Cardiac ultrasound image segmentation method and system based on artificial intelligence

Abstract

The present invention relates to the field of image segmentation technology, and in particular to an artificial intelligence-based cardiac ultrasound image segmentation method and system, which obtains each candidate cardiac tissue region in each frame of a cardiac ultrasound image sequence, determines a regional feature value according to edge change characteristics and grayscale expression characteristics of each candidate cardiac tissue region; determines a target region matching degree by analyzing edge differences and distance differences between two candidate cardiac tissue regions located in different cardiac ultrasound images according to the regional feature value; uses the target region matching degree to correct an initial clustering radius of each candidate cardiac tissue region to obtain a corrected clustering radius of each candidate cardiac tissue region; and uses a clustering algorithm in combination with the corrected clustering radius to segment each frame of the cardiac ultrasound image to obtain each frame of the cardiac ultrasound image with a marked artifact region.

Description

Technical Field

[0001] The present invention relates to the technical field of image segmentation, and in particular to a cardiac ultrasound image segmentation method and system based on artificial intelligence. Background Art

[0002] Cardiac ultrasound imaging technology is a non-invasive examination method that uses the principle of ultrasound to image the heart. Through the cardiac ultrasound images obtained by cardiac ultrasound imaging technology, doctors can observe the structure, function and blood flow of the heart to determine whether there is heart disease. Cardiac ultrasound imaging technology has the advantages of low cost, no radiation, and non-invasiveness, and has become one of the important means of diagnosing heart disease. However, due to the influence of factors such as the physical properties of ultrasound, the interaction between ultrasound and human tissue, and the performance of the ultrasound instrument itself, a variety of artifacts will appear in ultrasound imaging. The presence of artifacts may interfere with the diagnosis of the disease and lead to misdiagnosis. Among them, artifacts refer to non-random interference that occurs during image processing or medical imaging. The interference appears in the image as images of tissues or lesions that do not exist originally.

[0003] By performing image segmentation on cardiac ultrasound images, it is possible to distinguish between normal cardiac tissue areas, artifact areas, or areas affected by artifact images. Existing deep learning-based image segmentation methods train models based on training data, input the image to be segmented into the trained model, and output the segmentation results. Existing deep learning-based image segmentation methods have the following shortcomings when processing cardiac ultrasound images:

[0004] The image segmentation algorithm performs inconsistently in different patients or under different imaging conditions. The model has not been fully trained to deal with different artifact situations and has limited generalization capabilities. Secondly, due to the complexity of artifact types, the annotations of artifact areas in the training data may not be accurate, which affects the learning and segmentation performance of the algorithm, that is, whether the features learned by the model can effectively identify and process artifacts. Finally, the noise in cardiac ultrasound images will reduce the segmentation accuracy and make the processing of artifact areas more difficult. Summary of the invention

[0005] In order to solve the technical problem of low segmentation accuracy of the above-mentioned cardiac ultrasound image segmentation method, the purpose of the present invention is to provide a cardiac ultrasound image segmentation method and system based on artificial intelligence, and the technical solutions adopted are as follows:

[0006] An embodiment of the present invention provides a cardiac ultrasound image segmentation method based on artificial intelligence, the method comprising the following steps:

[0007] Acquire a cardiac ultrasound image sequence by using cardiac ultrasound imaging technology, and then determine each candidate cardiac tissue region in each frame of the cardiac ultrasound image in the cardiac ultrasound image sequence;

[0008] Determining a regional feature value of each candidate heart tissue region according to edge change characteristics and grayscale performance characteristics of each candidate heart tissue region;

[0009] Analyze edge differences and distance differences of two candidate cardiac tissue regions located in different cardiac ultrasound images to determine first regional difference factors of the two candidate cardiac tissue regions;

[0010] Determining a target region matching degree of each candidate cardiac tissue region according to the region feature value of each candidate cardiac tissue region and a plurality of the first region difference factors;

[0011] Determining an initial clustering radius of each candidate cardiac tissue region; using the target region matching degree, correcting the initial clustering radius of each candidate cardiac tissue region to obtain a corrected clustering radius of each candidate cardiac tissue region;

[0012] In combination with the modified clustering radius, a clustering algorithm is used to perform segmentation processing on each frame of cardiac ultrasound image to obtain each frame of cardiac ultrasound image with a marked artifact area.

[0013] Furthermore, determining the regional feature value of each candidate cardiac tissue region according to the edge variation feature and the grayscale representation feature of each candidate cardiac tissue region includes:

[0014] For any candidate cardiac tissue region, edge detection is performed on the candidate cardiac tissue region to obtain edge pixel points of the candidate cardiac tissue region; and a first region characteristic factor of the candidate cardiac tissue region is determined according to a difference in gradient amplitude between two adjacent edge pixel points;

[0015] Determine a grayscale performance index of each pixel point according to the grayscale value of each pixel point in the candidate heart tissue area;

[0016] Determine a second region characteristic factor of the candidate heart tissue region according to an average value of the grayscale performance indexes of all pixels in the candidate heart tissue region and a difference in the grayscale performance indexes between the two adjacent edge pixels;

[0017] The first region characteristic factor and the second region characteristic factor of the candidate heart tissue region are combined to determine the region characteristic value of the candidate heart tissue region.

[0018] Furthermore, determining the grayscale performance index of each pixel point according to the grayscale value includes:

[0019] For any pixel, calculate the grayscale difference between the pixel and any adjacent pixel;

[0020] Determine the grayscale performance index of the pixel point by combining the grayscale value of the pixel point and the grayscale difference value;

[0021] The grayscale value of the pixel point is positively correlated with the grayscale performance index, and the grayscale difference value is negatively correlated with the grayscale performance index.

[0022] Furthermore, determining the second region characteristic factor of the candidate heart tissue region according to the average value of the grayscale performance index of all pixels in the candidate heart tissue region and the difference of the grayscale performance index between the two adjacent edge pixels includes:

[0023] Calculate a first ratio of the grayscale performance index between two adjacent edge pixels, and calculate the absolute value of the difference between 1 and the first ratio; calculate a second ratio of the average value of the grayscale performance index of all pixels in the candidate heart tissue area and the absolute value of the difference between 1 and the first ratio; add all of the second ratios to obtain a second area characteristic factor of the candidate heart tissue area.

[0024] Further, analyzing the edge difference and distance difference of two candidate cardiac tissue regions located in different cardiac ultrasound images to determine the first region difference factor of the two candidate cardiac tissue regions includes:

[0025] Taking a candidate cardiac tissue region in any cardiac ultrasound image as a region to be analyzed, and taking a candidate cardiac tissue region in cardiac ultrasound images other than the cardiac ultrasound image as a comparison region;

[0026] For any area to be analyzed and any comparison area, determine the centroid position of the area to be analyzed and the comparison area and the position of each edge pixel point on the area boundary;

[0027] According to the centroid position and the position of each edge pixel point on the region boundary, the average distance from the centroid of the region to be analyzed and the comparison region to all edge pixel points on the corresponding region boundary is calculated;

[0028] Determining the degree of difference in distance between the area to be analyzed and the comparison area according to the difference between the distance mean of the area to be analyzed and the distance mean of the comparison area;

[0029] A first area difference factor between the area to be analyzed and the comparison area is determined according to the distance difference degree, the centroid distance and the difference in the number of edge pixels of the area boundaries between the area to be analyzed and the comparison area.

[0030] Furthermore, the determining of the first area difference factor between the area to be analyzed and the comparison area according to the distance difference degree, the centroid distance and the difference in the number of edge pixels of the area boundaries between the area to be analyzed and the comparison area includes:

[0031] Mapping the centroid of the comparison area to the area to be analyzed, and calculating the distance between the two centroids, which is recorded as the centroid distance;

[0032] Counting the number of edge pixels on the boundary of the area to be analyzed and the area to be compared, and taking the difference between the two numbers of edge pixels as the difference in the number of edge pixels;

[0033] The product of the distance difference degree, the centroid distance and the difference in the number of edge pixels is used as the first area difference factor between the area to be analyzed and the area to be compared.

[0034] Furthermore, determining the target region matching degree of each candidate cardiac tissue region according to the region feature value of each candidate cardiac tissue region and a plurality of the first region difference factors includes:

[0035] Determine a second regional difference factor between the region to be analyzed and the comparison region according to the regional characteristic value of the region to be analyzed and the difference between the regional characteristic values ​​of the region to be analyzed and the comparison region;

[0036] Determine the degree of regional matching between the region to be analyzed and the comparison region by combining the first regional difference factor and the second regional difference factor of the region to be analyzed and the comparison region;

[0037] Wherein, the first regional difference factor, the second regional difference factor and the regional matching degree are both negatively correlated;

[0038] The region matching degree between the region to be analyzed and each comparison region is obtained, and the maximum value among all the region matching degrees is selected as the target region matching degree of the region to be analyzed.

[0039] Furthermore, determining a second regional difference factor between the region to be analyzed and the comparison region according to the regional characteristic value of the region to be analyzed and the difference between the regional characteristic values ​​of the region to be analyzed and the comparison region includes:

[0040] The absolute value of the difference between the regional characteristic value of the region to be analyzed and the regional characteristic value of the comparison region is calculated, and the ratio of the absolute value of the difference to the regional characteristic value of the region to be analyzed is used as the second regional difference factor between the region to be analyzed and the comparison region.

[0041] Furthermore, the determining of the initial clustering radius of each candidate cardiac tissue region includes:

[0042] For any candidate heart tissue region, calculating the difference between the maximum grayscale value and the minimum grayscale value in the candidate heart tissue region as the first clustering influence factor;

[0043] Calculate the distance value between any two pixels at the same gray level, and then calculate the average value of all the distance values ​​at the same gray level as the distance mean, and take the average value of the distance means corresponding to all gray levels as the second clustering influence factor;

[0044] Determine an initial clustering radius of the candidate heart tissue region by combining the first clustering influence factor and the second clustering influence factor;

[0045] The first clustering influence factor, the second clustering influence factor and the initial clustering radius are negatively correlated.

[0046] An embodiment of the present invention also provides an artificial intelligence-based cardiac ultrasound image segmentation system, including a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement an artificial intelligence-based cardiac ultrasound image segmentation method.

[0047] The present invention has the following beneficial effects:

[0048] The present invention provides a cardiac ultrasound image segmentation method and system based on artificial intelligence. First, each candidate cardiac tissue region in each frame of cardiac ultrasound image is determined, which can reduce the amount of data for image analysis to a certain extent and save image segmentation steps. Secondly, for each candidate cardiac tissue region, a target region matching degree is determined for each candidate cardiac tissue region in combination with the image features of the normal cardiac tissue region, which can achieve in-depth analysis of the degree to which the candidate cardiac tissue region fits the normal cardiac tissue region. The numerical accuracy and reliability of the target region matching degree determined by multiple steps are stronger. Finally, an initial clustering radius is determined, and the initial clustering radius is corrected using the target region matching degree to obtain a corrected clustering radius. The corrected clustering radius combined with the target region matching degree can better cluster pixel points with cardiac tissue regions together, which is convenient for subsequent high-precision segmentation processing. The segmentation processing of the cardiac ultrasound image is implemented based on the corrected clustering radius, and an accurate segmentation result can be obtained. Subsequently, based on the accurate segmentation result, the artifact region or the region affected by the artifact can be marked, which reduces the degree to which the cardiac ultrasound image is affected by the artifact to a certain extent, and further effectively improves the accuracy of the doctor's diagnosis result. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 It is a flow chart of the cardiac ultrasound image segmentation method based on artificial intelligence of the present invention;

[0051] Figure 2 is an example diagram of any frame of a cardiac ultrasound image in an embodiment of the present invention;

[0052] Figure 3 is an example diagram of a cardiac ultrasound image after threshold segmentation processing in an embodiment of the present invention;

[0053] Figure 4 is a flowchart for implementing step S3 in an embodiment of the present invention;

[0054] Figure 5 4 is a flowchart for implementing step S4 in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the technical solutions proposed by the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0056] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0057] The application scenarios targeted by the present invention may be:

[0058] Cardiac ultrasound imaging can specifically be two-dimensional echocardiography, which is an imaging technology commonly used in the clinical diagnosis of heart disease. However, due to the influence of factors such as the physical properties of ultrasound, the interaction between ultrasound and human tissue, and the performance of the ultrasound instrument itself, various artifacts may appear in ultrasound imaging, and cardiac artifacts will affect the accuracy of heart segmentation in ultrasound images, thereby affecting the accuracy of doctors' diagnostic results.

[0059] In order to overcome the adverse effects of artifacts on cardiac ultrasound image segmentation, the present invention corrects the initial clustering radius to obtain a corrected clustering radius through the characteristics of the cardiac tissue area and the artifact area in the cardiac ultrasound image, as well as the regional feature differences between multiple cardiac ultrasound images. The corrected clustering radius is used to perform regional segmentation on the cardiac ultrasound image to obtain a cardiac ultrasound image with the artifact area marked. The marked cardiac ultrasound image can provide data support for doctors to make accurate diagnoses.

[0060] Specifically, an embodiment of the present invention provides a cardiac ultrasound image segmentation method based on artificial intelligence, such as Figure 1 As shown, the following steps are included:

[0061] S1, obtaining a cardiac ultrasound image sequence by using cardiac ultrasound imaging technology, and then determining each candidate cardiac tissue region in each frame of the cardiac ultrasound image in the cardiac ultrasound image sequence.

[0062] Here, the cardiac ultrasound image sequence includes multiple frames of cardiac ultrasound images, and the position of the artifact area in each frame of the cardiac ultrasound image may be different. In order to accurately distinguish the normal cardiac tissue area and the artifact area in the cardiac ultrasound image, it is necessary to obtain multiple frames of cardiac ultrasound images and determine all areas that may be cardiac tissues in each frame of the cardiac ultrasound image, that is, determine candidate cardiac tissue areas. The candidate cardiac tissue areas may be normal cardiac tissue areas, artifact areas, areas affected by artifacts, and abnormal areas.

[0063] The above step S1 can be implemented through steps S11 to S12 (not shown in the figure):

[0064] S11, obtaining a cardiac ultrasound image sequence by using cardiac ultrasound imaging technology.

[0065] In this embodiment, in order to subsequently mark the artifact area or the area affected by the artifact in the cardiac ultrasound image, it is first necessary to obtain an ultrasound image of the heart through an ultrasound instrument. The ultrasound image of the heart is a single static image obtained through cardiac ultrasound examination. It is a part of the image, showing the image of the heart at a specific moment, and is usually used to observe the structure of the heart and evaluate abnormalities; in addition, a single ultrasound image can provide detailed information on a specific section or perspective of the heart, and multiple images are usually required to obtain comprehensive diagnostic information. Therefore, in the actual operation process, the ultrasound image of each moment is continuously obtained by the ultrasound instrument to form a dynamic sequence, and the dynamic sequence can provide detailed images of the heart at each moment.

[0066] Specifically, by scanning the patient's heart with an ultrasonic instrument, a corresponding heart scan video of the patient can be obtained, and the heart scan video can be converted into an image to form an initial heart ultrasound image sequence; due to the dynamic characteristics of the scan, continuous artifacts may appear in the images at adjacent time points in the heart scan video. In order to facilitate the subsequent analysis of the differences in artifact areas in different heart ultrasound images, all images in the initial heart ultrasound image sequence are divided into groups of 15 frames, and the first frame of each group is selected as the image representing the corresponding group. The representative images of each group are organized into a new image set in chronological order, namely, a heart ultrasound image sequence. Subsequently, image analysis will be performed on each heart ultrasound image in the heart ultrasound image sequence. Among them, the number of image divisions in the initial heart ultrasound image sequence can be set by the implementer according to the specific actual situation, and is not specifically limited; an example of a heart ultrasound image of any frame is shown in the figure below. Figure 2 shown.

[0067] It should be noted that regarding ultrasound images and ultrasound images, doctors use ultrasound images when evaluating the state of the heart because it includes all types of image data. Ultrasound images are part of ultrasound images, and doctors use multiple ultrasound images for detailed analysis and diagnosis, so image analysis of ultrasound images is required to achieve more accurate ultrasound image segmentation.

[0068] S12, determining each candidate cardiac tissue region in each frame of the cardiac ultrasound image in the cardiac ultrasound image sequence.

[0069] In this embodiment, in order to determine the properties of the candidate cardiac tissue regions and mark the artifact regions in the image, it is first necessary to select from the cardiac ultrasound image the cardiac tissue regions that may be normal, artifacts, or cardiac tissue regions affected by artifacts, that is, to determine the candidate cardiac tissue regions in the cardiac ultrasound image.

[0070] Specifically, in order to analyze the grayscale features of the image, each frame of the cardiac ultrasound image is firstly grayscale processed to obtain the grayscale processed cardiac ultrasound image; then the connected domain algorithm is used to identify the connected domains in the cardiac ultrasound image to obtain each connected domain in each frame of the cardiac ultrasound image; then, each frame of the cardiac ultrasound image is threshold segmented, and the area with a grayscale value greater than the threshold is recorded as the foreground area, and the area with a grayscale value less than the threshold is recorded as the background area; since the grayscale values ​​of the cardiac tissue area and the artifact area are relatively high, the foreground area is used as a candidate cardiac tissue area, while the grayscale values ​​of the ventricles, atria and other areas are relatively low and are located in the background area, so in general, the background area will not be confused with the artifact area.

[0071] Among them, the implementation processes of grayscale processing, connected domain algorithm and threshold segmentation algorithm are all prior arts, which are not within the scope of protection of the present invention and will not be elaborated in detail here; an example of a cardiac ultrasound image after threshold segmentation processing is shown in FIG. Figure 3 shown.

[0072] So far, this embodiment has obtained each candidate cardiac tissue region in each frame of cardiac ultrasound image.

[0073] S2, determining a regional feature value of each candidate heart tissue region according to edge variation characteristics and grayscale representation characteristics of each candidate heart tissue region.

[0074] Here, the regional eigenvalue refers to the degree of image features of the cardiac tissue exhibited by the candidate cardiac tissue region. The larger the regional eigenvalue, the greater the possibility that the candidate cardiac tissue region belongs to the normal cardiac tissue region, and the smaller the possibility that it belongs to the artifact region or the region affected by the artifact. The image features of the cardiac tissue region are quantified from two aspects, namely, the edge changes of the region and the grayscale expression. Determining the eigenvalue from these two aspects can effectively enhance the numerical accuracy and representative significance of the regional eigenvalue.

[0075] The above step S2 can be implemented through steps S21 to S24 (not shown in the figure):

[0076] S21, for any candidate cardiac tissue region, edge detection is performed on the candidate cardiac tissue region to obtain each edge pixel point of the candidate cardiac tissue region; based on the difference in gradient amplitude between two adjacent edge pixel points, a first region characteristic factor of the candidate cardiac tissue region is determined.

[0077] In this embodiment, since the edge of the heart tissue region in the ultrasound image is usually regular and clear, the grayscale gradient amplitude of the edge in the normal heart tissue region will be relatively consistent, that is, the gradient change degree of the edge pixel point is relatively uniform; the artifact is usually caused by the noise of the ultrasound signal, reflection error or other imaging problems, so the edge of the artifact region in the image is relatively irregular, and the edge of the artifact region may appear chaotic or blurred. Therefore, there is a clear difference in the distribution pattern of edge pixels between the artifact region and the normal heart tissue region.

[0078] Specifically, the Canny edge detection operator is used to perform edge detection on any candidate cardiac tissue region to obtain each edge in the candidate cardiac tissue region, and then each edge pixel on each edge is obtained; the gradient amplitude is calculated based on the grayscale value of each edge pixel, and the gradient amplitude difference values ​​of all two adjacent edge pixels are negatively correlated to obtain the values ​​after negative correlation processing; all the values ​​after negative correlation processing are cumulatively calculated to obtain the first regional characteristic factor of the candidate cardiac tissue region. Among them, the gradient amplitude difference value refers to the absolute value of the difference between the gradient amplitudes of two adjacent edge pixels; the process of implementing edge detection by the Canny edge detection operator is a prior art, which is not within the protection scope of the present invention and will not be elaborated in detail here.

[0079] As an example, the calculation formula of the first region characteristic factor of the i-th candidate cardiac tissue region in the k-th frame of cardiac ultrasound image may be:

[0080] ; In the formula, represents the first regional characteristic factor of the i-th candidate cardiac tissue region in the k-th frame of cardiac ultrasound image, n represents the number of edge pixels in the i-th candidate cardiac tissue region, represents the gradient amplitude of the wth edge pixel in the ith candidate cardiac tissue region in the kth frame of cardiac ultrasound image, represents the gradient amplitude of the w+1th edge pixel in the ith candidate cardiac tissue region in the kth frame of cardiac ultrasound image, represents the absolute value function, represents the gradient amplitude difference value, Express Negative correlation processing.

[0081] It is worth noting that under normal circumstances, the gradient amplitudes of two adjacent edge pixels are not completely equal. Once a special case occurs, in order to ensure that the denominator of the fraction is not zero, a constant, such as 0.01, needs to be added to the denominator of the fraction.

[0082] S22, determining a grayscale performance index for each pixel point according to the grayscale value of each pixel point in the candidate heart tissue region.

[0083] In this embodiment, for the real heart tissue region, because the echo signal of normal heart tissue is relatively stable, the heart is shown in the ultrasound image as a lighter and uniform area, showing the characteristics of a higher grayscale and a relatively gentle change in the region; compared with the heart tissue region, the artifact may be caused by signal noise or other imaging problems, so the artifact or other abnormal area usually appears in a darker color in the ultrasound image, with a lower grayscale value and a certain change, and the change between pixels in the artifact area is usually larger than that in the heart tissue area. Therefore, the grayscale representation value can quantify the possibility that the candidate heart tissue region is a normal heart tissue region based on the grayscale characteristics of the candidate heart tissue region. The larger the grayscale representation value, the greater the possibility that the candidate heart tissue region is a normal heart tissue region.

[0084] Specifically, for any pixel, the grayscale difference value between the pixel and any adjacent pixel is calculated; the grayscale performance index of the pixel is determined by combining the grayscale value and the grayscale difference value of the pixel; wherein the grayscale value of the pixel is positively correlated with the grayscale performance index, and the grayscale difference value is negatively correlated with the grayscale performance index. The grayscale difference value refers to the absolute value of the difference between the grayscale values ​​of two pixels.

[0085] As an example, the calculation formula for the grayscale representation value of the rth pixel in the ith candidate tissue region in the kth frame of the cardiac ultrasound image may be:

[0086] ; In the formula, represents the grayscale value of the rth pixel in the i-th candidate tissue area in the k-th frame of cardiac ultrasound image, represents the gray value of the rth pixel in the ith candidate tissue area in the kth frame of cardiac ultrasound image, represents the gray value of the r+1th pixel in the ith candidate tissue region in the kth frame of cardiac ultrasound image, represents the grayscale difference value of the rth pixel in the i-th candidate tissue area in the k-th frame of cardiac ultrasound image. +0.1 is to prevent Zero occurs.

[0087] In the calculation formula of the grayscale expression value, the smaller the grayscale difference value is, the smaller the difference between the grayscale value of the rth pixel point in the area and the grayscale value of the r+1th pixel point is, that is, the grayscale change is relatively smooth and uniform; the larger the grayscale value of the rth pixel point in the area, the lighter the color of the pixel point; when the grayscale change of the pixel point is smoother and the grayscale value is larger, it means that the grayscale expression value of the corresponding pixel point is larger, and the possibility that the pixel point in the area is a pixel point in the cardiac tissue area is greater.

[0088] S23, determining a second region characteristic factor of the candidate heart tissue region according to an average value of the grayscale performance indexes of all pixels in the candidate heart tissue region and a difference in the grayscale performance indexes between two adjacent edge pixels.

[0089] Specifically, a first ratio of the grayscale performance index between two adjacent edge pixels is calculated, and the absolute value of the difference between 1 and the first ratio is calculated; a second ratio of the average value of the grayscale performance index of all pixels in the candidate heart tissue area and the absolute value of the difference is calculated; all the second ratios are added together to obtain the second area characteristic factor of the candidate heart tissue area.

[0090] As an example, the calculation formula for the second region characteristic factor of the i-th candidate cardiac tissue region in the k-th frame of cardiac ultrasound image may be:

[0091] ; In the formula, represents the second regional characteristic factor of the i-th candidate cardiac tissue region in the k-th frame of cardiac ultrasound image, n represents the number of edge pixels in the i-th candidate cardiac tissue region, represents the average value of the grayscale performance index of all pixels in the ith candidate cardiac tissue region in the kth frame of cardiac ultrasound image, It represents the grayscale performance index of the wth edge pixel in the ith candidate cardiac tissue region in the kth frame of cardiac ultrasound image. It represents the gray performance index of the w+1th edge pixel in the ith candidate cardiac tissue region in the kth frame of cardiac ultrasound image. represents the first ratio, Represents the second ratio.

[0092] In the calculation formula of the second regional characteristic factor, There is no possibility of zero, the first ratio The closer it is to 1, the more similar the grayscale values ​​of adjacent edge pixels are, and the more it can express the characteristics of the heart tissue area. Therefore, the absolute value of the difference between 1 and the first ratio is negatively correlated with the second regional characteristic factor. It represents the grayscale expression degree of the cardiac tissue characteristics of the candidate cardiac tissue region. The larger the overall grayscale value of the region, the closer the grayscale image characteristics of the candidate cardiac tissue region are to the grayscale characteristics of the cardiac tissue region. Therefore, the average grayscale expression value of all pixels in the candidate cardiac tissue region is positively correlated with the second regional characteristic factor.

[0093] S24, combining the first regional characteristic factor and the second regional characteristic factor of the candidate cardiac tissue region to determine the regional characteristic value of the candidate cardiac tissue region.

[0094] Specifically, the first regional characteristic factor, the second regional characteristic factor and the regional characteristic value are all positively correlated, so the product of the first regional characteristic factor and the second regional characteristic factor can be used as the regional characteristic value corresponding to the candidate heart tissue region.

[0095] Of course, the value obtained by adding the first regional characteristic factor and the second regional characteristic factor may also be used as the regional characteristic value of the candidate cardiac tissue region.

[0096] So far, this embodiment can obtain the regional feature value of each candidate cardiac tissue region in each frame of cardiac ultrasound image.

[0097] S3, analyzing edge differences and distance differences of two candidate cardiac tissue regions located in different cardiac ultrasound images, and determining first region difference factors of the two candidate cardiac tissue regions.

[0098] Here, when using an ultrasound instrument, a slight movement of the probe may cause a slight change in the shape and grayscale representation of the image, but the relative position movement of the same normal heart tissue area corresponding to each frame of the heart ultrasound image is usually not large, and the same heart ultrasound image in different heart ultrasound images is similar in grayscale representation. Therefore, when performing a matching analysis on two candidate heart tissue areas located in different heart ultrasound images, it is necessary to consider the regional differences between each frame of the image, that is, to determine the first regional difference factor. Among them, the first regional difference factor can characterize the edge matching and distance between the two candidate heart tissue areas located in different heart ultrasound images. The larger the first regional difference factor, the smaller the matching degree of the two candidate heart tissue areas located in different heart ultrasound images.

[0099] The above step S3 can be Figure 4 The steps shown achieve:

[0100] S31, taking a candidate cardiac tissue region in any cardiac ultrasound image as a region to be analyzed, and taking a candidate cardiac tissue region in cardiac ultrasound images other than the cardiac ultrasound image as a comparison region.

[0101] In this embodiment, when analyzing the first regional difference factor, it is determined based on the candidate cardiac tissue regions in two different cardiac ultrasound images. Each cardiac ultrasound image needs to perform regional matching analysis with other cardiac ultrasound images except itself. For ease of understanding, all candidate cardiac tissue regions in any cardiac ultrasound image are used as regions to be analyzed, and candidate cardiac tissue regions in cardiac ultrasound images other than the cardiac ultrasound image are used as comparison regions. The comparison region is the comparison region of each region to be analyzed, and each region to be analyzed has corresponding multiple comparison regions.

[0102] S32, for any region to be analyzed and any comparison region, determine the centroid position of the region to be analyzed and the comparison region and the position of each edge pixel point on the region boundary.

[0103] In this embodiment, taking any area to be analyzed and any comparison area as examples, the first area difference factor of the area to be analyzed and the comparison area is determined. Before analyzing the first area difference factor, in order to facilitate the subsequent determination of the distance difference degree, it is first necessary to determine the centroid position of the area to be analyzed and the comparison area and the position of each edge pixel point on the area boundary. Among them, the process of determining the centroid and area boundary in the area is a prior art and is not within the scope of protection of the present invention, and will not be elaborated here; the area boundary specifically refers to the edge located at the outermost layer of the area, which can be obtained by an edge detection algorithm.

[0104] S33, calculating the average of the distances from the centroid of the area to be analyzed and the comparison area to all edge pixels on the border of the corresponding area according to the centroid position and the position of each edge pixel on the border of the area.

[0105] In this embodiment, the distance mean can be used to characterize the position of the centroid of the region relative to the region boundary. The distance mean is a data indicator used to calculate the degree of distance difference later. The smaller the difference between the distance means of the region to be analyzed and the comparison region, the greater the possibility that the region to be analyzed and the comparison region belong to the same region, and the higher the first regional difference factor of the region to be analyzed and the comparison region.

[0106] Specifically, the Euclidean distance from the centroid of the area to be analyzed to all edge pixels on the boundary of the corresponding area is calculated, and the average value of all Euclidean distances is used as the distance mean of the area to be analyzed; the Euclidean distance from the centroid of the comparison area to all edge pixels on the boundary of the corresponding area is calculated, and the average value of all Euclidean distances is used as the distance mean of the comparison area. Among them, the calculation formula of the Euclidean distance is a prior art, which is not within the protection scope of the present invention and will not be elaborated in detail here.

[0107] S34, determining the degree of difference in distance between the area to be analyzed and the comparison area according to the difference between the distance mean of the area to be analyzed and the distance mean of the comparison area.

[0108] Specifically, the absolute value of the difference between the distance mean of the area to be analyzed and the distance mean of the comparison area is calculated, and the absolute value of the difference between the two distance means is used as the distance difference degree between the area to be analyzed and the comparison area.

[0109] S35, determining a first area difference factor between the area to be analyzed and the comparison area according to the distance difference degree, the centroid distance, and the difference in the number of edge pixels of the area boundaries between the area to be analyzed and the comparison area.

[0110] The above step S35 can be implemented through steps S351 to S353 (not shown in the figure):

[0111] S351, mapping the centroid of the comparison area to the area to be analyzed, and calculating the distance between the two centroids, which is recorded as the centroid distance.

[0112] S352, counting the number of edge pixels on the boundary of the area to be analyzed and the area to be compared, and taking the difference between the two numbers of edge pixels as the difference in the number of edge pixels.

[0113] S353: taking the product of the distance difference degree, the centroid distance and the difference in the number of edge pixels as the first area difference factor between the area to be analyzed and the comparison area.

[0114] As an example, the calculation formula of the first regional difference factor between the region to be analyzed and the comparison region may be:

[0115] ; Where D represents the first regional difference factor between the area to be analyzed and the comparison area, d represents the distance difference between the area to be analyzed and the comparison area, It represents the centroid distance between the area to be analyzed and the comparison area, and N represents the difference in the number of edge pixels between the area to be analyzed and the comparison area.

[0116] In the calculation formula of the first regional difference factor, It indicates the degree of position difference between the area to be analyzed and the comparison area in the cardiac ultrasound image. The smaller the position difference, the closer the position is, which means it is more likely to be the same area. N indicates the difference between the number of edge pixels of the area to be analyzed and the number of edge pixels of the comparison area. The smaller N is, the more similar the edge states of the area to be analyzed and the comparison area are, and the area to be analyzed is more likely to be a normal cardiac tissue area.

[0117] It should be noted that the first region difference factors are calculated and analyzed for the region to be analyzed and each comparison region, and several first region difference factors corresponding to the region to be analyzed can be obtained. By referring to the calculation method of the first region difference factors, several first region difference factors of each candidate cardiac tissue region can be obtained.

[0118] So far, this embodiment has obtained several first region difference factors of each candidate heart tissue region.

[0119] S4, determining the target region matching degree of each candidate heart tissue region according to the region feature value of each candidate heart tissue region and a plurality of first region difference factors.

[0120] Here, when marking the artifact area, the artifact area is false information that appears in the image due to various factors, which will affect the accurate segmentation of the normal heart tissue area. Artifacts are formed for various reasons, and they usually do not appear or repeat in the same form in multiple images. Therefore, the appearance of artifacts in different images may be different. Specifically, the artifact may be connected to the normal heart tissue area, thereby changing the boundary and width of the normal area, or the artifact may be independent of the normal tissue area, forming an area different from the surrounding tissue, which will distort the characteristics of the normal heart tissue area, thereby affecting the accuracy of regional segmentation.

[0121] Therefore, it is necessary to compare the candidate cardiac tissue region in the current cardiac ultrasound image with the candidate cardiac tissue regions in other cardiac ultrasound images. If the candidate cardiac tissue region in the current cardiac ultrasound image has a high degree of match with a candidate cardiac tissue region in other cardiac ultrasound images, it indicates that the candidate cardiac tissue region in the current cardiac ultrasound image is likely to be a real cardiac tissue region and conforms to the characteristics of normal cardiac tissue; if the matching result shows that the current candidate cardiac tissue region has a low degree of match with the candidate cardiac tissue regions in other cardiac ultrasound images, it indicates that the current candidate cardiac tissue region may be affected by the image of the artifact region, or may be a separate artifact region.

[0122] In this embodiment, the target region matching degree is determined by combining the regional feature value and the first regional difference factor. The target region matching degree refers to the matching degree corresponding to the region that best matches the candidate cardiac tissue region. The target region matching degree can characterize the degree of image characteristics of the cardiac tissue region exhibited by the candidate cardiac tissue region. The greater the target region matching degree, the greater the possibility that the candidate cardiac tissue region is a normal cardiac tissue region.

[0123] The above step S4 can be performed by Figure 5 The steps shown achieve:

[0124] S41, determining a second regional difference factor between the region to be analyzed and the comparison region according to the regional characteristic value of the region to be analyzed and the difference between the regional characteristic values ​​of the region to be analyzed and the comparison region.

[0125] Specifically, the absolute value of the difference between the regional characteristic value of the region to be analyzed and the regional characteristic value of the comparison region is calculated, and the ratio of the absolute value of the difference to the regional characteristic value of the region to be analyzed is used as the second regional difference factor between the region to be analyzed and the comparison region.

[0126] S42, combining the first regional difference factor and the second regional difference factor of the region to be analyzed and the comparison region to determine the regional matching degree between the region to be analyzed and the comparison region.

[0127] In this embodiment, the first regional difference factor, the second regional difference factor and the regional matching degree are negatively correlated.

[0128] As an example, the calculation formula for the area matching degree between the area to be analyzed and the comparison area can be:

[0129] ; In the formula, E represents the regional matching degree between the area to be analyzed and the comparison area, exp represents the exponential function with the natural constant as the base, and R represents the regional characteristic value of the area to be analyzed. represents the regional characteristic value of the contrast area, Indicates the absolute value of the difference between the regional characteristic value of the area to be analyzed and the regional characteristic value of the comparison area, represents the second regional difference factor, and D represents the first regional difference factor.

[0130] In the calculation formula of regional matching degree, the area to be analyzed and each comparison area have a regional matching degree. It represents the difference in regional characteristic values ​​between the region to be analyzed and the comparison region. The smaller the difference in regional characteristic values, the higher the similarity between the two regions, the more likely the region to be analyzed is a normal heart tissue region, and the higher the regional matching degree. The larger the regional characteristic value of the region to be analyzed, the more obvious the characteristics of the heart tissue region, the greater the possibility of matching, and the higher the regional matching degree. The smaller the first regional difference factor, the smaller the central difference between the region to be analyzed and the comparison region, the more likely the two regions are the same region, and the corresponding regional matching degree is greater. exp (-) can be used to achieve negative correlation of data. Of course, the implementer can also use other calculation methods to achieve negative correlation of data, such as finding the inverse.

[0131] S43, obtaining the region matching degree between the region to be analyzed and each comparison region, and selecting the maximum value among all the region matching degrees as the target region matching degree of the region to be analyzed.

[0132] In this embodiment, the calculation process of the regional matching degree between the region to be analyzed and any comparison region can be referred to to obtain the regional matching degree between the region to be analyzed and each comparison region. Since the greater the regional matching degree, the more similar the region to be analyzed is to a comparison region, and the greater the possibility of using it to subsequently correct the initial clustering radius of the region to be analyzed, the maximum value can be used as the target regional matching degree of the region to be analyzed.

[0133] So far, this embodiment obtains the target region matching degree of each candidate heart tissue region.

[0134] S5, determining the initial clustering radius of each candidate heart tissue region; and using the target region matching degree, correcting the initial clustering radius of each candidate heart tissue region to obtain a corrected clustering radius of each candidate heart tissue region.

[0135] Here, the initial clustering radius can be quantified by the grayscale distribution of pixels. The initial clustering radius does not take into account the various image feature differences between the normal heart tissue area and the artifact area, which makes the clustering segmentation result determined by the initial clustering radius biased. It is difficult to accurately segment the heart tissue area affected by the artifact area, which increases the difficulty for doctors to diagnose the heart condition. Therefore, it is necessary to use the target area matching degree to correct the initial clustering radius to obtain a more reliable and accurate corrected clustering radius.

[0136] The above step S5 can be implemented through steps S51 to S52 (not shown in the figure):

[0137] S51, determining an initial clustering radius of each candidate heart tissue region.

[0138] Specifically, for any candidate cardiac tissue region, the difference between the maximum grayscale value and the minimum grayscale value in the candidate cardiac tissue region is calculated as the first clustering influence factor; the distance value between any two pixel points at the same grayscale is calculated, and then the average value of all distance values ​​at the same grayscale is calculated and recorded as the distance mean, and the average value of the distance means corresponding to all grayscales is used as the second clustering influence factor; the initial clustering radius of the candidate cardiac tissue region is determined by combining the first clustering influence factor and the second clustering influence factor.

[0139] Among them, the first clustering influence factor and the second clustering influence factor are negatively correlated with the initial clustering radius.

[0140] As an example, the calculation formula for the initial clustering radius of the i-th candidate cardiac tissue region in the u-th frame cardiac ultrasound image may be:

[0141] ; In the formula, represents the initial clustering radius of the i-th candidate cardiac tissue region in the u-th frame of cardiac ultrasound image, represents the maximum grayscale value in the i-th candidate cardiac tissue region in the u-th frame cardiac ultrasound image, represents the minimum gray value in the i-th candidate cardiac tissue region in the u-th frame cardiac ultrasound image, represents the first clustering influencing factor, I represents the number of gray levels in the i-th candidate heart tissue region, m represents the number of pixels at the j-th gray level, represents the distance between the rth pixel and the fth pixel at the jth gray level in the i-th candidate cardiac tissue region, Represents the second clustering image factor.

[0142] In the calculation formula of the initial cluster radius, It represents the difference between the maximum grayscale value and the minimum grayscale value in the i-th candidate cardiac tissue region. The larger the difference, the more likely the i-th candidate cardiac tissue region is to be an artifact region or a region affected by an artifact. In order to distinguish between the artifact region and the normal cardiac tissue region, the value of the initial clustering radius should be smaller. It represents the average distance between all pixels with the same gray value in the i-th candidate cardiac tissue region. The smaller the average distance, the more concentrated the distribution of pixels with similar gray values ​​in the candidate cardiac tissue region is, and the more likely it is a normal cardiac tissue region, and the larger the initial clustering radius should be.

[0143] S52, using the target region matching degree, correcting the initial clustering radius of each candidate heart tissue region to obtain a corrected clustering radius of each candidate heart tissue region.

[0144] Specifically, the initial clustering radius of each candidate heart tissue region is multiplied by the matching degree of the target region to obtain the corrected clustering radius of the corresponding candidate heart tissue region, thereby obtaining the corrected clustering radius of each candidate heart tissue region.

[0145] It should be noted that by implementing segmentation processing of cardiac ultrasound images based on the modified clustering radius, accurate segmentation results can be obtained. Subsequently, based on the accurate segmentation results, the artifact areas and areas affected by artifacts can be marked, which to a certain extent reduces the degree to which cardiac ultrasound images are affected by artifacts, and can further effectively improve the accuracy of doctors' diagnostic results.

[0146] So far, this embodiment has obtained the corrected clustering radius of each candidate heart tissue region.

[0147] S6, combining the modified clustering radius, using a clustering algorithm to segment each frame of the cardiac ultrasound image, to obtain each frame of the cardiac ultrasound image with a marked artifact area.

[0148] In this embodiment, the DPC (Density Peaks Clustering) clustering algorithm is used to cluster the image in combination with the modified clustering radius. The modified clustering radius can be used to calculate the neighborhood range of the local density. The local density is calculated using the obtained modified clustering radius, and then the relative distance is obtained. A decision diagram is drawn according to the local density and the relative distance to obtain the cluster center, and other pixels are assigned to the closest cluster center to obtain different clusters. After the image is clustered, the cardiac tissue area and the artifact area can be segmented, and the cardiac tissue area and the artifact area are marked respectively, so as to obtain each frame of the cardiac ultrasound image with the marked artifact area, and the marked artifact area includes an independent artifact area and an area affected by the artifact. Among them, the implementation process of the DPC clustering algorithm is a prior art, which is not within the scope of protection of the present invention and will not be elaborated in detail here.

[0149] Specifically, an embodiment of the present invention also provides an artificial intelligence-based cardiac ultrasound image segmentation system, including a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement the artificial intelligence-based cardiac ultrasound image segmentation method as described above.

[0150] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A cardiac ultrasound image segmentation method based on artificial intelligence, characterized in that: The following steps are involved: Acquire a cardiac ultrasound image sequence by using cardiac ultrasound imaging technology, and then determine each candidate cardiac tissue region in each frame of the cardiac ultrasound image in the cardiac ultrasound image sequence; Determining a regional feature value of each candidate heart tissue region according to edge change characteristics and grayscale performance characteristics of each candidate heart tissue region; Analyze edge differences and distance differences of two candidate cardiac tissue regions located in different cardiac ultrasound images to determine first regional difference factors of the two candidate cardiac tissue regions; Determining a target region matching degree of each candidate cardiac tissue region according to the region feature value of each candidate cardiac tissue region and a plurality of the first region difference factors; determining an initial clustering radius of each candidate heart tissue region; Using the target region matching degree, correcting the initial clustering radius of each candidate heart tissue region to obtain a corrected clustering radius of each candidate heart tissue region; In combination with the modified clustering radius, a clustering algorithm is used to segment each frame of the cardiac ultrasound image to obtain each frame of the cardiac ultrasound image with an artifact region marked; Determining the regional feature value of each candidate cardiac tissue region according to the edge change feature and the grayscale representation feature of each candidate cardiac tissue region includes: For any candidate cardiac tissue region, edge detection is performed on the candidate cardiac tissue region to obtain edge pixel points of the candidate cardiac tissue region; and a first region characteristic factor of the candidate cardiac tissue region is determined according to a difference in gradient amplitude between two adjacent edge pixel points; Determine a grayscale performance index of each pixel point according to the grayscale value of each pixel point in the candidate heart tissue area; Determine a second region characteristic factor of the candidate heart tissue region according to an average value of the grayscale performance indexes of all pixels in the candidate heart tissue region and a difference in the grayscale performance indexes between the two adjacent edge pixels; Determine a regional characteristic value of the candidate cardiac tissue region by combining the first regional characteristic factor and the second regional characteristic factor of the candidate cardiac tissue region; Determining the grayscale performance index of each pixel point according to the grayscale value includes: For any pixel, calculate the grayscale difference between the pixel and any adjacent pixel; Determine the grayscale performance index of the pixel point by combining the grayscale value of the pixel point and the grayscale difference value; The grayscale value of the pixel point is positively correlated with the grayscale performance index, and the grayscale difference value is negatively correlated with the grayscale performance index. The step of analyzing edge differences and distance differences of two candidate cardiac tissue regions located in different cardiac ultrasound images to determine a first regional difference factor of the two candidate cardiac tissue regions includes: Taking a candidate cardiac tissue region in any cardiac ultrasound image as a region to be analyzed, and taking a candidate cardiac tissue region in cardiac ultrasound images other than the cardiac ultrasound image as a comparison region; For any area to be analyzed and any comparison area, determine the centroid position of the area to be analyzed and the comparison area and the position of each edge pixel point on the area boundary; According to the centroid position and the position of each edge pixel point on the region boundary, the average distance from the centroid of the region to be analyzed and the comparison region to all edge pixel points on the corresponding region boundary is calculated; Determining the degree of difference in distance between the area to be analyzed and the comparison area according to the difference between the distance mean of the area to be analyzed and the distance mean of the comparison area; Determine a first area difference factor between the area to be analyzed and the comparison area according to the distance difference degree, the centroid distance and the difference in the number of edge pixels of the area boundaries between the area to be analyzed and the comparison area; Determining the first area difference factor between the area to be analyzed and the comparison area according to the distance difference degree, the centroid distance and the difference in the number of edge pixels of the area boundaries between the area to be analyzed and the comparison area includes: Mapping the centroid of the comparison area to the area to be analyzed, and calculating the distance between the two centroids, which is recorded as the centroid distance; Counting the number of edge pixels on the boundary of the area to be analyzed and the area to be compared, and taking the difference between the two numbers of edge pixels as the difference in the number of edge pixels; The product of the distance difference degree, the centroid distance and the difference in the number of edge pixels is used as the first area difference factor between the area to be analyzed and the comparison area; The step of determining the target region matching degree of each candidate cardiac tissue region according to the region feature value of each candidate cardiac tissue region and a plurality of the first region difference factors includes: Determine a second regional difference factor between the region to be analyzed and the comparison region according to the regional characteristic value of the region to be analyzed and the difference between the regional characteristic values ​​of the region to be analyzed and the comparison region; Determine the degree of regional matching between the region to be analyzed and the comparison region by combining the first regional difference factor and the second regional difference factor of the region to be analyzed and the comparison region; Wherein, the first regional difference factor, the second regional difference factor and the regional matching degree are both negatively correlated; The region matching degree between the region to be analyzed and each comparison region is obtained, and the maximum value among all the region matching degrees is selected as the target region matching degree of the region to be analyzed.

2. The method for segmenting cardiac ultrasound images based on artificial intelligence according to claim 1, characterized in that: The determining of the second region characteristic factor of the candidate heart tissue region according to the average value of the grayscale performance index of all pixels in the candidate heart tissue region and the difference of the grayscale performance index between the two adjacent edge pixels comprises: Calculate a first ratio of the grayscale performance index between two adjacent edge pixels, and calculate the absolute value of the difference between 1 and the first ratio; calculate a second ratio of the average value of the grayscale performance index of all pixels in the candidate heart tissue area and the absolute value of the difference between 1 and the first ratio; add all of the second ratios to obtain a second area characteristic factor of the candidate heart tissue area.

3. The method for segmenting cardiac ultrasound images based on artificial intelligence according to claim 1, characterized in that: The determining, according to the regional characteristic value of the region to be analyzed and the difference between the regional characteristic values ​​of the region to be analyzed and the comparison region, a second regional difference factor between the region to be analyzed and the comparison region comprises: The absolute value of the difference between the regional characteristic value of the region to be analyzed and the regional characteristic value of the comparison region is calculated, and the ratio of the absolute value of the difference to the regional characteristic value of the region to be analyzed is used as the second regional difference factor between the region to be analyzed and the comparison region.

4. The method for segmenting cardiac ultrasound images based on artificial intelligence according to claim 1, characterized in that: The step of determining the initial clustering radius of each candidate cardiac tissue region comprises: For any candidate heart tissue region, calculating the difference between the maximum grayscale value and the minimum grayscale value in the candidate heart tissue region as the first clustering influence factor; Calculate the distance value between any two pixels at the same gray level, and then calculate the average value of all the distance values ​​at the same gray level as the distance mean, and take the average value of the distance means corresponding to all gray levels as the second clustering influence factor; Determine an initial clustering radius of the candidate heart tissue region by combining the first clustering influence factor and the second clustering influence factor; The first clustering influence factor, the second clustering influence factor and the initial clustering radius are negatively correlated.

5. An artificial intelligence-based cardiac ultrasound image segmentation system, characterized in that: It comprises a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement the artificial intelligence-based cardiac ultrasound image segmentation method as described in any one of claims 1 to 4.

Images