross-section based on a ResNet deep learning model. Image segmentation of classified ultrasonic sections is performed based on the UNet segmentation model; Identify the mitral valve state in each frame segmentation section, and select the end-diastolic target frame according to the mitral valve state corresponding to each frame; Calculate cardiac parameters in the end-diastolic target frame; An ultrasound report generation model is used to generate ultrasound reports based on cardiac parameters in multiple segmented sections. The step of identifying the mitral valve state in each frame segmentation section and selecting the end-diastolic target frame based on the corresponding mitral valve state for each frame includes: Extract all mitral lobe contours in each frame segmentation section based on a contour extraction algorithm; When the number of mitral valve contours in a single frame segmentation plane is greater than or equal to a preset threshold, the two mitral valve contours with the largest area are selected, and the mitral valve state is determined based on the Euclidean distance between the centroids of the two selected mitral valve contours. When the number of mitral valve contours in a single frame segmentation plane is less than a preset threshold, the mitral valve state is determined based on the aspect ratio of the circumscribed rectangle of the mitral valve contour. Among all the segmented sections that indicate the mitral valve is closed, the frame with the largest left ventricular region is selected as the end-diastolic target frame.

2. The automatic measurement and analysis method for cardiac ultrasound images as described in claim 1, characterized in that, The step of determining the mitral valve condition based on the Euclidean distance between the centroids of the two selected mitral valve profiles includes: If the Euclidean distance between the centroids of the two mitral valve profiles is less than a preset distance threshold, then the mitral valve state is determined to be valve closure. The step of determining the mitral valve condition based on the aspect ratio of the circumscribed rectangle of the mitral valve profile includes: If the aspect ratio of the circumscribed rectangle of the mitral valve profile is within a preset range, the mitral valve state is determined to be valve closure.

3. The method of claim 1, wherein the step of automatically measuring and analyzing the cardiac ultrasound image is performed by a computer. The step of calculating cardiac parameters in the end-diastolic target frame includes: The right ventricular outflow tract contour is extracted from the target frame at the end of diastole, and the maximum and minimum values ​​of the ordinate of all points in the right ventricular outflow tract contour are calculated. The pixel distance between the front and rear diameters of the RVOT is obtained by subtracting the maximum and minimum values ​​of the ordinate, and then the pixel distance is converted into the actual distance between the front and rear diameters of the RVOT.

4. The method of claim 1, wherein the method further comprises: The step of calculating cardiac parameters in the end-diastolic target frame includes: Extract the aortic contour point set from the target frame at the end of diastole; Define a center point, divide the aortic contour point set into four quadrants based on the center point, and find the target point farthest from the center point in each quadrant; The longitudinal distance between the two target points furthest from the center point is taken as the diameter of the ascending aorta.

5. The method of claim 1, wherein the step of automatically measuring and analyzing the cardiac ultrasound image is performed by a computer. The functional expression of the ResNet deep learning model is as follows: ； In the formula, X is an input feature map; F is a residual mapping function; is a trainable convolution weight parameter in the network; and y is an output feature map.

6. The method of automatic measurement and analysis of cardiac ultrasound images of claim 1, wherein, The functional expression of the UNet segmentation model is as follows: ； The loss function L of the UNet segmentation model Dice as follows: ； In the formula, X is the input image; It is a nonlinear mapping function; Y is the predicted mask; θ is the actual mask; and θ is the model parameter.

7. The automatic measurement and analysis method for cardiac ultrasound images as described in claim 1, characterized in that, The functional expression of the ultrasound report generation model is as follows: ； In the formula, X is the input image; It is a nonlinear mapping function; The mask is for prediction; θ represents the model parameters. The reward loss function of the ultrasound report generation model as follows: ； In the formula, For prompts and reply With parameters The reward value under the reward model; To indicate preferred response results; To mark the responses you don't like; The number of responses displayed for each prompt.

8. An automatic measurement and analysis system for cardiac ultrasound images, characterized in that, include: Image acquisition module, used to acquire cardiac ultrasound images; The classification module is communicatively connected to the image acquisition module and is used to perform ultrasound section classification on the cardiac ultrasound images based on the ResNet deep learning model. The segmentation module is communicatively connected to the classification module and is used to segment the classified ultrasound cross-section based on the UNet segmentation model. The target frame selection module is communicatively connected to the segmentation module and is used to identify the mitral valve state in each frame segmentation section and select the end-diastolic target frame according to the mitral valve state corresponding to each frame. The parameter calculation module is communicatively connected to the target frame selection module and is used to calculate cardiac parameters in the end-diastolic target frame. The report generation module is communicatively connected to the parameter calculation module and is used to generate an ultrasound report based on cardiac parameters in multiple segmented sections using the ultrasound report generation model. The step of identifying the mitral valve state in each frame segmentation section and selecting the end-diastolic target frame based on the corresponding mitral valve state for each frame includes: Extract all mitral lobe contours in each frame segmentation section based on a contour extraction algorithm; When the number of mitral valve contours in a single frame segmentation plane is greater than or equal to a preset threshold, the two mitral valve contours with the largest area are selected, and the mitral valve state is determined based on the Euclidean distance between the centroids of the two selected mitral valve contours. When the number of mitral valve contours in a single frame segmentation plane is less than a preset threshold, the mitral valve state is determined based on the aspect ratio of the circumscribed rectangle of the mitral valve contour. Among all the segmented sections that indicate the mitral valve is closed, the frame with the largest left ventricular region is selected as the end-diastolic target frame.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by the processor, it implements the automatic measurement and analysis method for cardiac ultrasound images as described in any one of claims 1 to 7.

Images