Section classification method and device for cardiac ultrasound image and computer equipment

Abstract

The invention relates to a section classification method and device for a cardiac ultrasound image, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: performing feature extraction on a cardiac ultrasound image to obtain image features; semantic segmentation and position detection are carried out based on the image features, a semantic segmentation map and angular point position information are obtained, the semantic segmentation map comprises a plurality of pixel points and tissues to which each pixel point belongs, the angular point position information represents the position of a heart in the cardiac ultrasound image, and the heart comprises a plurality of tissues; based on the angular point position information, converting initial position information of each pixel point in the semantic segmentation map into polar coordinate information; and determining the section type of the cardiac ultrasound image based on the polar coordinate information of each pixel point and the tissue to which each pixel point belongs. By adopting the method, the problems of weak generalization, insufficient classification granularity, limited quality control effect and the like in current echocardiogram classification can be effectively solved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, apparatus, computer device, computer-readable storage medium and computer program product for classifying sections of cardiac ultrasound images. Background Art

[0002] Echocardiography is the most widely used cardiac examination method. In an echocardiography examination, ultrasound physicians follow clinical guidelines to manipulate the ultrasound probe to try to scan different sections of the heart from multiple angles. Therefore, the use of echocardiography requires users to pay a very high learning cost. If ultrasound physicians can automatically identify the type of cardiac section and evaluate the quality of the current section to achieve automated recording, the cardiac ultrasound examination process can be greatly simplified.

[0003] However, the existing echocardiography classification technology directly uses the original image as input, which has the defect of weak generalization. At the same time, the existing methods do not explicitly use the angle and depth information of the anatomical structure, making it impossible to distinguish the subdivided sections with similar structures, and the classification granularity is insufficient. Finally, the echocardiography quality control based on the existing target detection scheme does not make full use of the anatomical structure angle information, and the effect is limited. Therefore, there is an urgent need for an echocardiography classification method that can generalize well and support fine-grained classification of the whole heart section. Summary of the invention

[0004] Based on this, it is necessary to provide a section classification method, device, computer equipment, computer-readable storage medium and computer program product for cardiac ultrasound images that can generalize well and support fine-grained classification of all heart sections in order to address the above technical problems.

[0005] In a first aspect, the present application provides a method for classifying sections of cardiac ultrasound images, comprising:

[0006] Extract features from cardiac ultrasound images to obtain image features;

[0007] Perform semantic segmentation and position detection based on the image features to obtain a semantic segmentation map and corner point position information, wherein the semantic segmentation map includes a plurality of pixels and a tissue to which each pixel belongs, and the corner point position information represents the position of the heart in the cardiac ultrasound image, wherein the heart includes a plurality of tissues;

[0008] Based on the corner point position information, converting the initial position information of each pixel point in the semantic segmentation map into polar coordinate information;

[0009] Based on the polar coordinate information of each pixel point and the tissue to which each pixel point belongs, the section type of the cardiac ultrasound image is determined.

[0010] In one embodiment, determining the section type of the cardiac ultrasound image based on the polar coordinate information of each pixel and the tissue to which each pixel belongs includes:

[0011] The semantic segmentation map after polar coordinate conversion is input into a feature extractor, and the feature extractor is used to extract features from the semantic segmentation map after polar coordinate conversion to obtain a target feature representation; the target feature representation is input into a section classifier, and the section classifier is used to determine the section type of the cardiac ultrasound image based on the target feature representation.

[0012] In one embodiment, the feature extractor extracts features from the semantic segmentation map after polar coordinate conversion to obtain a target feature representation, including:

[0013] Through the feature extractor, the semantic segmentation map after polar coordinate conversion is divided into multiple local images of the same size; for each local image, the local image is converted into an initial feature representation of a preset dimension; based on the polar coordinate information of the pixel points in the local image, the self-attention weight corresponding to the local image is determined, and the initial feature representation and the self-attention weight are weighted to obtain the target feature representation.

[0014] In one embodiment, determining the section type of the cardiac ultrasound image based on the target feature representation by the section classifier includes:

[0015] The target feature representation is flattened and straightened to obtain a one-dimensional feature representation; for each section type, a weight value and a bias value corresponding to the section type are obtained; the one-dimensional feature representation and the weight value are weighted to obtain a linear combination feature representation; the linear combination feature representation and the bias value are added to obtain an activation value corresponding to the section type; according to the activation value corresponding to each section type, a section type probability distribution feature vector is obtained; according to the section type probability distribution feature vector, the section type of the cardiac ultrasound image is determined.

[0016] In one embodiment, determining the section type of the cardiac ultrasound image according to the section type probability distribution feature vector includes:

[0017] The activation value corresponding to each section type is converted into the probability that the cardiac ultrasound image is classified into each section type; and the section type of the cardiac ultrasound image is obtained according to the probability that the cardiac ultrasound image is classified into each section type.

[0018] In one of the embodiments, the target feature representation is input into a quality regressor, and through the quality regressor, a detection frame is generated in the cardiac ultrasound image based on the target feature representation, and the detection frame is used to characterize the positions of different tissue structures in the heart; a section quality evaluation rule corresponding to the section type of the cardiac ultrasound image is obtained; and the section quality of the cardiac ultrasound image is determined based on the section quality evaluation rule and the detection frame generated in the cardiac ultrasound image.

[0019] In one of the embodiments, the target feature representation is input into a quality assessor, and the quality assessor generates a defect language text based on the target feature representation, wherein the defect language text is used to describe the section defects of the cardiac ultrasound image.

[0020] In a second aspect, the present application further provides a slice classification device for cardiac ultrasound images, comprising:

[0021] An extraction module is used to extract features from cardiac ultrasound images to obtain image features;

[0022] an acquisition module, configured to perform semantic segmentation and position detection based on the image features to obtain a semantic segmentation map and corner point position information, wherein the semantic segmentation map includes a plurality of pixels and a tissue to which each pixel belongs, and the corner point position information represents a position of a heart in the cardiac ultrasound image, wherein the heart includes a plurality of tissues;

[0023] A conversion module, used for converting the initial position information of each pixel point in the semantic segmentation map into polar coordinate information based on the corner point position information;

[0024] A determination module is used to determine the section type of the cardiac ultrasound image based on the polar coordinate information of each pixel point and the tissue to which each pixel point belongs.

[0025] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0026] Extract features from cardiac ultrasound images to obtain image features;

[0027] Perform semantic segmentation and position detection based on the image features to obtain a semantic segmentation map and corner point position information, wherein the semantic segmentation map includes a plurality of pixels and a tissue to which each pixel belongs, and the corner point position information represents the position of the heart in the cardiac ultrasound image, wherein the heart includes a plurality of tissues;

[0028] Based on the corner point position information, converting the initial position information of each pixel point in the semantic segmentation map into polar coordinate information;

[0029] Based on the polar coordinate information of each pixel point and the tissue to which each pixel point belongs, the section type of the cardiac ultrasound image is determined.

[0030] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0031] Extract features from cardiac ultrasound images to obtain image features;

[0032] Perform semantic segmentation and position detection based on the image features to obtain a semantic segmentation map and corner point position information, wherein the semantic segmentation map includes a plurality of pixels and a tissue to which each pixel belongs, and the corner point position information represents the position of the heart in the cardiac ultrasound image, wherein the heart includes a plurality of tissues;

[0033] Based on the corner point position information, converting the initial position information of each pixel point in the semantic segmentation map into polar coordinate information;

[0034] Based on the polar coordinate information of each pixel point and the tissue to which each pixel point belongs, the section type of the cardiac ultrasound image is determined.

[0035] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0036] Extract features from cardiac ultrasound images to obtain image features;

[0037] Perform semantic segmentation and position detection based on the image features to obtain a semantic segmentation map and corner point position information, wherein the semantic segmentation map includes a plurality of pixels and a tissue to which each pixel belongs, and the corner point position information represents the position of the heart in the cardiac ultrasound image, wherein the heart includes a plurality of tissues;

[0038] Based on the corner point position information, converting the initial position information of each pixel point in the semantic segmentation map into polar coordinate information;

[0039] Based on the polar coordinate information of each pixel point and the tissue to which each pixel point belongs, the section type of the cardiac ultrasound image is determined.

[0040] The above-mentioned section classification method, device, computer device, computer-readable storage medium and computer program product for cardiac ultrasound images obtain image features by extracting features from cardiac ultrasound images, and effectively extract the features of the target anatomical structure from the cardiac ultrasound images. This not only helps to improve the segmentation accuracy, but also increases the generalization ability of the model to different devices and imaging conditions, thereby reducing the impact caused by device differences.

[0041] Based on the image features, semantic segmentation and position detection are performed to obtain a semantic segmentation map and corner point position information, wherein the semantic segmentation map includes multiple pixels and the tissue to which each pixel belongs, and the corner point position information represents the position of the heart in the cardiac ultrasound image, wherein the heart includes multiple tissues; a detailed anatomical structure label is generated, and the position of the heart in the image is accurately located. It solves the generalization problem that is prone to occur when traditional methods rely only on the original image for classification, and ensures that the algorithm can focus on the target structure itself rather than background noise or the imaging style of a specific machine. At the same time, the detection of the fan-shaped corner points provides the necessary geometric reference points for the subsequent conversion to the polar coordinate system.

[0042] Based on the corner point position information, the initial position information of each pixel in the semantic segmentation map is converted into polar coordinate information, so that the angle and depth information of the anatomical structure can be explicitly utilized. This is crucial for distinguishing fine-grained sections with similar anatomical structures but different angles, greatly improving the granularity and accuracy of classification. In addition, the polar coordinate representation helps to better capture the spatial relationship of the internal structure of the heart, further enhancing the expressiveness of the model.

[0043] Based on the polar coordinate information of each pixel and the tissue to which each pixel belongs, the section type of the cardiac ultrasound image is determined. Polar coordinate data containing depth and angle information is used to achieve high-precision classification of cardiac ultrasound section types. Compared with the existing technology, this method can identify various standard and derived cardiac sections at a finer granularity, greatly simplifying the workflow of ultrasound doctors and improving the quality control level of automated recording.

[0044] Through the above steps, the introduction of anatomical semantic segmentation and polar coordinate transformation effectively solves the problems of weak generalization, insufficient classification granularity and limited quality control effect in the current echocardiography classification, providing strong technical support for realizing more intelligent and efficient cardiac ultrasound examinations. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 is an application environment diagram of a section classification method for cardiac ultrasound images in one embodiment;

[0047] Figure 2 is a schematic flow chart of a method for classifying sections of cardiac ultrasound images in one embodiment;

[0048] Figure 3 A schematic diagram of polar coordinate conversion for a semantic segmentation map in one embodiment;

[0049] Figure 4 A feature extraction network diagram for one embodiment;

[0050] Figure 5 is a flow chart of a method for classifying sections of cardiac ultrasound images in another embodiment;

[0051] Figure 6 is a structural block diagram of a slice classification device for cardiac ultrasound images in one embodiment;

[0052] Figure 7 is a structural block diagram of a slice classification device for cardiac ultrasound images in another embodiment;

[0053] Figure 8 A standard left ventricular long axis section and a derived section with a deeper left ventricular long axis section in one embodiment;

[0054] Fig. 9 The standard section of the apical four-chamber heart (centered on the left ventricle) and the derived section of the apical four-chamber heart (centered on the right ventricle) in one embodiment are shown;

[0055] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0057] The section classification method for cardiac ultrasound images provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The terminal 102 is used to generate a section classification request for the cardiac ultrasound image, and send the section classification request for the cardiac ultrasound image to the server 104, so that the server 104 determines the section type of the cardiac ultrasound image based on the polar coordinate information of each pixel point and the organization to which each pixel point belongs. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car devices, projection devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server 104 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.

[0058] In an exemplary embodiment, Figure 2 As shown in FIG. 1 , a cross-sectional classification method for cardiac ultrasound images is provided, and the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate, including the following steps 202 to 208. Among them:

[0059] Step 202: extract features from the cardiac ultrasound image to obtain image features.

[0060] Among them, image features refer to information extracted from the original cardiac ultrasound image that can characterize the image content. This information is used in subsequent processing steps such as semantic segmentation, position detection, etc. It can be understood that in this application, image features are multi-scale features, and each scale corresponds to a level. For cardiac ultrasound images, "image features" particularly emphasize those characteristics related to the anatomical structure of the heart, such as the location, size, and morphology of the ventricles, atria, valves, and other blood vessels. These features are crucial for accurate semantic segmentation (i.e., distinguishing different cardiac tissues).

[0061] Cardiac ultrasound images refer to two-dimensional grayscale images of the internal structure of the heart obtained by an echocardiography device. These images reflect different sections of the heart, such as the long axis of the left ventricle, the apical four-chamber heart, etc., and can be static images or a frame in a dynamic video. Optionally, an ultrasound probe can be used to scan the heart from different angles and positions to obtain cardiac ultrasound images.

[0062] Specifically, the cardiac ultrasound image is input into the feature extraction network. It can be understood that the feature extraction network in the present application includes at least an encoder, a decoder, and an output head. First, in the encoder stage, the encoder part gradually reduces the spatial size (width and height) of the image through a series of convolutional layers and pooling layers, while increasing the number of feature channels. Each layer captures a higher level of feature representation. And in the encoder stage, the feature map of each level is also saved for use in the subsequent decoder stage.

[0063] Then, in the decoder stage, deconvolution (upsampling) is used to restore the spatial size of the image and gradually reconstruct the high-resolution segmentation map. In this process, the feature maps from the corresponding level of the encoder are merged with the output of the current decoder level to ensure that the position information and details are preserved. It can be understood that the output of the encoder and decoder at each level is merged into the image feature. Therefore, the image feature is a multi-scale feature to ensure that the position information and details are preserved.

[0064] Step 204, performing semantic segmentation and position detection based on image features to obtain a semantic segmentation map and corner point position information, wherein the semantic segmentation map includes multiple pixels and the tissue to which each pixel belongs, and the corner point position information represents the position of the heart in the cardiac ultrasound image, and the heart includes multiple tissues.

[0065] The semantic segmentation map is to classify each pixel in the original cardiac ultrasound image and assign a label to indicate which specific target structure (such as ventricle, atrium, etc.) or background the pixel belongs to. It can be understood that the semantic segmentation map is used to clearly mark the specific location and range of each anatomical structure in the image, providing a basis for subsequent polar coordinate conversion.

[0066] Optionally, a semantic segmentation map is an image of the same size as the original image, but in which the color or intensity values ​​no longer represent the true tissue density or blood flow velocity, but instead represent the categories of different anatomical structures. For example, all parts belonging to the left ventricle may be red, the right ventricle is blue, the mitral valve is green, and so on; background or other non-target areas are usually represented by black or other specified colors.

[0067] Corner point location information refers to the location information of three key points within the fan area - two endpoints and the vertex (i.e., the center of the fan). These three points can uniquely determine the shape and orientation of the fan, which helps to describe the geometric characteristics in cardiac ultrasound images. Understandably, corner point location information is crucial for subsequent polar coordinate transformation because it provides the necessary geometric reference points, enabling accurate calculation of the angle and distance of any given point relative to the vertex of the fan, which is very important for understanding the angle and depth of anatomical structures.

[0068] Tissue refers to the different anatomical structures or regions inside the heart, such as ventricles, atria, valves, large blood vessels, etc. These structures make up the main components of the heart. In the semantic segmentation map, each pixel is assigned a label that indicates the specific anatomical structure (i.e. tissue) to which the pixel belongs. For example, all pixels belonging to the left ventricle will be labeled as one class, all pixels belonging to the right ventricle will be labeled as another class, and so on.

[0069] Specifically, the feature extraction network includes two output heads, one of which ultimately generates a semantic segmentation map in which each pixel is assigned a label indicating the anatomical structure or background to which it belongs. The other output head is specifically used to predict the positions of the three corner points within the fan-shaped area. Optionally, the coordinates of the three points, i.e., the corner point position information, are inferred directly from the image features through a fully connected network.

[0070] Step 206: based on the corner point position information, convert the initial position information of each pixel in the semantic segmentation map into polar coordinate information.

[0071] The initial position information refers to the position of each pixel in the original cardiac ultrasound image without any transformation processing.

[0072] Polar coordinate information refers to the representation of each pixel after converting from a plane rectangular coordinate system to a polar coordinate system. The polar coordinate system uses two parameters to define a point: the polar coordinate r (also called the radial distance) and the polar angle θ (angle). In this application, the polar coordinate r represents the distance from the vertex of the fan (usually the probe position) to a certain point, reflecting the "depth" of the cardiac structure at the point relative to the probe. The polar angle θ represents the angle formed by the line segment from the reference direction (such as the positive x-axis) to the origin and the point, and is used to describe the directionality of the anatomical structure.

[0073] Specifically, refer to Figure 3 First, determine the origin and reference direction of the polar coordinate system, including: taking the vertex of the three corner points (including the polar coordinate origin, the left pole and the right pole) as the origin of the polar coordinate system (i.e., the polar coordinate origin), i.e., the position of the ultrasound probe; selecting a reference direction, usually the direction from the vertex to the horizontal right can be selected as the 0-degree reference line. This step is to ensure the consistency and accuracy of all angle measurements. Understandably, the corner point position information refers to the position information of the polar coordinate origin, the left pole and the right pole.

[0074] Then the polar radius r and polar angle θ of each pixel are calculated, including: for each pixel, the straight-line distance from it to the origin (i.e., vertex) of the polar coordinate system is calculated as the polar radius r. This distance reflects the depth of the structure relative to the probe. Then the angle formed by the line segment from the reference direction to the origin and the pixel is calculated as the polar angle θ. This angle describes the directionality of the anatomical structure.

[0075] Finally, a new representation is created for each pixel, which contains the polar radius r and polar angle θ. Optionally, this can be achieved by adding additional channels or metadata to the original semantic segmentation map, or by directly saving this information in a new image format so that these spatial features can be further utilized in subsequent processing steps.

[0076] Step 208: Determine the section type of the cardiac ultrasound image based on the polar coordinate information of each pixel point and the tissue to which each pixel point belongs.

[0077] Among them, the slice type refers to the specific cardiac anatomical view or perspective represented by the cardiac ultrasound image. During cardiac ultrasound examination, different views of the cardiac structure are obtained by adjusting the position and angle of the ultrasound probe. These specific angle and position combinations are called "slices." Understandably, each slice type provides a unique way to observe different parts of the heart and its functions. For example, cardiac ultrasound examinations can involve up to 27 two-dimensional slices, and an additional 55 Doppler slices can be derived from these basic slices, for a total of 72 slices. Each slice has its own specific standards, such as the long axis of the left ventricle, the apical four-chamber heart, etc., which are used to evaluate specific cardiac structures such as valves, myocardial wall motion, blood flow direction, etc. Reference Figure 7 , showing a standard view obtained by scanning the heart at a specific angle with an ultrasound probe, showing the long axis direction of the left ventricle, including structures such as the aortic valve, left ventricular wall, mitral valve, and part of the right ventricular outflow tract. This view is a basic and commonly used standard section in echocardiography, used to evaluate the function and structure of the heart. Figure 8 Based on the standard left ventricular long axis section, the position or parameters of the ultrasound probe are adjusted to allow the scan to go deeper into the heart, thereby obtaining an image with a deeper field of view. It can provide more information about the state of the deep structures of the heart, such as better display of the top area of ​​the left ventricle or more clear view of certain structures located behind the heart. Such a derivative section may be necessary for the diagnosis of certain diseases because it can provide additional information to help doctors make more accurate judgments.

[0078] Specifically, the cardiac ultrasound semantic segmentation map after polar coordinate conversion is input into the cardiac section prediction network. The cardiac section prediction network includes at least a feature extractor, a section classifier, a quality regressor, and a quality evaluator. The feature extractor processes the cardiac ultrasound semantic segmentation map after polar coordinate conversion, and outputs a feature representation rich in contextual information (i.e., a target feature representation). It can be understood that the target feature representation not only contains the spatial information of the original image, but also incorporates depth and angle information. The target feature representation output by the feature extractor is processed by the section classifier to output the section type of the cardiac ultrasound image.

[0079] In one embodiment, the target feature representation output by the feature extractor is processed by a quality regressor to output the quality of the section of the cardiac ultrasound image. The target feature representation output by the feature extractor is processed by a quality assessor to output a quality text description of the quality of the section of the cardiac ultrasound image.

[0080] In one of the embodiments, the semantic segmentation map after polar coordinate conversion is input into a feature extractor, and the feature extractor performs feature extraction on the semantic segmentation map after polar coordinate conversion to obtain a target feature representation; the target feature representation is input into a section classifier, and the section classifier determines the section type of the cardiac ultrasound image based on the target feature representation.

[0081] The feature extractor is a part of the cardiac slice prediction network, which is used to automatically extract useful features from the input data. In this specific application scenario, the feature extractor is designed to process the cardiac ultrasound semantic segmentation map after polar coordinate transformation.

[0082] The section classifier is a module specifically used to identify and classify which standard section or derived section the cardiac ultrasound image belongs to. It can be understood that based on the feature representation provided by the feature extractor, the section classifier performs a multi-class classification task to determine which specific standard section or more fine-grained derived section the input cardiac ultrasound image belongs to.

[0083] The target feature representation refers to a feature vector or feature matrix that is rich in contextual information. It integrates the anatomical structure information in the original cardiac ultrasound image as well as their angle and depth information. It is a high-level abstract representation that can effectively support subsequent tasks such as section classification. These feature representations are crucial for achieving fine-grained cardiac ultrasound section classification because they contain the necessary details for distinguishing different section types.

[0084] Specifically, the input data is first prepared, including: ensuring that each pixel in the semantic segmentation map after polar coordinate conversion contains not only its initial position information (plane rectangular coordinates), but also the polar radius r and polar angle θ obtained by calculation. And each pixel has been assigned a label, indicating the specific anatomical structure to which the pixel belongs. Then the feature extractor receives the semantic segmentation map after polar coordinate conversion as input. The feature extractor processes the cardiac ultrasound semantic segmentation map after polar coordinate conversion, and outputs a feature representation rich in contextual information (i.e., target feature representation). And the output target feature representation is sent to the section classifier. Finally, the section classifier receives the target feature representation sent by the feature extractor. The target feature representation is processed by the section classifier to output the section type of the cardiac ultrasound image.

[0085] Since the anatomical semantic segmentation map is used instead of the original image as input, the difference factors caused by the imaging texture and background noise of different devices are eliminated, so that the algorithm can better adapt to ultrasound devices of different brands and models, and enhance the generalization ability of the system. And through polar coordinate transformation and semantic segmentation, the model can more accurately capture the angle and depth information of the anatomical structure. This helps to solve the problem that traditional methods have difficulty in distinguishing fine-grained section types, thereby improving the accuracy of classification.

[0086] In one of the embodiments, a feature extractor is used to segment a semantic segmentation map after polar coordinate conversion into multiple local images of the same size; for each local image, the local image is converted into an initial feature representation of a preset dimension; based on the polar coordinate information of the pixel points in the local image, the self-attention weight corresponding to the local image is determined, and the initial feature representation and the self-attention weight are weighted to obtain a target feature representation.

[0087] Among them, the local image means that when processing the semantic segmentation map after polar coordinate transformation, in order to facilitate the model to process and learn more fine-grained spatial information, the image is divided into multiple small blocks of the same size, and each such small block is called a "local image".

[0088] The initial feature representation of a preset dimension refers to converting each local image into a vector form of a fixed dimension (preset dimension), and this vector is the "initial feature representation" of the local image. Optionally, each local image is mapped to a low-dimensional space through a specific encoding method (such as a convolutional layer or a linear transformation). For example, if the original local image is h×w in size, it can be converted into a vector of length d, where d is the preset dimension. It can be understood that despite the dimensionality reduction process, these initial feature representations still retain the key features of the local image, such as color, texture, etc., and can reflect the relative position relationship between pixels.

[0089] Self-attention weight refers to the self-attention weight matrix, which is used to measure the importance between different pixels or local images. It allows the model to dynamically adjust the influence of each part on the final output. Understandably, the importance of each pixel in the local image is calculated using the polar radius r and polar angle θ information of each pixel. The directionality and depth information of the anatomical structure are taken into account to ensure that the model can focus on key areas.

[0090] Specifically, the semantic segmentation map after polar coordinate conversion is first input into the feature extractor, and the semantic segmentation map is divided into multiple small blocks of the same size (i.e., local images) by the feature extractor. Each small block represents a part of the original image, which helps the model capture local features more carefully. Optionally, the segmentation method can be uniform division or non-uniform division based on the key areas of the anatomical structure to ensure that important structures are not over-segmented.

[0091] Then, for each local image, a specific encoding method (such as a convolutional layer or a linear transformation) is used to map it to a vector space of a fixed dimension (preset dimension) to obtain the initial feature representation. It can be understood that the choice of preset dimension depends on the model design and computing resource constraints to ensure that the local image features can be fully expressed while maintaining computational efficiency.

[0092] Next, the polar radius r and polar angle θ of each pixel in the local image are used to calculate the self-attention weight. The polar coordinate information helps the model understand the specific position and direction of each pixel in the heart structure. Different pixels correspond to different self-attention weight matrices to determine the importance of each pixel in each local image.

[0093] Finally, for each local image, its initial feature representation is multiplied by the corresponding self-attention weight matrix for weighted processing. This means that those pixels or local images that are considered more important will have a greater influence in subsequent tasks. The weighted feature representation is the target feature representation, which more accurately reflects the important structural information in the local image while retaining the global contextual relationship.

[0094] Since the semantic segmentation map after polar coordinate conversion is divided into multiple local images of the same size, each small block can be analyzed and processed independently. This method can capture more detailed anatomical information and improve the understanding of different areas of the heart. In addition, the introduction of self-attention weights and weighting them with the initial feature representation allows the model to dynamically adjust the influence of each part on the final output. This helps to highlight key anatomical structures and reduce the impact of irrelevant or redundant information, thereby improving classification accuracy.

[0095] In one of the embodiments, the target feature representation is flattened and straightened to obtain a one-dimensional feature representation; for each section type, a weight value and a bias value corresponding to the section type are obtained; the one-dimensional feature representation and the weight value are weighted to obtain a linear combination feature representation; the linear combination feature representation and the bias value are added to obtain an activation value corresponding to the section type; based on the activation value corresponding to each section type, a section type probability distribution feature vector is obtained; based on the section type probability distribution feature vector, the section type of the cardiac ultrasound image is determined.

[0096] Among them, the one-dimensional feature representation refers to the one-dimensional vector obtained by flattening and straightening the target feature representation (usually multi-dimensional). It can be understood that despite the dimensionality conversion, all the information in the original feature representation is still retained to ensure that important anatomical structure features are not lost. The weight value refers to a set of numerical parameters corresponding to each section type, which is used to measure the importance of different features. The bias value refers to an additional parameter added to the linear combination feature representation to adjust the position of the output result. The linear combination feature representation refers to the new feature representation obtained by weighted processing of the one-dimensional feature representation and the weight value. The activation value refers to the result of adding the bias value to the linear combination feature representation. The section type probability distribution feature vector refers to the probability distribution obtained by applying the activation function to the activation values ​​of all section types.

[0097] Specifically, the target feature representation is first input into the section classifier, which flattens the multi-dimensional target feature representation into a one-dimensional vector, i.e., a one-dimensional feature representation. This step simplifies the input format, making subsequent linear transformations and calculations easier to implement. It ensures that important anatomical structure information is not lost during the flattening process. For each possible section type, a set of corresponding weight values ​​W is pre-trained and stored. i and the bias value b i , where i represents different aspect types.

[0098] Next, for each slice type i, the linear combination feature representation Z is calculated i =X×W i , by calculating the linear combination feature representation corresponding to each section type to measure the contribution of each feature to a specific section type in the one-dimensional feature representation. Where X is the one-dimensional feature representation, Z i The linear combination feature representation corresponding to each section type. For each section type i, calculate the activation value ai=Z i +b i This step comprehensively considers the impact of all features and their weights, as well as the role of bias values, to form a preliminary score or rating for a specific aspect type.

[0099] Finally, the activation function is applied to convert all activation values ​​into a normalized probability distribution pi= / Where j is the possible slice type. This step generates a probability distribution that reflects the relative likelihood that the input image belongs to each slice type. Finally, the slice type probability distribution feature vector p = [p 1 , p 2 ,…,p n ], where n is the total number of aspect types. Select the aspect type with the highest probability as the final classification result. That is, find the maximum value p in p max , and its corresponding index i is the predicted section type of the cardiac ultrasound image. Optionally, if the maximum probability value is lower than a certain threshold, it can be marked as uncertain or requires further examination.

[0100] Since the one-dimensional feature representation is weighted with the weight value and the activation value is calculated in combination with the bias value, this approach enables the model to flexibly respond to different types of cardiac ultrasound images and maintain good classification performance even for data collected with new equipment or under different conditions. The uncertainty of the classification results can be evaluated through probability distribution. If the maximum probability value is low, it indicates that there may be ambiguity or abnormality, which requires further inspection or confirmation, enhancing the robustness and reliability of the system.

[0101] In one of the embodiments, the activation value corresponding to each section type is converted into the probability of the cardiac ultrasound image being classified into each section type; based on the probability of the cardiac ultrasound image being classified into each section type, the section type of the cardiac ultrasound image is obtained.

[0102] Specifically, the activation function is applied to convert all activation values ​​into a normalized probability distribution pi = / Where j is the possible slice type. This step generates a probability distribution that reflects the relative likelihood that the input image belongs to each slice type. Finally, the slice type probability distribution feature vector p = [p 1 , p 2 ,…,p n ], where n is the total number of aspect types. Select the aspect type with the highest probability as the final classification result. That is, find the maximum value p in p max , and its corresponding index i is the predicted section type of the cardiac ultrasound image. Optionally, if the maximum probability value is lower than a certain threshold, it can be marked as uncertain or requires further examination.

[0103] By converting the activation value into a probability distribution, not only the most likely aspect type is given, but also other possibilities and their probability sizes are provided. This not only improves the accuracy of classification, but also provides a quantitative assessment of the reliability of the classification results.

[0104] In one of the embodiments, a target feature representation is input into a quality regressor, and a detection frame is generated in a cardiac ultrasound image based on the target feature representation through the quality regressor, wherein the detection frame is used to characterize the positions of different tissue structures in the heart; a section quality evaluation rule corresponding to the section type of the cardiac ultrasound image is obtained; and the section quality of the cardiac ultrasound image is determined based on the section quality evaluation rule and the detection frame generated in the cardiac ultrasound image.

[0105] The quality regressor is a module responsible for evaluating the quality of cardiac ultrasound images and giving a quantitative evaluation. By processing the target feature representation output by the feature extractor, a continuous value is estimated as a scoring indicator to reflect the quality of the current section. Optionally, a numerical value is output to represent a quantitative evaluation of the quality of the current section (e.g., a score between 1 and 5).

[0106] A detection box is a rectangular box that marks the location of a specific anatomical structure in a cardiac ultrasound image. It can be understood that each detection box corresponds to a specific tissue structure in the heart, such as a ventricle, atrium, valve, etc. They specify the specific location of these structures in the image.

[0107] Slice quality evaluation rules refer to a series of standards or criteria developed for each possible cardiac ultrasound slice type to measure the quality of that type of image. Understandably, specific scoring criteria are set for different slice types, such as image clarity, visibility of anatomical structures, and other factors. These rules guide how to assess image quality based on detection boxes and other features.

[0108] Section quality refers to the overall quality level of cardiac ultrasound images after considering multiple factors. It is expressed as one or more numerical values, which intuitively reflect the quality level of the image. High-quality images mean clearer and more complete display of anatomical structures, which is conducive to accurate medical diagnosis.

[0109] Specifically, the target feature representation is first input into the mass regressor, and the mass regressor generates multiple detection boxes in the cardiac ultrasound image based on the target feature representation. Each detection box corresponds to a specific cardiac anatomical structure (such as ventricle, atrium, valve, etc.) and clearly indicates the specific location of the structure in the image.

[0110] Then, according to the slice type previously determined by the slice classifier, the slice quality evaluation rule corresponding to the slice type is obtained from the preset rule library. It can be understood that the slice quality evaluation rule defines the scoring criteria for the slice type, such as factors such as image clarity and anatomical structure visibility.

[0111] Finally, the quality regressor comprehensively considers various factors, such as image clarity, anatomical integrity, and position accuracy, by combining the slice quality evaluation rules and the anatomical structure information provided by the detection frame. One or more continuous values ​​are calculated as scoring indicators to reflect the slice quality of the cardiac ultrasound image. Optionally, the slice quality is expressed as a score within a numerical range (e.g., a score between 1 and 5). If the score is lower than a preset threshold, the image quality can be marked as low, indicating that further inspection or re-acquisition is required.

[0112] By obtaining the slice quality evaluation rules corresponding to the slice type of the cardiac ultrasound image, it is ensured that different doctors or systems use the same standards when evaluating the same type of images. This reduces subjective differences and enhances the consistency and reliability of the evaluation results. And the evaluation based on the target feature representation not only considers the presence or absence of anatomical structures, but also incorporates depth and angle information, so that even slices with similar structures but slightly different angles or positions can be effectively distinguished.

[0113] In one of the embodiments, the target feature representation is input into a quality assessor, and the quality assessor generates defect language text based on the target feature representation, and the defect language text is used to describe the section defects of the cardiac ultrasound image.

[0114] The quality assessor refers to a module for describing the section defects of cardiac ultrasound images, which receives the target feature representation from the feature extractor as input, and outputs the defect description of the cardiac ultrasound section, i.e., the defect language text, through a series of processing steps.

[0115] Defect language text refers to a text description generated by the quality evaluator based on the target feature representation, which is used to point out the problems or areas for improvement in the current cardiac ultrasound section. Specifically, the cardiac section prediction network also includes a quality evaluator, which is used to generate a text description of the current section defect. Optionally, the target feature representation is input into the quality evaluator, and the quality evaluator uses a trained section defect language model, which uses an encoding-decoding architecture or a conditional generative adversarial network. The section defect language model generates descriptive text about the current section defect based on the information in the target feature representation. For example, output: a text description pointing out the problems or areas for improvement in the current section, such as "the mitral valve is not clearly displayed" and "the apex position is offset."

[0116] In one embodiment, reference Figure 4 , describing the processing of the feature extraction network.

[0117] In the encoder part: the input image size is 572x572, and it is subjected to a 3x3 convolution layer for feature extraction. It is downsampled through a max pooling layer and the size becomes 282x282. It continues to be downsampled through convolutional layers and max pooling layers, and the sizes become 140x140, 68x68, and 32x32 respectively.

[0118] In the bottleneck layer part: In the deepest downsampling layer, feature extraction is performed through a convolutional layer with a size of 32x32.

[0119] In the decoder part: upsample through the upsampling layer to a size of 64x64. Concatenate the upsampled feature map with the feature map of the corresponding layer of the encoder, and then perform feature fusion through the convolution layer. Continue upsampling through the upsampling layer and feature fusion, and the size becomes 128x128, 256x256, and 512x512 in turn.

[0120] In the output layer, the final upsampling is performed through the upsampling layer, and the size becomes 388x388. The final segmentation map is generated through a 1x1 convolution layer, and the size is 388x388. Through the above architecture, the feature extraction network can effectively perform image segmentation tasks while maintaining high resolution.

[0121] In an exemplary embodiment, Figure 5 As shown, it includes steps 502 to 510. Among them:

[0122] Step 502, extracting features from the cardiac ultrasound image to obtain image features;

[0123] Step 504, performing semantic segmentation and position detection based on the image features to obtain a semantic segmentation map and corner point position information, wherein the semantic segmentation map includes a plurality of pixels and a tissue to which each pixel belongs, and the corner point position information represents the position of the heart in the cardiac ultrasound image, and the heart includes a plurality of tissues;

[0124] Step 506, based on the corner point position information, convert the initial position information of each pixel in the semantic segmentation map into polar coordinate information;

[0125] Step 508: Input the semantic segmentation map after polar coordinate conversion into a feature extractor, and divide the semantic segmentation map after polar coordinate conversion into multiple local images of the same size through the feature extractor; for each local image, convert the local image into an initial feature representation of a preset dimension; based on the polar coordinate information of the pixel points in the local image, determine the self-attention weight corresponding to the local image, and perform weighted processing on the initial feature representation and the self-attention weight to obtain the target feature representation; input the target feature representation into the section classifier, and flatten and straighten the target feature representation through the section classifier. to obtain a one-dimensional feature representation; for each section type, obtain a weight value and a bias value corresponding to the section type; perform weighted processing on the one-dimensional feature representation and the weight value to obtain a linear combination feature representation; add the linear combination feature representation and the bias value to obtain an activation value corresponding to the section type; according to the activation value corresponding to each section type, obtain a section type probability distribution feature vector; convert the activation value corresponding to each section type into the probability that the cardiac ultrasound image is classified into each section type; according to the probability that the cardiac ultrasound image is classified into each section type, obtain the section type of the cardiac ultrasound image;

[0126] Step 510, input the target feature representation to the quality regressor, and generate a detection frame in the cardiac ultrasound image based on the target feature representation through the quality regressor, the detection frame is used to characterize the location of different tissue structures in the heart; obtain the section quality evaluation rule corresponding to the section type of the cardiac ultrasound image; determine the section quality of the cardiac ultrasound image according to the section quality evaluation rule and the detection frame generated in the cardiac ultrasound image; input the target feature representation to the quality evaluator, and generate a defect language text based on the target feature representation through the quality evaluator, the defect language text is used to describe the section defects of the cardiac ultrasound image.

[0127] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0128] Based on the same inventive concept, the embodiment of the present application also provides a device for classifying sections of cardiac ultrasound images for implementing the above-mentioned method for classifying sections of cardiac ultrasound images. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the device for classifying sections of cardiac ultrasound images provided below can refer to the limitations of the method for classifying sections of cardiac ultrasound images above, and will not be repeated here.

[0129] In an exemplary embodiment, Figure 6 As shown, a cross-section classification device 600 for cardiac ultrasound images is provided, comprising: an extraction module 602, an acquisition module 604, a conversion module 606 and a determination module 608, wherein:

[0130] An extraction module 602 is used to extract features from cardiac ultrasound images to obtain image features;

[0131] An acquisition module 604 is used to perform semantic segmentation and position detection based on image features to obtain a semantic segmentation map and corner point position information, wherein the semantic segmentation map includes multiple pixels and the tissue to which each pixel belongs, and the corner point position information represents the position of the heart in the cardiac ultrasound image, and the heart includes multiple tissues;

[0132] A conversion module 606, configured to convert the initial position information of each pixel point in the semantic segmentation map into polar coordinate information based on the corner point position information;

[0133] The determination module 608 is used to determine the section type of the cardiac ultrasound image based on the polar coordinate information of each pixel point and the tissue to which each pixel point belongs.

[0134] In one embodiment, the determination module 608 is used to input the semantic segmentation map after polar coordinate conversion into a feature extractor, and perform feature extraction on the semantic segmentation map after polar coordinate conversion through the feature extractor to obtain a target feature representation; the target feature representation is input into a section classifier, and the section classifier determines the section type of the cardiac ultrasound image based on the target feature representation.

[0135] In one embodiment, the determination module 608 is used to divide the semantic segmentation map after polar coordinate conversion into multiple local images of the same size through a feature extractor; for each local image, convert the local image into an initial feature representation of a preset dimension; based on the polar coordinate information of the pixel points in the local image, determine the self-attention weight corresponding to the local image, and weight the initial feature representation and the self-attention weight to obtain the target feature representation.

[0136] In one embodiment, a determination module 608 is used to flatten and straighten the target feature representation to obtain a one-dimensional feature representation; for each section type, a weight value and a bias value corresponding to the section type are obtained; the one-dimensional feature representation and the weight value are weighted to obtain a linear combination feature representation; the linear combination feature representation and the bias value are added to obtain an activation value corresponding to the section type; based on the activation value corresponding to each section type, a section type probability distribution feature vector is obtained; and based on the section type probability distribution feature vector, the section type of the cardiac ultrasound image is determined.

[0137] In one embodiment, the determination module 608 is used to convert the activation value corresponding to each section type into the probability that the cardiac ultrasound image is classified into each section type; and obtain the section type of the cardiac ultrasound image according to the probability that the cardiac ultrasound image is classified into each section type.

[0138] In one of the embodiments, the slice classification device for cardiac ultrasound images also includes a quality module 610, which is used to input the target feature representation into a quality regressor, and generate a detection frame in the cardiac ultrasound image based on the target feature representation through the quality regressor, and the detection frame is used to characterize the positions of different tissue structures in the heart; obtain a slice quality evaluation rule corresponding to the slice type of the cardiac ultrasound image; and determine the slice quality of the cardiac ultrasound image according to the slice quality evaluation rule and the detection frame generated in the cardiac ultrasound image.

[0139] In one of the embodiments, the section classification device for cardiac ultrasound images also includes an evaluation module 612, which is used to input the target feature representation into the quality evaluator, and generate a defect language text based on the target feature representation through the quality evaluator, and the defect language text is used to describe the section defects of the cardiac ultrasound image.

[0140] In another embodiment, if Figure 7 As shown, Figure 7The structural block diagram of the section classification device for cardiac ultrasound images in another embodiment includes: an extraction module 602, an acquisition module 604, a conversion module 606 and a determination module 608, wherein: the section classification device 600 for cardiac ultrasound images also includes a quality module 610, which is used to input the target feature representation into the quality regressor, and generate a detection frame in the cardiac ultrasound image based on the target feature representation through the quality regressor, and the detection frame is used to characterize the position of different tissue structures in the heart; obtain the section quality evaluation rule corresponding to the section type of the cardiac ultrasound image; determine the section quality of the cardiac ultrasound image according to the section quality evaluation rule and the detection frame generated in the cardiac ultrasound image. And an evaluation module 612, which is used to input the target feature representation into the quality evaluator, and generate a defect language text based on the target feature representation through the quality evaluator, and the defect language text is used to describe the section defects of the cardiac ultrasound image.

[0141] Each module in the above-mentioned device for classifying sections of cardiac ultrasound images can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each of the above modules.

[0142] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 8 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data related to the section classification of cardiac ultrasound images. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a section classification method for cardiac ultrasound images is implemented.

[0143] Those skilled in the art will understand that Figure 8The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0144] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0145] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0146] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0147] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0148] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0149] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0150] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for classifying sections of cardiac ultrasound images, characterized in that: The method comprises: Extract features from cardiac ultrasound images to obtain image features; Perform semantic segmentation and position detection based on the image features to obtain a semantic segmentation map and corner point position information, wherein the semantic segmentation map includes a plurality of pixels and a tissue to which each pixel belongs, and the corner point position information represents the position of the heart in the cardiac ultrasound image, wherein the heart includes a plurality of tissues; Based on the corner point position information, converting the initial position information of each pixel point in the semantic segmentation map into polar coordinate information; Based on the polar coordinate information of each pixel point and the tissue to which each pixel point belongs, the section type of the cardiac ultrasound image is determined.

2. The method according to claim 1, characterized in that The determining the section type of the cardiac ultrasound image based on the polar coordinate information of each pixel and the tissue to which each pixel belongs includes: Inputting the semantic segmentation map after polar coordinate conversion into a feature extractor, and performing feature extraction on the semantic segmentation map after polar coordinate conversion by the feature extractor to obtain a target feature representation; The target feature representation is input into a slice classifier, and the slice classifier determines the slice type of the cardiac ultrasound image based on the target feature representation.

3. The method according to claim 2, characterized in that The feature extractor extracts features from the semantic segmentation map after polar coordinate conversion to obtain target feature representation, including: The semantic segmentation map after polar coordinate transformation is divided into multiple local images of the same size through the feature extractor; For each local image, convert the local image into an initial feature representation of a preset dimension; Based on the polar coordinate information of the pixel points in the targeted local image, the self-attention weight corresponding to the targeted local image is determined, and the initial feature representation and the self-attention weight are weighted to obtain the target feature representation.

4. The method according to claim 2, characterized in that: Determining the section type of the cardiac ultrasound image based on the target feature representation by the section classifier includes: Flattening and straightening the target feature representation to obtain a one-dimensional feature representation; For each aspect type, obtain the weight value and bias value corresponding to the aspect type; Performing weighted processing on the one-dimensional feature representation and the weight value to obtain a linear combination feature representation; Adding the linear combination feature representation and the bias value to obtain an activation value corresponding to the targeted section type; According to the activation value corresponding to each section type, the section type probability distribution feature vector is obtained; The section type of the cardiac ultrasound image is determined according to the section type probability distribution feature vector.

5. The method according to claim 4, characterized in that The step of determining the section type of the cardiac ultrasound image according to the section type probability distribution feature vector comprises: Convert the activation value corresponding to each section type into the probability that the cardiac ultrasound image is classified into each section type; The section type of the cardiac ultrasound image is obtained according to the probability that the cardiac ultrasound image is classified into each section type.

6. The method according to claim 2, characterized in that The method further comprises: Inputting the target feature representation into a mass regressor, and generating a detection frame in the cardiac ultrasound image based on the target feature representation through the mass regressor, wherein the detection frame is used to characterize the positions of different tissue structures in the heart; Acquire a section quality evaluation rule corresponding to the section type of the cardiac ultrasound image; The section quality of the cardiac ultrasound image is determined according to the section quality evaluation rule and a detection frame generated in the cardiac ultrasound image.

7. The method according to claim 6, characterized in that The method further comprises: The target feature representation is input into a quality assessor, and the quality assessor generates a defect language text based on the target feature representation, wherein the defect language text is used to describe the section defect of the cardiac ultrasound image.

8. A device for classifying sections of cardiac ultrasound images, characterized in that: The device comprises: An extraction module is used to extract features from cardiac ultrasound images to obtain image features; an acquisition module, configured to perform semantic segmentation and position detection based on the image features to obtain a semantic segmentation map and corner point position information, wherein the semantic segmentation map includes a plurality of pixels and a tissue to which each pixel belongs, and the corner point position information represents a position of a heart in the cardiac ultrasound image, wherein the heart includes a plurality of tissues; A conversion module, used for converting the initial position information of each pixel point in the semantic segmentation map into polar coordinate information based on the corner point position information; A determination module is used to determine the section type of the cardiac ultrasound image based on the polar coordinate information of each pixel point and the tissue to which each pixel point belongs.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images