Coronary artery calcification score estimation method and device, equipment and storage medium

Abstract

The invention provides a coronary artery calcification score estimation method and device, equipment and a storage medium. The method comprises the steps that a to-be-evaluated target three-dimensional chest CT image is acquired; performing coronary artery calcification region segmentation on the target three-dimensional chest CT image to obtain a coronary artery calcification image corresponding to the target three-dimensional chest CT image, and performing coronary artery calcification classification on the target three-dimensional chest CT image to obtain a coronary artery calcification classification result corresponding to the target three-dimensional chest CT image, the coronary artery calcification classification result is used for indicating whether coronary artery calcification exists in the target three-dimensional chest CT image or not; and if the coronary artery calcification classification result is used for indicating that the target three-dimensional chest CT image has coronary artery calcification, performing coronary artery calcification score estimation based on the coronary artery calcification image to obtain a coronary artery calcification score of the target three-dimensional chest CT image. According to the technical scheme, the evaluation efficiency can be improved, and the accuracy of coronary artery calcification scoring is improved.

Description

Technical Field

[0001] The present application relates to the field of automatic scoring, and in particular to a coronary artery calcification scoring estimation method, apparatus, device and storage medium. Background Art

[0002] Coronary artery calcification is one of the main forms of coronary artery disease and one of the main risk factors for heart disease and stroke. Coronary artery calcification scoring is an important method for risk assessment of patients with coronary artery disease. However, due to the complex distribution and morphology of calcification, existing coronary artery calcification scoring methods often rely on the experience and knowledge of experts for detection and evaluation, and the scoring efficiency is low. Summary of the invention

[0003] The present application provides a coronary artery calcification scoring estimation method, apparatus, device and storage medium to solve the technical problem of low scoring efficiency caused by manual coronary artery calcification scoring.

[0004] In a first aspect, a method for estimating a coronary artery calcification score is provided, comprising:

[0005] Acquire a three-dimensional chest computed tomography (CT) image of the target to be evaluated;

[0006] performing coronary calcification region segmentation on the target three-dimensional chest CT image to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and performing coronary calcification classification on the target three-dimensional chest CT image to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image, wherein the coronary calcification classification result is used to indicate whether coronary calcification exists in the target three-dimensional chest CT image;

[0007] If the coronary calcification classification result is used to indicate that coronary calcification exists in the target three-dimensional chest CT image, a coronary calcification score is estimated based on the coronary calcification image to obtain a coronary calcification score of the target three-dimensional chest CT image.

[0008] In this technical solution, after obtaining a target three-dimensional chest CT image to be evaluated, the target three-dimensional chest CT image is segmented for coronary calcification regions to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and the target three-dimensional chest CT image is classified for coronary calcification to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image. If the coronary calcification classification result is used to indicate that coronary calcification exists in the target three-dimensional chest CT image, a coronary calcification score is estimated based on the coronary calcification image to obtain a coronary calcification score of the target three-dimensional chest CT image; that is, automatic scoring of coronary calcification is achieved by image region segmentation and image classification, without the need for manual detection and evaluation, thereby improving evaluation efficiency; in addition, when performing automatic scoring, the target three-dimensional chest CT image is also classified for coronary calcification to obtain a coronary calcification classification result. If the coronary calcification classification result is used to indicate that coronary calcification exists in the target three-dimensional chest CT image, a score estimation is performed based on the coronary calcification image, thereby avoiding the problem of inaccurate scoring caused by mis-segmentation of the coronary region due to noise, thereby improving the accuracy of the scoring.

[0009] In combination with the first aspect, in a possible implementation, the method further includes: if the coronary calcification classification result is used to indicate that the target three-dimensional chest CT image does not have coronary calcification, using a preset score as the coronary calcification score of the target three-dimensional chest CT image. In the case where the coronary calcification classification result is used to indicate that the target three-dimensional chest CT image does not have coronary calcification, using the preset score as the coronary calcification score of the target three-dimensional chest CT image can avoid the problem of inaccurate scoring caused by regional missegmentation caused by noise, thereby improving the accuracy of the scoring.

[0010] In combination with the first aspect, in a possible implementation, the segmentation of the target three-dimensional chest CT image for coronary calcification regions to obtain the coronary calcification image corresponding to the target three-dimensional chest CT image, and the classification of the coronary calcification of the target three-dimensional chest CT image to obtain the classification result of the coronary calcification corresponding to the target three-dimensional chest CT image include: inputting the target three-dimensional chest CT image into a target multi-task coronary calcification processing model, segmenting the target three-dimensional chest CT image for coronary calcification regions through the target multi-task coronary calcification processing model to obtain the coronary calcification image corresponding to the target three-dimensional chest CT image, and classifying the coronary calcification of the target three-dimensional chest CT image to obtain the classification result of the coronary calcification corresponding to the target three-dimensional chest CT image. Using the model to segment the target three-dimensional chest CT image for coronary calcification regions and classify coronary calcification can improve efficiency.

[0011] In combination with the first aspect, in a possible implementation, the target multi-task coronary calcification processing model includes an encoder, a decoder, a region segmentation module and a classifier; the target three-dimensional chest CT image is input into the target multi-task coronary calcification processing model, the target three-dimensional chest CT image is segmented into coronary calcification regions by the target multi-task coronary calcification processing model to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and the target three-dimensional chest CT image is classified for coronary calcification to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image, including: inputting the target three-dimensional chest CT image into the target multi-task coronary calcification processing model, performing coronary calcification region segmentation on the target three-dimensional chest CT image to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and performing coronary calcification classification on the target three-dimensional chest CT image to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image. The first feature map is input into the encoder, and the encoder is used to extract features of the target three-dimensional chest CT image to obtain the first feature map corresponding to the target three-dimensional chest CT image; the first feature map is input into the decoder, and the decoder is used to decode the first feature map to obtain a feature decoding map; the feature decoding map is input into the regional segmentation module, and the regional segmentation module is used to segment the feature decoding map into coronary calcification regions to obtain the coronary calcification image; the first feature map is input into the classifier, and the classifier is used to classify the first feature map into coronary calcification to obtain the coronary calcification classification result. Since the regional segmentation module and the classifier in the target multi-task coronary calcification processing model share the features extracted by the encoder, they can share information between the coronary calcification segmentation task and the coronary classification task, thereby improving the multi-task coronary calcification processing model's ability to discriminate coronary calcification, and improving the accuracy of coronary calcification regional segmentation and coronary calcification classification.

[0012] In combination with the first aspect, in a possible implementation, the first feature map is input into the decoder, and the first feature map is feature decoded by the decoder to obtain a feature decoding map, and further includes: downsampling the feature decoding map to obtain a second feature map, the size of the second feature map is the same as that of the first feature map; feature fusion of the first feature map and the second feature map to obtain a third feature map; the first feature map is input into the classifier, and the first feature map is classified by the classifier for coronary calcification to obtain the coronary calcification classification result, including: inputting the third feature map into the classifier, and the third feature map is classified by the classifier for coronary calcification to obtain the coronary calcification classification result. By downsampling the feature decoding map output by the decoder and fusing it with the first feature map output by the encoder before performing coronary calcification classification, context information can be increased, thereby improving the accuracy of coronary calcification classification.

[0013] In combination with the first aspect, in a possible implementation, the coronary calcification image includes a coronary calcification region and a region category corresponding to the coronary calcification region; the region segmentation module includes a first convolution block and a second convolution block; the inputting the feature decoding image into the region segmentation module, performing coronary calcification region segmentation on the feature decoding image through the region segmentation module, and obtaining the coronary calcification image includes: inputting the feature decoding image into the first convolution block, performing coronary calcification region segmentation on the feature decoding image through the first convolution block, and obtaining the coronary calcification region; inputting the feature decoding image into the second convolution block, and performing coronary calcification region category recognition on the feature decoding image through the second convolution block, and obtaining the region category. By identifying the coronary calcification region and the region category, different region categories can be scored to improve evaluation efficiency.

[0014] In combination with the first aspect, in a possible implementation method, the target three-dimensional chest CT image is input into a target multi-task coronary calcification processing model, the target three-dimensional chest CT image is segmented into coronary calcification areas by the target multi-task coronary calcification processing model to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and before the target three-dimensional chest CT image is classified for coronary calcification, it also includes: obtaining a sample three-dimensional chest CT image and a sample label corresponding to the sample three-dimensional chest CT image, the sample label including a coronary calcification area label, a region category label and a coronary calcification classification result label; and training the target multi-task coronary calcification processing model according to the sample three-dimensional chest CT image and the sample label corresponding to the sample three-dimensional chest CT image.

[0015] In combination with the first aspect, in a possible implementation, the training of the target multi-task coronary calcification processing model based on the sample three-dimensional chest CT image and the sample label corresponding to the sample three-dimensional chest CT image includes: inputting the sample three-dimensional chest CT image into the multi-task coronary calcification processing model to obtain the coronary calcification prediction area, coronary calcification prediction area category and coronary calcification classification prediction result output by the multi-task coronary calcification processing model; calculating a first loss based on the coronary calcification prediction area and the coronary calcification area label; calculating a second loss based on the coronary calcification prediction area category and the area category label; calculating a third loss based on the coronary calcification classification prediction result and the coronary calcification classification result label; determining the total loss of the multi-task coronary calcification processing model based on the first loss, the second loss and the third loss; adjusting the model parameters of the multi-task coronary calcification processing model based on the total loss so that the total loss is reduced, thereby obtaining the target multi-task coronary calcification processing model.

[0016] In a second aspect, a coronary artery calcification score estimation device is provided, comprising:

[0017] An image acquisition module, used for acquiring a target three-dimensional chest CT image to be evaluated;

[0018] a calcification processing module, configured to segment the target three-dimensional chest CT image into a coronary calcification region to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and to classify the coronary calcification of the target three-dimensional chest CT image to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image, wherein the coronary calcification classification result is used to indicate whether the target three-dimensional chest CT image has coronary calcification;

[0019] an estimation module, for estimating a coronary calcification score based on the coronary calcification image to obtain a coronary calcification score of the target three-dimensional chest CT image if the coronary calcification classification result indicates that coronary calcification exists in the target three-dimensional chest CT image.

[0020] In a third aspect, a computer device is provided, comprising a memory and one or more processors, wherein the memory is connected to the one or more processors, and the one or more processors are used to execute one or more computer programs stored in the memory, and when the one or more processors execute the one or more computer programs, the computer device implements the coronary calcification score estimation method of the first aspect.

[0021] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the coronary artery calcification score estimation method of the first aspect.

[0022] The present application can achieve the following technical effects: automatic scoring of coronary calcification is achieved through image region segmentation and image classification, without the need for manual detection and evaluation, which can improve evaluation efficiency; in addition, when performing automatic scoring, the target three-dimensional chest CT image is also classified for coronary calcification to obtain a coronary calcification classification result. When the coronary calcification classification result is used to indicate the presence of coronary calcification in the target three-dimensional chest CT image, a score estimation is performed based on the coronary calcification image, which can avoid the problem of inaccurate scoring caused by mis-segmentation of the coronary region due to noise, thereby improving the accuracy of the scoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments of the present application 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 drawings can be obtained based on these drawings without paying any creative labor.

[0024] Figure 1 A feasible technical solution for automated coronary artery calcification scoring provided in the embodiments of the present application;

[0025] Figure 2 A schematic diagram of a flow chart of a coronary artery calcification score estimation method provided in an embodiment of the present application;

[0026] Figure 3A-3B A schematic diagram of the composition structure of the target multi-task coronary calcification processing model provided in an embodiment of the present application;

[0027] Figure 4 A flowchart of a training method for a multi-task coronary calcification processing model provided in an embodiment of the present application;

[0028] Figure 5 is a schematic diagram of the structure of a coronary artery calcification scoring estimation device provided in an embodiment of the present application;

[0029] Figure 6 It is a structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions 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 intended to limit the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0031] It should be noted that, if there is no conflict, the various features in the embodiments of the present application can be combined with each other, all within the scope of protection of the present application. In addition, although the functional module division is performed in the device schematic diagram and the logical order is shown in the flow chart, in some cases, the steps shown or described can be performed in a sequence different from the module division in the device or the flow chart. Furthermore, the words "first", "second", "third", etc. used in this application do not limit the data and execution order, but only distinguish the same items or similar items with basically the same functions and effects.

[0032] The technical solution of this application is applicable to the scenario of automated coronary calcification scoring. Automated coronary calcification scoring refers to coronary calcification scoring of chest CT images based on computer technology. A feasible technical solution for automated coronary calcification scoring can be as follows: Figure 1 As shown, the encoder is first used to extract the coronary calcification features of the chest plain scan CT image, and then the decoder is used to output the coronary calcification image, which includes the coronary calcification area. Finally, the coronary calcification score is performed based on the coronary calcification area in the coronary calcification image, and the corresponding coronary calcification score is output.

[0033] When the coronary calcification area in the coronary calcification image is small, based on Figure 1 The technical solution shown may miss segmentation when segmenting the coronary calcification area through the encoder, that is, the coronary calcification area is not segmented from the chest plain scan CT image; and, since there may be noise, calcification of other tissues and organs, or coronary calcification artifacts in the chest plain scan CT image, these areas may be mistakenly segmented as coronary calcification areas through the encoder, resulting in inaccurate subsequent coronary calcification scores.

[0034] In view of this, the present application proposes a new automatic coronary calcification scoring scheme. On the basis of coronary calcification segmentation, the results of coronary calcification classification are introduced to correct the coronary calcification score, thereby reducing the error of the coronary calcification score and improving the accuracy of the coronary calcification score. The technical solution of the present application is specifically introduced below.

[0035] The technical solution of the present application can be applied to any computer device with image processing function.

[0036] See also Figure 2 , Figure 2 A schematic diagram of a coronary artery calcification score estimation method provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the method comprises the following steps:

[0037] S101, obtaining a target three-dimensional chest CT image to be evaluated.

[0038] Here, the target three-dimensional chest CT image refers to a chest scan CT image in three-dimensional form that needs to be scored for coronary artery calcium. The target three-dimensional chest CT image can be represented by X c1×d1×h1×w1 , c1 is the number of channels of the target three-dimensional chest CT image, c1 is 1, d1 represents the number of axial slices contained in the target three-dimensional chest CT image, that is, the number of two-dimensional images obtained by dividing the target three-dimensional chest CT image into two-dimensional images, h1 and w1 represent the height and width of the two-dimensional image obtained by dividing the target two-dimensional chest CT image into two-dimensional images.

[0039] The target three-dimensional chest CT image to be evaluated can be obtained from a local computer device (such as a computer device that scans and generates a target three-dimensional chest CT image), a chest scanning device, or the cloud.

[0040] S102, performing coronary calcification region segmentation on the target three-dimensional chest CT image to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and performing coronary calcification classification on the target three-dimensional chest CT image to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image.

[0041] Here, performing coronary calcification segmentation on the target three-dimensional chest CT image refers to identifying the coronary calcification area of ​​the target three-dimensional chest CT image through image segmentation technology, thereby identifying the coronary calcification area in the target three-dimensional chest CT image, so that the coronary calcification image corresponding to the target three-dimensional chest CT image includes the coronary calcification area. The coronary calcification image corresponding to the target three-dimensional chest CT image may also include the regional category of the coronary calcification area. The distinction category of the coronary calcification area is used to indicate which specific area of ​​the coronary artery has coronary calcification. The regional category of the coronary calcification area can be divided into four types: left main coronary artery, left circumflex artery, left anterior descending artery and right coronary artery. The regional category of the coronary calcification area can be obtained by image recognition technology by distinguishing the coronary calcification in the target three-dimensional chest CT image.

[0042] Classifying the target three-dimensional chest CT image for coronary artery calcification refers to identifying whether the target three-dimensional chest CT image has coronary artery calcification through image classification technology, thereby determining whether the target three-dimensional chest CT image is an image of coronary artery calcification, that is, whether the target three-dimensional chest CT image has coronary artery calcification. That is, the coronary artery calcification classification result corresponding to the target three-dimensional chest CT image is used to indicate whether the target three-dimensional chest CT image has coronary artery calcification.

[0043] In some possible situations, the target three-dimensional chest CT image can be segmented and classified for coronary calcification regions by a deep learning model. The target three-dimensional chest CT image can be input into a target multi-task coronary calcification processing model, and the target three-dimensional chest CT image can be segmented for coronary calcification regions by the target multi-task coronary calcification processing model to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and the target three-dimensional chest CT image can be classified for coronary calcification to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image.

[0044] In one possible implementation, the target multi-task coronary calcification processing model can be as follows: Figure 3AAs shown, the target multi-task coronary calcification processing model may include a coronary calcification region segmentation model and a coronary calcification classification model. The coronary calcification region segmentation model and the coronary calcification classification model are independent of each other. After the target three-dimensional chest CT image is input into the target multi-task coronary calcification processing model, the coronary region segmentation and coronary region recognition of the target three-dimensional chest CT image can be performed by the coronary calcification region segmentation model to obtain the coronary calcification image corresponding to the target three-dimensional chest CT image; and the coronary calcification classification model is used to classify the coronary calcification of the target three-dimensional chest CT image to obtain the coronary calcification classification result corresponding to the target three-dimensional chest CT image. The coronary calcification region segmentation model includes but is not limited to a Segnet model, a Deeplab model, a Mask R-CNN model, a Gated SCNN model or a fully convolutional neural network model. The coronary calcification classification model includes but is not limited to an ImageNet model, an AlexNet model, a VGG model, etc.

[0045] In another feasible implementation, the target multi-task coronary artery calcification model may also be a model that associates coronary artery region segmentation with coronary artery calcification classification.

[0046] In one embodiment, the target multi-task coronary calcification processing model can be as follows Figure 3B As shown, the target multi-task coronary calcification processing model includes an encoder, a decoder, a regional segmentation module and a classifier. After the target three-dimensional chest CT image is input into the target multi-task coronary calcification processing model, the encoder can be used to extract features of the target three-dimensional chest CT image to obtain a first feature map corresponding to the target three-dimensional chest CT image; then the first feature map output by the encoder is input into the decoder, and the first feature map is feature decoded by the decoder to obtain a feature decoding map; then the feature decoding map output by the decoder is input into the regional segmentation module, and the feature decoding map is segmented into coronary calcification regions by the regional segmentation module to obtain a coronary calcification image; and the first feature map output by the encoder is input into the classifier, and the first feature map is classified by the classifier to obtain a coronary calcification classification result.

[0047] The encoder may be an encoder of any structure, for example, an encoder network in a ResNet model, an encoder network in a U-Net model, etc. After the target three-dimensional chest CT image is input to the encoder, the first feature map output by the encoder can be represented as F c2×d2×h2×w2, c2 represents the number of channels output by the last layer of the encoder, d2 = d1 / n, n is the number of times the encoder downsamples the target three-dimensional chest CT image, h2 and w2 are the height and width of the first feature map, h2≤h1, w≤w1. Since the regional segmentation module and classifier in the target multi-task coronary calcification processing model share the features extracted by the encoder, they can share information between the coronary calcification segmentation task and the coronary classification task, thereby improving the multi-task coronary calcification processing model's ability to discriminate coronary calcification, and improving the accuracy of coronary calcification regional segmentation and coronary calcification classification.

[0048] The decoder can be a decoder of any structure, and the decoder can be, for example, an encoding network in a U-Net model. After the first feature map output by the encoder is input into the decoder, the decoded feature map output by the decoder can be expressed as P c3×d3×h3×w3 , c3=c1, represents the number of channels of the last layer output of the encoder, d2=d1, represents the number of axial slices of the decoded feature map, h2 and w2 are the height and width of the decoded feature map, h2=h1, w3=w1.

[0049] The regional segmentation module includes a first convolution block and a second convolution block. After the feature decoding image output by the decoder is input into the regional segmentation module, the first convolution block is used to segment the coronary calcification region of the feature decoding image to obtain the coronary calcification region in the coronary calcification image; the second convolution block is used to decode the feature decoding image and identify the coronary calcification region category to obtain the region category corresponding to the coronary calcification region in the coronary calcification image. By identifying the coronary calcification region and region category, different region categories can be scored to improve the evaluation efficiency.

[0050] The classifier may be composed of a global average pooling layer and a fully connected layer. After the first feature map output by the encoder is input into the classifier, the first feature map is first pooled by the global average pooling layer to obtain a pooled feature map; the pooled feature map is then processed by the fully connected layer to output the probability of the presence of coronary calcification in the target three-dimensional chest CT image and the probability of the absence of coronary calcification in the target three-dimensional chest CT image, thereby obtaining a coronary calcification classification result. If the probability of the presence of coronary calcification in the target three-dimensional chest CT image is greater than the probability of the absence of coronary calcification in the target three-dimensional chest CT image, the coronary calcification classification result is used to indicate that the target three-dimensional chest CT image has coronary calcification; if the probability of the presence of coronary calcification in the target three-dimensional chest CT image is less than the probability of the absence of coronary calcification in the target three-dimensional chest CT image, the coronary calcification classification result is used to indicate that the target three-dimensional chest CT image does not have coronary calcification.

[0051] Optionally, in the process of inputting the first feature map output by the encoder into the classifier, classifying the first feature map by the classifier to obtain the coronary calcification classification result, the feature decoding map output by the decoder can also be downsampled to obtain a second feature map. The size of the second feature map is the same as that of the first feature map. The second feature map can be expressed as T c4×d4×h4×w4 , c4=c2, d4=d2, h4=h2, w4=w2; feature fusion is performed on the first feature map and the second feature map to obtain a third feature map; the third feature map is then input into the classifier, and the third feature map is classified for coronary calcification by the classifier to obtain a coronary calcification classification result. Among them, the first feature map and the second feature map can be feature fused by splicing to obtain the third feature map. After obtaining the third feature map, the first feature map can be pooled by a global average pooling layer to obtain a pooled feature map; the pooled feature map is then processed by a fully connected layer to output the probability of coronary calcification in the target three-dimensional chest CT image and the probability of coronary calcification in the target three-dimensional chest CT image, thereby obtaining a coronary calcification classification result. By downsampling the feature decoding map output by the decoder and then fusing it with the first feature map output by the encoder before performing coronary calcification classification, context information can be increased, thereby improving the accuracy of coronary calcification classification.

[0052] In some other possible situations, other methods can also be used to segment the coronary calcification region and classify the coronary calcification of the target three-dimensional chest CT image. For example, the coronary calcification region of the target three-dimensional chest CT image can be segmented by traditional image processing methods such as region growing method, region classification merge method (RSM), contour fitting method, histogram method, etc., to obtain the coronary calcification image corresponding to the target three-dimensional chest CT image. The coronary calcification of the target three-dimensional chest CT image can also be classified by classifiers such as support vector machine (SVM) and decision tree (DT), to obtain the coronary calcification classification result corresponding to the target three-dimensional chest CT image.

[0053] S103: If the coronary calcification classification result corresponding to the target three-dimensional chest CT image indicates that coronary calcification exists in the target three-dimensional chest CT image, a coronary calcification score is estimated based on the coronary calcification image corresponding to the target three-dimensional chest CT image to obtain a coronary calcification score of the target three-dimensional chest CT image.

[0054] Here, the target coronary calcification image includes a coronary calcification region, and the area of ​​the coronary calcification region in the coronary calcification image can be counted as the coronary calcification score of the target three-dimensional chest CT image; or, the ratio between the area of ​​the coronary calcification image in the coronary calcification image and the area of ​​the coronary calcification image can be calculated, and the product of the ratio and the preset total score can be used as the coronary calcification score of the target three-dimensional chest CT image. This application does not limit the specific scoring method.

[0055] If the coronary calcification classification result corresponding to the target three-dimensional chest CT image is used to indicate the presence of coronary calcification in the target three-dimensional chest CT image, the pixel value of each pixel in the coronary calcification image corresponding to the target three-dimensional chest CT image can be multiplied by the first indication value used to indicate the presence of coronary calcification in the target three-dimensional chest CT image, so as to correct the coronary calcification image corresponding to the target three-dimensional chest CT image, thereby performing a coronary calcification score estimation based on the coronary calcification image corresponding to the target three-dimensional chest CT image, and obtaining the coronary calcification score of the target three-dimensional chest CT image. The first indication value can be 1. Multiplying each pixel in the coronary calcification image corresponding to the target three-dimensional chest CT image by the first indication value will not change the pixel value of each pixel in the coronary calcification image corresponding to the target three-dimensional chest CT image. Scoring based on the corrected coronary calcification image is essentially to estimate the coronary calcification score of the coronary calcification image, and obtain the coronary calcification score of the target three-dimensional chest CT image.

[0056] S104: If the coronary artery calcification classification result corresponding to the target three-dimensional chest CT image indicates that the target three-dimensional chest CT image does not have coronary artery calcification, a preset score is used as the coronary artery calcification score of the target three-dimensional chest CT image.

[0057] Here, the preset score may be 0.

[0058] If the coronary calcification classification result corresponding to the target three-dimensional chest CT image is used to indicate that the target three-dimensional chest CT image does not have coronary calcification, the pixel value of each pixel in the coronary calcification image corresponding to the target three-dimensional chest CT image can be multiplied by the second indication value used to indicate that the target three-dimensional chest CT image does not have coronary calcification, so as to correct the coronary calcification image corresponding to the target three-dimensional chest CT image, thereby using the preset score as the coronary calcification score of the target chest CT image. The second indication value can be 0. Multiplying each pixel in the coronary calcification image corresponding to the target three-dimensional chest CT image by the second indication value will make the pixel value of each pixel in the coronary calcification image corresponding to the target three-dimensional chest CT image 0, so that when scoring is performed based on the corrected coronary calcification image, the coronary calcification score estimated by the score is 0.

[0059] In the above Figure 2In the corresponding technical solution, after obtaining the target three-dimensional chest CT image to be evaluated, the coronary calcification region of the target three-dimensional chest CT image is segmented to obtain the coronary calcification image corresponding to the target three-dimensional chest CT image, and the coronary calcification classification result of the target three-dimensional chest CT image is obtained. If the coronary calcification classification result is used to indicate that the target three-dimensional chest CT image has coronary calcification, the coronary calcification score is estimated based on the coronary calcification image to obtain the coronary calcification score of the target three-dimensional chest CT image; that is, the automatic scoring of coronary calcification is achieved by image region segmentation and image classification, without the need for manual detection and evaluation, which can improve the evaluation efficiency; in addition, when performing automatic scoring, the coronary calcification classification of the target three-dimensional chest CT image is also performed to obtain the coronary calcification classification result. When the coronary calcification classification result is used to indicate that the target three-dimensional chest CT image has coronary calcification, the score estimation is performed based on the coronary calcification image, which can avoid the problem of inaccurate scoring caused by mis-segmentation of the coronary region due to noise, thereby improving the accuracy of the scoring.

[0060] In the case of using a deep learning model to perform coronary artery region segmentation and coronary calcification classification on a target three-dimensional chest CT image, the target three-dimensional chest CT image is input into a target multi-task coronary calcification processing model, the target three-dimensional chest CT image is segmented into coronary artery calcification regions by the target multi-task coronary calcification processing model to obtain a coronary artery calcification image corresponding to the target three-dimensional chest CT image, and the target three-dimensional chest CT image is classified for coronary artery calcification. Before obtaining the coronary artery calcification classification result corresponding to the target three-dimensional chest CT image, it is necessary to train the target multi-task coronary calcification processing model. The specific composition structure of the target multi-task coronary calcification processing model is different, so the method of training the target multi-task coronary calcification processing model is different. The following introduces the training method. Figure 3B The method of the target multi-task coronary processing model is shown.

[0061] See also Figure 4 , Figure 4 A flowchart of a training method for a multi-task coronary calcification processing model provided in an embodiment of the present application is shown in FIG. Figure 4 As shown, the method comprises the following steps:

[0062] S201, obtaining a sample three-dimensional chest CT image and a sample label corresponding to the sample three-dimensional chest CT image.

[0063] Here, the sample three-dimensional chest CT image refers to a three-dimensional chest CT image used as a training sample, the sample three-dimensional chest CT image refers to a training label pre-calibrated for the sample three-dimensional chest CT image, and the sample label corresponding to the sample three-dimensional chest CT image includes a coronary calcification area label, an area category label, and a coronary classification result label. The coronary calcification area label is used to indicate the coronary calcification area in the sample three-dimensional chest CT image. The coronary calcification area label may include the probability that each pixel in the sample three-dimensional chest CT image belongs to a background element and a calcification element. The coronary calcification area label can be expressed as P f1 The regional category label is used to indicate the regional category of the coronary artery calcification region in the sample three-dimensional chest CT image. The regional category label can be expressed as P k1 , the regional category label may include the regional category of each pixel in the sample three-dimensional chest CT image, and the regional category label may be presented in a one-hot encoding manner. The coronary calcification classification result label is used to indicate whether there is coronary calcification in the sample three-dimensional chest CT image, and the coronary calcification classification result label may be represented as y, and the coronary calcification classification result label may also be presented in a one-hot encoding manner.

[0064] S202, training a target multi-task coronary artery calcification processing model based on the sample three-dimensional chest CT image and the sample label corresponding to the sample three-dimensional chest CT image.

[0065] Among them, the target multi-task coronary artery calcification processing model can be obtained through training through the following steps A1-A6.

[0066] A1. Input the sample three-dimensional chest CT image into the multi-task coronary calcification processing model to obtain the coronary calcification prediction area, coronary calcification prediction area category and coronary calcification classification prediction results output by the multi-task coronary calcification processing model.

[0067] The multi-task coronary calcification processing model here is an untrained multi-task coronary calcification processing model. The structure of the multi-task coronary calcification processing model can be found in Figure 3B .

[0068] The coronary calcification prediction region refers to the coronary calcification region in the sample three-dimensional chest CT image output by the multi-task coronary calcification processing model performing coronary calcification region segmentation on the sample three-dimensional chest CT image; the coronary calcification prediction region category refers to the region category of the coronary calcification region in the sample three-dimensional chest CT image output by the multi-task coronary calcification processing model performing coronary calcification region category identification on the sample three-dimensional chest CT image; the coronary calcification classification prediction result refers to the coronary calcification classification result corresponding to the sample three-dimensional chest CT image output by the multi-task coronary calcification processing model performing coronary calcification classification on the sample three-dimensional chest CT model. The specific implementation method of the multi-task coronary calcification processing model outputting the coronary calcification region in the sample three-dimensional chest CT image, the region category of the coronary calcification region in the sample three-dimensional chest CT image, and the coronary calcification classification result in the sample three-dimensional chest CT image is the same as that in the aforementioned step S102 based on Figure 3B Similarly, the multi-task coronary calcification processing model shown outputs the coronary calcification area in the target three-dimensional chest CT image, the regional category of the coronary calcification area in the target three-dimensional chest CT image, and the coronary calcification classification result in the target three-dimensional chest CT image. Please refer to the description of the aforementioned step S102, which will not be repeated here.

[0069] A2. Calculate the first loss based on the coronary calcification prediction area and the coronary calcification area label output by the multi-task coronary calcification processing model.

[0070] Here, the first loss is used to reflect the gap between the coronary calcification prediction area and the coronary calcification area label. After the sample three-dimensional chest CT image is input into the multi-task coronary calcification processing model, the multi-task coronary calcification processing model can output the probability of each pixel in the sample three-dimensional chest CT image belonging to the background element and the probability of belonging to the calcified element. The sum of the probability of the pixel belonging to the background element and the probability of the pixel belonging to the calcified element is 1; if the probability of a pixel belonging to a calcified element is greater than the probability of belonging to the background element, then the pixel belongs to the calcified element; if the probability of a pixel belonging to a calcified element is less than the probability of belonging to the background element, then the pixel belongs to the background element; the area composed of calcified elements in the sample three-dimensional chest CT image is the coronary calcification prediction area.

[0071] In a feasible implementation, the first loss may be calculated based on a Dice loss function. The formula for calculating the first loss based on the Dice loss function is as follows:

[0072]

[0073] Among them, L1 represents the first loss, P f2 represents the predicted area of ​​coronary artery calcification, |P f1 ∩P f2| represents the number of intersection elements between the coronary calcification region label and the coronary calcification prediction region, |P f1 | represents the number of elements in the coronary calcification region label, |P f2 | represents the number of elements in the coronary artery calcification prediction area.

[0074] Optionally, the first loss may also be calculated based on a loss function such as an intersection over union (IOU) loss function, which is not limited in this application.

[0075] A3. Calculate the second loss based on the coronary calcification prediction region category and region category label output by the multi-task coronary calcification processing model.

[0076] Here, the second loss is used to reflect the gap between the predicted regional category of coronary calcification and the regional category label. After the sample three-dimensional chest CT image is input into the multi-task coronary calcification processing model, the multi-task coronary calcification processing model can output the probability that each element in the sample three-dimensional chest CT image belongs to various regional categories as the predicted regional category of coronary calcification, and the sum of the probabilities of each element belonging to various regional categories is 1.

[0077] In a feasible implementation, the second loss may be calculated based on a cross entropy loss function. The formula for calculating the second loss based on the cross entropy loss function is as follows:

[0078]

[0079] Where L2 represents the second loss, D represents the total number of regional categories of coronary artery calcification areas, and P k2 Represents the predicted regional category of coronary artery calcification.

[0080] Optionally, the second loss may also be calculated based on a loss function such as a Euclidean distance loss function, which is not limited in this application.

[0081] A4. Calculate the third loss based on the coronary calcification classification prediction results and coronary calcification classification result labels output by the multi-task coronary calcification processing model.

[0082] Here, the third loss is used to reflect the gap between the coronary calcification classification prediction result and the coronary calcification classification result label. After the sample three-dimensional chest CT image is input into the multi-task coronary calcification processing model, the multi-task coronary calcification processing model can output the coronary calcification probability of the sample three-dimensional chest CT image as the coronary calcification classification prediction result.

[0083] In a feasible implementation, the third loss may be calculated based on a cross entropy loss function. The formula for calculating the third loss based on the cross entropy loss function is as follows:

[0084] L3=-∑y i .log(p i )+(1-y i )log(1-p i )

[0085] Among them, L3 represents the third loss, p i represents the classification prediction result of coronary artery calcification, y i Represents the coronary artery calcification classification result label.

[0086] Optionally, the third loss may also be calculated based on a loss function such as a Euclidean distance loss function, which is not limited in this application.

[0087] A5. Determine the total loss of the multi-task coronary calcification processing model based on the first loss, the second loss and the third loss.

[0088] In a feasible implementation, the first loss, the second loss, and the third loss may be summed to obtain the total loss of the multi-task coronary artery calcification processing model. The calculation formula of the total loss is as follows:

[0089] L4=L1+L2+L3.

[0090] Among them, L4 represents the total loss.

[0091] In another feasible implementation, the first loss, the second loss and the third loss may be weighted and summed to obtain the total loss of the multi-task coronary artery calcification processing model. The calculation formula of the total loss is as follows:

[0092] L4=a1*(b1*L1+b2*L2)+a2*L3

[0093] Among them, a1, a2, b1, and b2 represent weighting coefficients.

[0094] A6. According to the total loss of the multi-task coronary calcification processing model, adjust the model parameters of the multi-task coronary calcification processing model so that the total loss of the multi-task coronary calcification processing model is reduced, and the target multi-task coronary calcification processing model is obtained.

[0095] Here, adjusting the model parameters of the multi-task coronary calcification processing model refers to adjusting Figure 3B The parameters of the encoder, decoder, classifier and region segmentation module are shown. The model parameters of the multi-task coronary artery calcification processing model can be adjusted based on the stochastic gradient descent method until the total loss of the multi-task coronary artery calcification processing model is less than the preset loss threshold, or the number of parameter adjustments reaches the preset number, and the multi-task coronary artery calcification processing model obtained by the last adjustment is determined as the target multi-task coronary artery calcification processing model.

[0096] In the above Figure 4 In the corresponding technical scheme, a target multi-task coronary calcification processing model is trained by obtaining a sample three-dimensional chest CT image and a sample label corresponding to the sample three-dimensional chest CT image, and then training the target multi-task coronary calcification processing model according to the sample three-dimensional chest CT image and the sample label corresponding to the sample three-dimensional chest CT image. Since the sample label corresponding to the sample three-dimensional chest CT image includes a coronary calcification area label, an area category label and a coronary classification result label, the multi-task coronary calcification processing model can learn the relationship between image features and coronary calcification areas, area categories of coronary calcification areas and coronary calcification categories, thereby obtaining a multi-task coronary calcification processing model that can accurately perform coronary calcification segmentation and coronary calcification classification.

[0097] It should be noted that, in each of the above-mentioned embodiments, there is not necessarily a certain order between the above-mentioned steps. A person skilled in the art can understand, based on the description of the embodiments of the present application, that in different embodiments, the above-mentioned steps may have different execution orders, that is, they may be executed in parallel, may be executed interchangeably, and so on.

[0098] The method of the present application is introduced above, and the device of the present application is introduced below.

[0099] See also Figure 5 , Figure 5 Schematic diagram of a coronary artery calcification scoring estimation device provided in an embodiment of the present application. Figure 5 As shown, the coronary artery calcification score estimation device 30 includes:

[0100] An image acquisition module 301 is used to acquire a target three-dimensional chest CT image to be evaluated;

[0101] The calcification processing module 302 is used to perform coronary calcification region segmentation on the target three-dimensional chest CT image to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and to perform coronary calcification classification on the target three-dimensional chest CT image to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image, wherein the coronary calcification classification result is used to indicate whether there is coronary calcification in the target three-dimensional chest CT image;

[0102] The estimation module 303 is used to estimate the coronary calcification score based on the coronary calcification image to obtain the coronary calcification score of the target three-dimensional chest CT image if the coronary calcification classification result indicates that coronary calcification exists in the target three-dimensional chest CT image.

[0103] In a possible design, the estimation module 303 is further used to: if the coronary calcification classification result is used to indicate that there is no coronary calcification in the target three-dimensional chest CT image, use a preset score as the coronary calcification score of the target three-dimensional chest CT image.

[0104] In one possible design, the calcification processing module 302 is specifically used to: input the target three-dimensional chest CT image into a target multi-task coronary calcification processing model, perform coronary calcification area segmentation on the target three-dimensional chest CT image through the target multi-task coronary calcification processing model to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and perform coronary calcification classification on the target three-dimensional chest CT image to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image.

[0105] In one possible design, the target multi-task coronary calcification processing model includes an encoder, a decoder, a regional segmentation module and a classifier; the above-mentioned calcification processing module 302 is specifically used to: input the target three-dimensional chest CT image into the encoder, and perform feature extraction on the target three-dimensional chest CT image through the encoder to obtain a first feature map corresponding to the target three-dimensional chest CT image; input the first feature map into the decoder, and perform feature decoding on the first feature map through the decoder to obtain a feature decoding map; input the feature decoding map into the regional segmentation module, and perform coronary calcification region segmentation on the feature decoding map through the regional segmentation module to obtain the coronary calcification image; input the first feature map into the classifier, and perform coronary calcification classification on the first feature map through the classifier to obtain the coronary calcification classification result.

[0106] In a possible design, the calcification processing module 302 is further used to: downsample the feature decoding image to obtain a second feature image, the size of the second feature image is the same as that of the first feature image; perform feature fusion on the first feature image and the second feature image to obtain a third feature image; the calcification processing module 302 is specifically used to: input the third feature image into the classifier, and classify the third feature image for coronary calcification through the classifier to obtain the coronary calcification classification result.

[0107] In a possible design, the coronary calcification image includes a coronary calcification area and a regional category corresponding to the coronary calcification area; the regional segmentation module includes a first convolution block and a second convolution block; the above-mentioned calcification processing module 302 is specifically used to: input the feature decoding image into the first convolution block, perform coronary calcification area segmentation on the feature decoding image through the first convolution block to obtain the coronary calcification area; input the feature decoding image into the second convolution block, perform coronary calcification area category identification on the feature decoding image through the second convolution block to obtain the region category.

[0108] In a possible design, the above-mentioned coronary calcification score estimation device 30 also includes a model training module 304, which is used to obtain a sample three-dimensional chest CT image and a sample label corresponding to the sample three-dimensional chest CT image, and the sample label includes a coronary calcification area label, a regional category label and a coronary calcification classification result label; according to the sample three-dimensional chest CT image and the sample label corresponding to the sample three-dimensional chest CT image, the target multi-task coronary calcification processing model is trained.

[0109] In a possible design, the above-mentioned model training module 304 is specifically used to: input the sample three-dimensional chest CT image into the multi-task coronary calcification processing model, obtain the coronary calcification prediction area, coronary calcification prediction area category and coronary calcification classification prediction result output by the multi-task coronary calcification processing model; calculate the first loss according to the coronary calcification prediction area and the coronary calcification area label; calculate the second loss according to the coronary calcification prediction area category and the area category label; calculate the third loss according to the coronary calcification classification prediction result and the coronary calcification classification result label; determine the total loss of the multi-task coronary calcification processing model according to the first loss, the second loss and the third loss; adjust the model parameters of the multi-task coronary calcification processing model according to the total loss so that the total loss is reduced, and obtain the target multi-task coronary calcification processing model.

[0110] It should be noted that Figure 5 For the contents not mentioned in the corresponding embodiments, please refer to the description of the aforementioned method embodiments, which will not be repeated here.

[0111] The above-mentioned device, after acquiring the target three-dimensional chest CT image to be evaluated, performs coronary calcification region segmentation on the target three-dimensional chest CT image to obtain the coronary calcification image corresponding to the target three-dimensional chest CT image, and performs coronary calcification classification on the target three-dimensional chest CT image to obtain the coronary calcification classification result corresponding to the target three-dimensional chest CT image. If the coronary calcification classification result is used to indicate that the target three-dimensional chest CT image has coronary calcification, the coronary calcification score is estimated based on the coronary calcification image to obtain the coronary calcification score of the target three-dimensional chest CT image; that is, the automatic scoring of coronary calcification is achieved by image region segmentation and image classification, without the need for manual detection and evaluation, which can improve the evaluation efficiency; in addition, when performing automatic scoring, the coronary calcification is also classified on the target three-dimensional chest CT image to obtain the coronary calcification classification result. When the coronary calcification classification result is used to indicate that the target three-dimensional chest CT image has coronary calcification, the score estimation is performed based on the coronary calcification image, which can avoid the problem of inaccurate scoring caused by mis-segmentation of the coronary region caused by noise, thereby improving the accuracy of the scoring.

[0112] See also Figure 6 , Figure 6 4 is a schematic diagram of a computer device provided in an embodiment of the present application, wherein the computer device 40 includes a processor 401 and a memory 402. The memory 402 is connected to the processor 401, for example, via a bus.

[0113] The processor 401 is configured to support the computer device 40 to perform the corresponding functions in the method in the above method embodiment. The processor 401 can be a central processing unit (CPU), a network processor (NP), a hardware chip or any combination thereof. The above hardware chip can be an application specific integrated circuit (ASIC), a programmable logic device (PLD) or a combination thereof. The above PLD can be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL) or any combination thereof.

[0114] The memory 402 is used to store program codes, etc. The memory 402 may include a volatile memory (VM), such as a random access memory (RAM); the memory 402 may also include a non-volatile memory (NVM), such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD); the memory 402 may also include a combination of the above-mentioned types of memories.

[0115] The processor 401 may call the program code to perform the following operations:

[0116] Acquire a three-dimensional chest CT image of a target to be evaluated;

[0117] performing coronary calcification region segmentation on the target three-dimensional chest CT image to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and performing coronary calcification classification on the target three-dimensional chest CT image to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image, wherein the coronary calcification classification result is used to indicate whether coronary calcification exists in the target three-dimensional chest CT image;

[0118] If the coronary calcification classification result is used to indicate that coronary calcification exists in the target three-dimensional chest CT image, a coronary calcification score is estimated based on the coronary calcification image to obtain a coronary calcification score of the target three-dimensional chest CT image.

[0119] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a computer, the computer executes the method described in the above embodiment.

[0120] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0121] The above disclosure is only the preferred embodiment of the present application, which certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A method for estimating coronary artery calcification score, characterized in that: include: Acquire a three-dimensional chest CT image of a target to be evaluated; performing coronary calcification region segmentation on the target three-dimensional chest CT image to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and performing coronary calcification classification on the target three-dimensional chest CT image to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image, wherein the coronary calcification classification result is used to indicate whether coronary calcification exists in the target three-dimensional chest CT image; If the coronary calcification classification result is used to indicate that coronary calcification exists in the target three-dimensional chest CT image, a coronary calcification score is estimated based on the coronary calcification image to obtain a coronary calcification score of the target three-dimensional chest CT image.

2. The method according to claim 1, characterized in that The method further comprises: If the coronary artery calcification classification result is used to indicate that there is no coronary artery calcification in the target three-dimensional chest CT image, a preset score is used as the coronary artery calcification score of the target three-dimensional chest CT image.

3. The method according to claim 1, characterized in that The segmenting of the target three-dimensional chest CT image to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and classifying the coronary calcification of the target three-dimensional chest CT image to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image include: The target three-dimensional chest CT image is input into a target multi-task coronary calcification processing model, and the target three-dimensional chest CT image is segmented into coronary calcification areas by the target multi-task coronary calcification processing model to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and the target three-dimensional chest CT image is classified for coronary calcification to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image.

4. The method according to claim 3, characterized in that The target multi-task coronary artery calcification processing model includes an encoder, a decoder, a region segmentation module and a classifier; The target three-dimensional chest CT image is input into a target multi-task coronary calcification processing model, the target three-dimensional chest CT image is segmented into a coronary calcification region by the target multi-task coronary calcification processing model to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and the target three-dimensional chest CT image is classified for coronary calcification to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image, including: Inputting the target three-dimensional chest CT image into the encoder, and extracting features of the target three-dimensional chest CT image through the encoder to obtain a first feature map corresponding to the target three-dimensional chest CT image; Inputting the first feature map into the decoder, and performing feature decoding on the first feature map by the decoder to obtain a feature decoding map; Inputting the characteristic decoding image into the region segmentation module, and performing coronary artery calcification region segmentation on the characteristic decoding image by the region segmentation module to obtain the coronary artery calcification image; The first feature map is input into the classifier, and the classifier is used to classify the first feature map for coronary artery calcification to obtain the coronary artery calcification classification result.

5. The method according to claim 4, characterized in that After inputting the first feature map into the decoder and performing feature decoding on the first feature map by the decoder to obtain a feature decoding map, the method further includes: Downsampling the feature decoding map to obtain a second feature map, where the size of the second feature map is the same as that of the first feature map; Performing feature fusion on the first feature map and the second feature map to obtain a third feature map; The step of inputting the first feature map into the classifier, and classifying the first feature map for coronary artery calcification by the classifier to obtain the coronary artery calcification classification result includes: The third feature map is input into the classifier, and the classifier is used to classify the third feature map for coronary artery calcification to obtain the coronary artery calcification classification result.

6. The method according to claim 4 or 5, characterized in that: The coronary calcification image includes a coronary calcification area and a region category corresponding to the coronary calcification area; the region segmentation module includes a first convolution block and a second convolution block; The step of inputting the feature decoding image into the region segmentation module, and performing coronary artery calcification region segmentation on the feature decoding image by the region segmentation module to obtain the coronary artery calcification image comprises: Inputting the feature decoding image into the first convolution block, and performing coronary calcification region segmentation on the feature decoding image through the first convolution block to obtain the coronary calcification region; The feature decoding image is input into the second convolution block, and the coronary calcification region category is identified on the feature decoding image by the second convolution block to obtain the region category.

7. The method according to claim 6, characterized in that Before inputting the target three-dimensional chest CT image into the target multi-task coronary calcification processing model, segmenting the target three-dimensional chest CT image into a coronary calcification region by the target multi-task coronary calcification processing model, obtaining a coronary calcification image corresponding to the target three-dimensional chest CT image, and classifying the target three-dimensional chest CT image for coronary calcification, the method further includes: Acquire a sample three-dimensional chest CT image and a sample label corresponding to the sample three-dimensional chest CT image, wherein the sample label includes a coronary artery calcification region label, a region category label, and a coronary artery calcification classification result label; The target multi-task coronary artery calcification processing model is trained based on the sample three-dimensional chest CT image and the sample label corresponding to the sample three-dimensional chest CT image.

8. The method according to claim 7, characterized in that The training of the target multi-task coronary artery calcification processing model according to the sample three-dimensional chest CT image and the sample label corresponding to the sample three-dimensional chest CT image includes: Inputting the sample three-dimensional chest CT image into a multi-task coronary calcification processing model, and obtaining the coronary calcification prediction area, coronary calcification prediction area category and coronary calcification classification prediction result output by the multi-task coronary calcification processing model; Calculating a first loss according to the predicted coronary calcification region and the coronary calcification region label; Calculating a second loss according to the coronary artery calcification prediction region category and the region category label; Calculating a third loss according to the coronary artery calcification classification prediction result and the coronary artery calcification classification result label; determining a total loss of the multi-task coronary calcification processing model according to the first loss, the second loss and the third loss; The model parameters of the multi-task coronary calcification processing model are adjusted according to the total loss so that the total loss is reduced, thereby obtaining the target multi-task coronary calcification processing model.

9. A coronary artery calcification score estimation device, characterized in that: include: An image acquisition module, used for acquiring a target three-dimensional chest CT image to be evaluated; a calcification processing module, configured to segment the target three-dimensional chest CT image into a coronary calcification region to obtain a coronary calcification image corresponding to the target three-dimensional chest CT image, and to classify the coronary calcification of the target three-dimensional chest CT image to obtain a coronary calcification classification result corresponding to the target three-dimensional chest CT image, wherein the coronary calcification classification result is used to indicate whether the target three-dimensional chest CT image has coronary calcification; an estimation module, for estimating a coronary calcification score based on the coronary calcification image to obtain a coronary calcification score of the target three-dimensional chest CT image if the coronary calcification classification result indicates that coronary calcification exists in the target three-dimensional chest CT image.

10. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory is connected to the processor, and the processor is used to execute one or more computer programs stored in the memory. When the processor executes the one or more computer programs, the computer device implements the method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor is caused to perform the method according to any one of claims 1 to 8.

Images