Method and device for constructing aortic dissection prediction model based on multimodal features

Abstract

The present invention provides a method, system, device, medium and program product for constructing a prediction model of aortic dissection based on multimodal features, and relates to the field of intelligent medical care. The method comprises: obtaining NCE-CT images of aortic dissection in a training set sample; training based on the NCE-CT images to obtain aortic tubular features; performing aortic segmentation based on the NCE-CT images to obtain aortic segmentation features; obtaining segmentation hints, inputting aortic segmentation features and segmentation hints, and aortic tubular features into a cross-attention network with a multi-head attention network to obtain fused segmentation features; inputting the fused segmentation features into a multi-head self-attention network, and adding the result obtained after the input to the aortic segmentation features element by element to obtain a prediction model showing the segmentation results.

Description

Technical Field

[0001] The present invention relates to the field of intelligent medical treatment, and more specifically, to a method, device, medium and program product for constructing an aortic dissection prediction model based on multimodal features. Background Art

[0002] Aortic dissection (AD) is a serious cardiovascular disease, and its early diagnosis and timely treatment are crucial to reduce mortality. Clinically, enhanced computed tomography (CE-CT) is the gold standard for diagnosing aortic dissection. However, CE-CT is expensive, complicated to operate, and patients are exposed to contrast agents, which may cause allergic reactions or renal damage, and in severe cases lead to death. Therefore, patients prefer to undergo NCE-CT for initial screening. If AD can be diagnosed with high enough reliability through NCE-CT, it will have important clinical significance. Studies have shown that non-contrast enhanced computed tomography (NCE-CT) can diagnose AD without the use of contrast agents. However, due to the lack of contrast agents, the contrast of NCE-CT images is low, and it is difficult to accurately distinguish the aorta from surrounding tissues, especially at the site of lesions, resulting in an increased risk of misdiagnosis and missed diagnosis.

[0003] NCE-CT is prone to misdiagnosis of AD due to the following reasons: (1) Due to the low contrast between the aorta and surrounding tissues, it is difficult for the network to focus on the tubular structure of the aorta; (2) The correlation between aortic geometry and AD is not certain, so it is difficult to establish a relationship model between aortic geometry and AD; (3) The intensity value of the aortic intimal valve is similar to that of blood, which poses a challenge to the network to identify the false lumen, which is also the criterion for radiologists to diagnose AD. Therefore, developing an automated system that can accurately identify the aortic structure on NCE-CT and accurately detect aortic dissection is conducive to reducing missed or misdiagnosed aortic dissection in early clinical screening. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention provides a method, device, medium and program product for constructing a prediction model of aortic dissection based on multimodal features; the method of the present invention is conducive to reducing missed diagnosis or misdiagnosis of diseases that require identification of tubular structures such as aortic dissection in early clinical screening by developing a method that can identify tubular structures with high precision in NCE-CT.

[0005] The first aspect of the present application discloses a method for constructing an aortic dissection prediction model based on multimodal features, the method comprising:

[0006] S1: Obtain NCE-CT images of aortic dissection in the training set samples;

[0007] S2: training based on NCE-CT images to obtain aortic tubular features; performing aortic segmentation based on the NCE-CT images to obtain aortic segmentation features;

[0008] S3: Obtain segmentation hints, input the aorta segmentation features, segmentation hints, and aorta tubular features into a cross-attention network with a multi-head attention network to obtain fused segmentation features;

[0009] S4: Input the fused segmentation feature into a multi-head self-attention network, add the result obtained after the input to the aorta segmentation feature element by element, and obtain a prediction model showing the segmentation result.

[0010] In some embodiments, between S2 and S3, the method further includes S21: reshaping and encoding the aortic tubular feature and the aortic segmentation feature to obtain a processed aortic tubular feature and a processed aortic segmentation feature; executing S3, replacing the aortic tubular feature in S3 with the processed aortic tubular feature, and replacing the aortic segmentation feature in S3 with the processed aortic segmentation feature;

[0011] Optionally, the formula for obtaining the tubular feature of the aorta after processing in S21 is:

[0012] X cls =En cls (Reshape(F cls ), Reshape(F cls )); Encls is the transformation encoder, Fcls is the aorta tubular feature, and Xcls is the processed aorta tubular feature;

[0013] The formula for obtaining the segmentation features of the aorta after processing in S21 is:

[0014] X seg =En seg (Reshape(F seg ), Reshape(F seg )); where Enseg is the transform encoder, Fseg is the aorta segmentation feature, and Xseg is the processed aorta segmentation feature.

[0015] In some embodiments, the formula for obtaining the fused tubular feature Ycls in S3 is:

[0016] Y cls =Pool(En share (Concat(X cls , prompt cls ), X seg)); Ycls is the fused tubular feature, Xcls is the aorta tubular feature, prompt cls is the tubular hint, Xseg is the aorta segmentation feature, and Enshare is the transform encoder;

[0017] The formula for obtaining the fused segmentation feature Yseg in S3 is:

[0018] Y seg =Reshape(En share (Concat(X seg , prompt seg ), X cls )); Yseg is the fusion segmentation feature, Xseg is the aorta segmentation feature, prompt seg is the segmentation hint, Xcls is the aorta tubular feature, and Enshare is the transformation encoder.

[0019] In some embodiments, the method for obtaining the aorta segmentation feature includes: performing aorta segmentation on the NCE-CT image by TotalSegmentator to obtain mask data; downsampling (scaling the image) the mask data by conventional U-Net to obtain a high-level segmentation feature map;

[0020] In some embodiments, the aortic tubular feature is obtained based on a tubular feature prediction model, and the method for constructing the tubular feature prediction model includes:

[0021] S201: Acquire CE-CT images of aortic dissection paired with NCE-CT images in training set samples;

[0022] S202: extracting high-scale features and low-scale features of the aorta in the NCE-CT image using a residual network;

[0023] S203: obtaining an offset based on the high-scale feature; guiding the position offset of the low-scale feature based on the offset, inputting the value of the low-scale feature after position offset and the offset into a multi-head self-attention network for fusion to obtain tubular feature data; obtaining a student model classification result based on the tubular feature data;

[0024] S204: Based on the CE-CT image, the pre-trained large model is adjusted to obtain a teacher model, and a corresponding teacher model classification result is obtained; based on the teacher model classification result and the student model classification result, the student model is optimized using knowledge distillation, and the constructed student model is the tubular feature prediction model.

[0025] Optionally, the method for obtaining the offset includes: S2031: sequentially inputting high-scale features into a convolution layer and a normalization layer to obtain a query vector Q to capture key features inside the aorta; S2032: mapping the query vector Q to obtain the offset;

[0026] Optionally, the query vector is obtained by the formula in S2031:

[0027] Q = LayerNorm (Conv (Fhigh)); where Q is the query vector, Fhigh is the high-scale feature, conv is the convolution layer, and LayerNorm is the normalization layer;

[0028] Optionally, the formula for obtaining the offset in S2032 is:

[0029] offset=Conv(gelu(LayerNorm(Conv(Q)))); where offset is the offset, Conv is the convolution layer, LayerNorm is the normalization layer, and gelu is the activation function;

[0030] Optionally, the value of the low-scale feature after position shift in S203 includes a deformed key K and value V, and the deformed key K and value V and the query vector Q are input into a multi-head self-attention network for fusion to obtain the tubular feature data FDefor;

[0031] Optionally, the method for constructing the tubular feature prediction model also includes: obtaining tubular prompts, inputting the aortic tubular features and tubular prompts and aortic segmentation features into a cross-attention network with a multi-head attention network to obtain a fused tubular feature; adding the fused tubular feature and the tubular feature data element by element, and outputting the student model classification result based on the result obtained by the element-by-element addition.

[0032] In some embodiments, after S1 and before S2, the method further comprises: performing standardized preprocessing on the paired NCE-CT and CE-CT images to obtain standardized NCE-CT data and standardized CE-CT data;

[0033] Optionally, the standardized preprocessing includes: extracting the ROI area of ​​the aorta in the image to obtain an aortic mask; increasing the proportion of the aorta in the image based on the aortic mask to obtain an image with an increased aortic proportion; downsampling, adjusting the CT threshold, and performing data enhancement on the image with the increased aortic proportion to obtain the standardized NCE-CT data and the standardized CE-CT data;

[0034] Optionally, the method of increasing the proportion of the aorta in the image includes: cropping the circumscribed matrix of the image based on the aorta mask to increase the proportion of the aorta mask in the image;

[0035] Optionally, the CT threshold of the NCE-CT data is set to [0,200];

[0036] Optionally, the CT threshold of the CE-CT data is set to [0,800];

[0037] Optionally, the teacher model and the student model adopt any of the following algorithms: Total Segmentator, ViT, CNN-BiLSTM, MobileViT; preferably Total Segmentator.

[0038] A second aspect of the present application provides a method for detecting aortic dissection, the method comprising:

[0039] Acquire the NCE-CT image to be tested;

[0040] Inputting the NCE-CT image into the prediction model described in the first aspect of the present application to obtain a segmentation result of the aorta;

[0041] Optionally, the method further includes: obtaining a classification result of whether an interlayer is contained based on the segmentation result.

[0042] The third aspect of the present application discloses a computer device, which includes: a memory and a processor; the memory is used to store a computer program; and the processor executes the computer program to implement the steps of the above method.

[0043] A fourth aspect of the present application discloses a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0044] A fifth aspect of the present application discloses a computer program product, including a computer program, which implements the steps of the above method when executed by a processor.

[0045] This application has the following beneficial effects:

[0046] The present invention designs a multi-task learning framework that can simultaneously learn and optimize multiple related tasks. Through shared representation and collaborative learning between tasks, the generalization ability and overall performance of the model are improved.

[0047] 1. This application innovatively discloses a method for constructing a prediction model for aortic dissection based on multimodal features. The method obtains aortic segmentation features obtained by coarse segmentation using a common network based on NCE-CT images, and aortic tubular features obtained by training based on NCE-CT images with deformable attention technology. Then, the segmentation prompts are used to input the above data into a cross-attention network with a multi-head attention network to obtain fused segmentation features. Subsequently, the fused segmentation features are added element by element to the aortic segmentation features obtained by coarse segmentation. After training, a prediction model showing the segmentation results is obtained. The prediction model can well improve the segmentation accuracy of the aorta, overcome the limitations of the prior art in complex AD geometry and low-contrast tissue recognition, and ensure high-accuracy diagnosis of AD without contrast agents.

[0048] 2. This application innovatively discloses a method for obtaining aortic tubular features based on NCE-CT images trained with deformable attention technology. This method extracts high-dimensional and low-dimensional features from NCE-CT images, learns a feature set from high-dimensional features, and adds it to low-dimensional features, which easily causes feature displacement. This displacement shifts the focus of the network to the inside of the tubular structure and introduces cross-scale information into the network output tubular feature data; it also enhances the tubular feature data to ensure the accuracy of the tubular feature data. This method can dynamically select and optimize features based on the specificity of the data and task requirements. By introducing this module, the deformable attention technology is introduced, allowing the network to focus on the inside of the aorta in the image.

[0049] 3. The model disclosed in this application has a wide range of applications. It is not only suitable for the detection of aortic dissection. When there is a segmentation mask of the vascular cavity, true cavity and false cavity, it can be used as the input of the module, and the three features of the vascular cavity, true cavity and false cavity are extracted through this module and then output as a higher-dimensional fusion feature. In addition, it can also be used for medical image registration, diagnosis of different diseases, etc. When it is applied to medical image registration, medical image registration is a technology that aligns two or more medical images with the aim of making the corresponding points in these images coincide in space. Registration can occur between different modalities (such as CT and MRI), between different time points (such as before and after treatment), or between different individuals (such as aligning a patient's image with a standard template). This mechanism is applied for this target task. First, features are extracted from the two images to be registered. Suppose we have two images, one is a CT image and the other is an MRI image.

[0050] Extracting feature Fct from CT images and feature Fmri from MRI images can be achieved through convolutional neural networks (CNN) or other feature extraction methods. Then, the extracted features are preliminarily encoded to generate higher-level feature representations. Fct and Fmri are passed through the Transformer encoder to obtain Xct and Xmri, and finally the cross-attention mechanism is used to fuse the features of different modalities.

[0051] The automated system developed in this application that can accurately detect aortic dissection by high-precision identification of aortic structure in NCE-CT is conducive to reducing missed or misdiagnosed aortic dissection in early clinical screening. The NCE-CT technology that combines the deformable self-attention module and the adaptive geometric information extraction module achieves highly reliable AD diagnosis by combining the two modules, significantly improving the recognition accuracy of aortic structure in NCE-CT images, overcoming the limitations of existing technologies in complex AD geometric morphology and low-contrast tissue recognition, and ensuring high-accuracy diagnosis of AD without contrast agents. Among them, the deformable self-attention module adaptively focuses on important features in the image at different scales, and enhances the ability to capture details and extract features of the aortic structure by adaptively adjusting the attention weight. The adaptive geometric information extraction module can effectively identify, extract and utilize the geometric information of the aorta, and effectively separate the geometric structure of the aorta from the NCE-CT image by using geometric constraints and morphological features, thereby improving the resolution of complex geometric morphology.

[0052] The technical solution of the present application can solve the following technical problems well: 1. The contrast between the aorta and the surrounding tissues is low, and it is difficult for the network to focus on the tubular structure of the aorta; 2. The correlation between the aortic geometric information and AD is not certain, and it is difficult to establish a relationship model between the aortic geometric information and AD. The existing modeling of aortic geometric information and AD simply extracts low-dimensional information and ignores the complexity of the relationship between the aortic geometric structure and AD. The existing vascular segmentation technology can extract the technical features of the blood vessels, but because low-scale features contain more morphological features and high-scale features contain more semantic information, the existing vascular segmentation does not take into account the lack of interaction between feature levels; 3. The intensity value of the aortic intimal valve is similar to the intensity value of blood, which brings challenges to the network to identify the false lumen. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0054] Figure 1 is a schematic diagram of a method flow chart provided by the first aspect of an embodiment of the present invention;

[0055] Figure 2 It is a schematic diagram of a system for constructing a multimodal feature aortic dissection prediction model provided in the second aspect of an embodiment of the present invention;

[0056] Figure 3 is a schematic diagram of a computer device provided by an embodiment of the present invention;

[0057] Figure 4 is a schematic diagram of the architecture of an exemplary computing device provided by an embodiment of the present invention;

[0058] Figure 5 is a schematic diagram of a storage medium provided by an embodiment of the present invention;

[0059] Figure 6 is a general network structure diagram provided by an embodiment of the present invention;

[0060] Figure 7 is a network structure diagram of an aorta tubular feature guidance module provided by an embodiment of the present invention;

[0061] Figure 8 is a network structure diagram of an adaptive feature fusion module provided by an embodiment of the present invention;

[0062] Fig. 9 It is a quantitative evaluation result compared with the commonly used segmentation model provided by the embodiment of the present invention;

[0063] Fig.10 A comparison diagram of the aorta segmentation structure between the model provided by the embodiment of the present invention and the commonly used segmentation model;

[0064] Fig.11 It is a quantitative evaluation result obtained by internal verification of a commonly used deep learning classification model provided in an embodiment of the present invention;

[0065] Fig.12 It is a quantitative evaluation result of external validation of a common deep learning classification model provided by an embodiment of the present invention using a data set from an external center;

[0066] Fig.13 It is a rendering of the aortic tubular feature guidance module provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0067] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0068] In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0069] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0070] Figure 1 The present invention provides a method for constructing a multimodal feature aortic dissection prediction model. Specifically, the method comprises the following steps:

[0071] S1: Obtain NCE-CT images of aortic dissection in the training set samples;

[0072] In some embodiments, after S1 and before S2, the method further comprises: performing standardized preprocessing on the paired NCE-CT and CE-CT images to obtain standardized NCE-CT data and standardized CE-CT data;

[0073] Optionally, the standardized preprocessing includes: extracting the ROI area of ​​the aorta in the image to obtain an aortic mask; increasing the proportion of the aorta in the image based on the aortic mask to obtain an image with an increased aortic proportion; downsampling, adjusting the CT threshold, and performing data enhancement on the image with the increased aortic proportion to obtain the standardized NCE-CT data and the standardized CE-CT data;

[0074] Optionally, the method of increasing the proportion of the aorta in the image includes: cropping the circumscribed matrix of the image based on the aorta mask to increase the proportion of the aorta mask in the image;

[0075] Optionally, the CT threshold of the NCE-CT data is set to [0,200];

[0076] Optionally, the CT threshold of the CE-CT data is set to [0,800];

[0077] Optionally, the teacher model and the student model adopt any of the following algorithms: Total Segmentator, nn-Unet, ViT, CNN-BiLSTM, MobileViT; preferably Total Segmentator.

[0078] S2: training based on NCE-CT images to obtain aortic tubular features; performing aortic segmentation based on the NCE-CT images to obtain aortic segmentation features;

[0079] In some embodiments, between S2 and S3, the method further includes S21: reshaping and encoding the aortic tubular feature and the aortic segmentation feature to obtain a processed aortic tubular feature and a processed aortic segmentation feature; executing S3, replacing the aortic tubular feature in S3 with the processed aortic tubular feature, and replacing the aortic segmentation feature in S3 with the processed aortic segmentation feature;

[0080] Optionally, the formula for obtaining the tubular feature of the aorta after processing in S21 is:

[0081] X cls =En cls (Reshape(F cls ), Reshape(F cls )); Encls is the transformation encoder, Fcls is the aorta tubular feature, and Xcls is the processed aorta tubular feature;

[0082] The formula for obtaining the segmentation features of the aorta after processing in S21 is:

[0083] X seg =En seg (Reshape(F seg ), Reshape(F seg )); where Enseg is the transform encoder, Fseg is the aorta segmentation feature, and Xseg is the processed aorta segmentation feature.

[0084] In some embodiments, the method for obtaining the aortic segmentation feature includes: performing aortic segmentation on the NCE-CT image through TotalSegmentator to obtain mask data; downsampling the mask data (scaling the image) through a conventional U-Net to obtain a high-level segmentation feature map; the aortic segmentation feature is a coarse segmentation feature obtained by roughly segmenting the standardized NCE-CT data through totalsegmentator. This method cannot accurately segment the patient's aortic structure.

[0085] In some embodiments, the aortic tubular feature is obtained based on a tubular feature prediction model, and the method for constructing the tubular feature prediction model includes:

[0086] S201: Acquire CE-CT images of aortic dissection paired with NCE-CT images in training set samples;

[0087] S202: extracting high-scale features and low-scale features of the aorta in the NCE-CT image using a residual network;

[0088] S203: obtaining an offset based on the high-scale feature; guiding the position offset of the low-scale feature based on the offset, inputting the value of the low-scale feature after position offset and the offset into a multi-head self-attention network for fusion to obtain tubular feature data; obtaining a student model classification result based on the tubular feature data;

[0089] S204: Based on the CE-CT image, the pre-trained large model is adjusted to obtain a teacher model, and a corresponding teacher model classification result is obtained; based on the teacher model classification result and the student model classification result, the student model is optimized using knowledge distillation, and the constructed student model is the tubular feature prediction model.

[0090] Optionally, the method for obtaining the offset includes: S2031: sequentially inputting high-scale features into a convolution layer and a normalization layer to obtain a query vector Q to capture key features inside the aorta; S2032: mapping the query vector Q to obtain the offset;

[0091] Optionally, the query vector is obtained by the formula in S2031:

[0092] Q = LayerNorm (Conv (Fhigh)); where Q is the query vector, Fhigh is the high-scale feature, conv is the convolution layer, and LayerNorm is the normalization layer;

[0093] Optionally, the formula for obtaining the offset in S2032 is:

[0094] offset=Conv(gelu(LayerNorm(Conv(Q)))); where offset is the offset, Conv is the convolution layer, LayerNorm is the normalization layer, and gelu is the activation function;

[0095] Optionally, the value of the low-scale feature after position shift in S203 includes a deformed key K and value V, and the deformed key K and value V and the query vector Q are input into a multi-head self-attention network for fusion to obtain the tubular feature data FDefor;

[0096] Optionally, the method for constructing the tubular feature prediction model also includes: obtaining tubular prompts, inputting the aortic tubular features and tubular prompts and aortic segmentation features into a cross-attention network with a multi-head attention network to obtain a fused tubular feature; adding the fused tubular feature and the tubular feature data element by element, and outputting the student model classification result based on the result obtained by the element-by-element addition.

[0097] S3: Obtain segmentation hints, input the aorta segmentation features, segmentation hints, and aorta tubular features into a cross-attention network with a multi-head attention network to obtain fused segmentation features;

[0098] In some embodiments, the formula for obtaining the fused tubular feature Ycls in S3 is:

[0099] Y cls =Pool(En share (Concat(X cls , prompt cls ), X seg )); Ycls is the fused tubular feature, Xcls is the aorta tubular feature, prompt cls is the tubular hint, Xseg is the aorta segmentation feature, and Enshare is the transform encoder;

[0100] The formula for obtaining the fused segmentation feature Yseg in S3 is:

[0101] Y seg =Reshape(En share (Concat(X seg , prompt seg ), X cls )); Yseg is the fusion segmentation feature, Xseg is the aorta segmentation feature, prompt seg is the segmentation hint, Xcls is the aorta tubular feature, and Enshare is the transformation encoder.

[0102] S4: Input the fused segmentation feature into a multi-head self-attention network, add the result obtained after the input to the aorta segmentation feature element by element, and obtain a prediction model showing the segmentation result.

[0103] In some embodiments, the aortic tubular feature guidance module dynamically selects and optimizes features according to the specificity of the data and task requirements. The deformable attention technology is introduced to enable the network to focus on the inside of the aorta in the image; learning a feature set from high-scale features and adding it to low-scale features can easily cause feature displacement, which shifts the focus of the network to the inside of the aorta and introduces cross-scale information into the network. The output tubular features are the features inside the aorta, that is, the overall features of the aorta. When there are segmentation masks of the vascular lumen, true lumen and false lumen, they can be used as the input of this module, and the three features of the vascular lumen, true lumen and false lumen are extracted through this module and then output as higher-dimensional fusion features. It can be used in intraoperative navigation scenarios.

[0104] After S203, the method further includes: performing enhancement processing on the tubular feature data, and obtaining a classification result based on the enhanced tubular feature data; optionally, the enhancement processing includes: adding the tubular feature data to the query vector to obtain an original intermediate feature Mid; sequentially inputting the original intermediate feature Mid into a normalization layer (LayerNorm), a multi-layer perceptron (MLP) and a dropout layer (Dropath) for processing to obtain a processed intermediate feature, taking the sum of the processed intermediate feature and the original intermediate feature to obtain the enhanced tubular feature data; this process enhances the nonlinear representation capability of the feature, so that the network can better capture the complex internal structure information of the aorta;

[0105] Optionally, the calculation formula of the original intermediate feature Mid is:

[0106] Mid=Q+Dropath(F Defor ), where FDefor is the tubular feature data;

[0107] Optionally, the formula for obtaining the enhanced tubular feature data is:

[0108] Out=Mid+Dropath(MLP(LayerNorm(Mid))), where out is the tubular feature data after the enhancement processing;

[0109] Optionally, the method further includes: applying path discarding to process the tubular feature data, and adding the tubular feature data after the path discarding process to the query vector Q to obtain an original intermediate feature.

[0110] Through this adaptive geometric feature fusion module, by learning the importance weights of features, effective screening and combination of features can be achieved. In addition to being used in the aorta segmentation field disclosed in this embodiment, the method of this module can also be used in specific application scenarios such as medical image registration and diagnosis of different diseases: obtain the first image and the second image to be tested, extract the features in the first image and the second image, and obtain the first image feature and the second image feature; fuse the first image feature and the second image feature. The first image and the second image are images of different modalities or images at different time points, or images of different individuals;

[0111] A second aspect of the present application discloses a method for detecting aortic dissection, the method comprising:

[0112] Acquire the NCE-CT image to be tested;

[0113] The NCE-CT image is input into the prediction model described in the first aspect of the present application to obtain a segmentation result of the aorta. Optionally, the method further includes: obtaining a classification result of whether the aorta contains a dissection based on the segmentation result.

[0114] In some embodiments, the term "subject" or "person to be tested" or "sample to be tested" as used herein refers to any animal (e.g., mammal), including but not limited to humans, non-human primates, rodents, etc., which will be the recipient of a specific treatment. Generally, the terms "subject" and "patient" are used interchangeably herein when referring to human subjects. Preferably, the subject is a human. In some embodiments, the sample to be tested is a patient clinically used to receive a prognostic assessment.

[0115] Figure 3 is a schematic diagram of a computer device provided by an embodiment of the present invention, such as Figure 3 As shown, the device may include: one or more processors, and one or more memories; wherein the memories store computer-readable codes, and when the computer-readable codes are run by the one or more processors, the method described above may be executed.

[0116] The processor in this embodiment can be an integrated circuit chip with signal processing capabilities. The above processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The disclosed methods, operations and logic block diagrams in the embodiments of the present disclosure can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc., which can be an X86 architecture or an ARM architecture.

[0117] In general, various example embodiments of the present disclosure may be implemented in hardware or dedicated circuits, software, firmware, logic, or any combination thereof. Certain aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor, or other computing device. When various aspects of the disclosed embodiments are illustrated or described as block diagrams, flow charts, or using some other graphical representation, it will be understood that the blocks, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuits or logic, general purpose hardware or controllers or other computing devices, or some combination thereof as non-limiting examples.

[0118] For example, the method or device according to the embodiment of the present disclosure may also be implemented by Figure 4 The architecture of the computing device 3000 shown in FIG. Figure 4 As shown, the computing device 3000 may include a bus 3010, one or more CPUs 3020, a read-only memory (ROM) 3030, a random access memory (RAM) 3040, a communication port 3050 connected to a network, an input / output component 3060, a hard disk 3070, etc. The storage device in the computing device 3000, such as ROM 3030 or hard disk 3070, may store various data or files used for processing and / or communication of the method provided by the present disclosure and program instructions executed by the CPU. The computing device 3000 may also include a user interface 3080. Of course, Figure 4 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 4 One or more components of a computing device are shown.

[0119] The embodiment of the present invention also provides a computer-readable storage medium, such as Figure 5As shown, it is a schematic diagram of a storage medium provided by an embodiment of the present invention, and a computer readable instruction 4010 is stored on the computer storage medium 4020. When the computer readable instruction 4010 is executed by the processor, the method according to the embodiment of the present disclosure described with reference to the above figures can be executed. The computer readable storage medium in the embodiment of the present disclosure can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. The non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus random access memory (DR RAM). It should be noted that the memory of the methods described herein is intended to include, but is not limited to, these and any other suitable types of memory. It should be noted that the memory of the methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0120] The embodiments of the present disclosure further provide a computer program product or system, including a computer program, which implements the steps of the above method when executed by a processor.

[0121] In some embodiments, this embodiment further discloses a system for constructing an aortic dissection prediction model based on multimodal features, the system comprising:

[0122] A first acquisition module 201 is used to acquire NCE-CT images of aortic dissection in a training set sample;

[0123] The feature extraction module 202 is used to obtain the tubular features of the aorta through training based on the NCE-CT images; and to obtain the aorta segmentation features through segmentation of the aorta based on the NCE-CT images;

[0124] A cross attention processing module 203 is used to obtain segmentation hints, input the aorta segmentation features, the segmentation hints, and the aorta tubular features into a cross attention network with a multi-head attention network to obtain a fused segmentation feature;

[0125] The segmentation result output module 204 is used to input the fused segmentation feature into the multi-head self-attention network, add the result obtained after the input to the aorta segmentation feature element by element, and obtain a prediction model showing the segmentation result.

[0126] In some embodiments, this embodiment further discloses a detection system for aortic dissection, the system comprising:

[0127] The second acquisition module is used to acquire the NCE-CT image to be tested;

[0128] The output module is used to input the NCE-CT image into the prediction model described in the first aspect of the present application to obtain the segmentation result of the aorta.

[0129] In some embodiments, the system further comprises a classification module for obtaining a classification result of whether a dissection is contained based on the segmentation result.

[0130] Example

[0131] The overall network structure diagram designed in this embodiment is as follows Figure 6 As shown, specifically, a flow chart of the end-to-end detection system for aortic dissection, wherein the high-dimensional feature is the high-scale feature Fhigh, the low-dimensional feature is the low-scale feature Flow, and the tubular feature is Fdefor.

[0132] The aortic dissection detection system proposed in the present invention mainly includes: first, paired NCE-CT and CE-CT images of aortic dissection patients from multiple centers are collected as system input, and the data are subjected to standardized preprocessing such as image cropping and resampling through a multi-center data preprocessing module, and then the standardized data is subjected to an aortic tubular feature guidance module and an adaptive geometric feature fusion module, and finally the fused features are sent to a deep learning classification network for automatic diagnosis to obtain a diagnosis result.

[0133] Specifically, 1) Multi-center data preprocessing module: First, the NCE-CT and CE-CT images are processed by a deep learning medical segmentation model to extract the region of interest (ROI) of the aorta, thereby obtaining the aortic mask. According to the mask, the external matrix of the image is cropped so that the proportion of the aorta in the image is increased. Then, the image is downsampled and the CT threshold is adjusted, and data augmentation operations are performed to ensure that the multi-center data is input in a uniform size of 128×128×256, and the CT thresholds of NCE-CT and CE-CT are set to [0, 200] and [0, 800], respectively. Finally, these preprocessed data are learned end-to-end through a neural network.

[0134] 2) Aortic tubular feature guidance module: After preprocessing, the image is input into the aortic tubular feature guidance module. This module uses a multi-scale convolutional neural network (CNN) and a self-attention structure to extract features of different scales from the image. The self-attention mechanism is an attention mechanism used to capture long-range dependencies in the input sequence. It plays an important role in the tubular feature guidance module of the end-to-end detection system for aortic dissection, which can help the model better learn the tubular features in the aortic image. Specifically, the input of this module includes high-scale features Fhigh and low-scale features Flow. We first map Fhigh to Q, and then map Q to the offset.

[0135] Q=LayerNorm(Conv(Fhigh))#(1)

[0136] offset=Conv(gelu(LayerNorm(Conv(Q))))#(2)

[0137] The offset is then applied to F low , and mapped to K and V, and then the deformation feature map is obtained through the multi-head self-attention module (MSA), and F Defor express.

[0138] K,V=Conv(grid sample (F low ,offset))#(3)

[0139] F Defor =MSA(Q, K, V)#(4)

[0140] Finally, perform the following coding to obtain the output result of the tubular feature.

[0141] Mid=Q+Dropath(F Defor )#(5)

[0142] Out=Mid+Dropath(MLP(LayerNorm(Mid)))#(6)

[0143] Among them, LayerNorm is the normalization layer, Conv is the convolution, gelu is the activation function, grid_sample is the Pytorch function: torch.nn.functional.grid sample, MSA is multi-head self-attention, Dropath is path elimination, and MLP is a multi-layer perceptron. The specific implementation network structure of this module is as follows Figure 7 As shown, the high-dimensional feature is the high-scale feature Fhigh, the low-dimensional feature is the low-scale feature Flow, and the tubular feature is Fdefor.

[0144] 3) Adaptive feature fusion module

[0145] Given F seg The high-level segmentation feature map and F cls After representing the high-level features of the aorta tubular shape, they are first reshaped and then sent to the transformer encoder for preliminary encoding.

[0146] X cls =En cls (Reshape(F cls ), Reshape(F cls ))#(7)

[0147] X seg =En seg (Reshape(F seg ), Reshape(F seg ))#(8)

[0148] Then a cross-attention module is designed. To enable this module to distinguish the sources of query features, segmentation and classification cues Xcls and Xseg are added for distinction.

[0149] Y cls =Pool(En share (Concat(X cls , prompt cls ), X seg ))#(9)

[0150] Y seg =Reshape(En share (Concat(X seg , prompt seg ), X cls ))#(10)

[0151] Among them, Encls, Enseg and Enshare are transform encoders. The two results Ycls and Yseg will return the classification network and segmentation network respectively to enhance their functions. The specific implementation network structure diagram of this module is as follows Figure 8 As shown; the segmentation feature at the input end is the tubular segmentation feature in S3021, and the tubular feature is Figure 7 The output tubular feature data, the fused segmentation feature at the output end is Yseg, and the fused tubular feature is Ycls.

[0152] 4) Deep learning classification module

[0153] During the training process of the network, the network parameters are updated by calculating the loss, gradient descent, and back propagation. The network loss consists of three parts, namely the classification loss Losscls, the classification loss Lossseg, and the similarity loss Losssim. Specifically, the cross entropy loss is used to calculate Losscls, as shown in Formula 11; the weighted sum of the Dice loss LossDice and the cross entropy loss LossCrEn is used to calculate Lossseg, as shown in Formula 14. The total loss of the model is shown in Formula 15.

[0154] Loss cls =-γ1ylog p-γ2(1-y)log(1-p)#(11)

[0155] Loss CrEn =-γ3ylog p-γ4(1-y)log(1-p)#(12)

[0156]

[0157] Loss seg =Loss CrEn +Loss Dice #(14)

[0158] Loss total =Loss cls +Loss seg #(15)

[0159] The AdamW optimizer is used in combination with the total loss function for network learning, and finally the diagnostic output results are obtained.

[0160] 5) Diagnostic output module

[0161] Classification In the embodiment of the present invention, quantitative evaluation is used to analyze the performance of the model, and the advantages of each module are demonstrated by combining ablation comparison experiments. For the segmentation task, we use Dice similarity coefficient (DSC), Jaccard index (JI), mean intersection over union (MIoU) and frequency weighted intersection over union (FWIoU) for quantitative evaluation. Dice is a method for measuring the overlap between a segmentation result and a true segmentation result. JI is an indicator similar to the Dice similarity coefficient, which is used to measure the overlap between a segmentation result and a true segmentation result. MIoU is the average of the DSC or JI indicators of multiple segmentation results. FWIoU is an indicator based on the area of ​​the region, which is used to measure the overlap between the segmentation result and the true segmentation result. The expression is as follows:

[0162]

[0163] For classification tasks, the following are used: Accuracy, Sensitivity, Specificity, F1 score (F1) and Area Under Curve (AUC). Accuracy refers to the proportion of the number of samples correctly predicted by the model to the total number of samples. Sensitivity, also known as recall, refers to the proportion of all true positive examples correctly identified by the model. Specificity, also known as precision, refers to the proportion of all negative examples correctly identified by the model. The F1 score is the harmonic mean of sensitivity and specificity, which can comprehensively measure the recall and precision of the model. Accuracy refers to the proportion of the number of samples correctly predicted by the model to the total number of samples. AUC is the value of the area under the ROC curve, which can measure the model's ability to classify positive and negative samples. The ROC curve is a curve that plots the relationship between sensitivity and specificity. The evaluation index expression is as follows:

[0164]

[0165]

[0166] Where TP: True Positive, the number of samples predicted to be positive and actually positive, FP: False Positive, the number of samples predicted to be positive and actually negative, FN: False Negative, the number of samples predicted to be negative and actually positive. A_i: The area of ​​the i-th true segmentation region, A_tot: The sum of the areas of all true segmentation regions.

[0167] The present invention is trained and internally validated on aortic dissection datasets obtained from three medical centers. For the segmentation task, the quantitative indicators of the segmentation results are as follows: Fig. 9 As shown. Compared with common medical image segmentation models, our framework achieved the best results. Specifically, the segmentation results of U-net obtained DSC of 0.864, JI of 0.762, MioU of 0.878, and FWIoU of 0.991. It is lower than nn-U-net, with DSC of 0.880, JI of 0.786, MIoU of 0.891, and FWIoU of 0.992. In contrast, compared with nn-U-net, the DSC of our proposed framework is improved by 4.8%, JI by 8.1%, MIoU by 4.1%, and FWIoU by 0.3%267. These segmentation results are also reflected in the visualization, as shown in Fig.10 The cavity segmentation result of U-net is not smooth enough as shown by the arrow, and the bifurcation segmentation result of nn-U-net shows obvious under-segmentation as shown by the arrow. Compared with them, our result is smoother and achieves better results on the aortic bifurcation.

[0168] In the classification task, we compared several of the most commonly used deep learning classification models. The quantitative classification results of internal verification are as follows: Fig.11 As shown in Figure 2, compared with common classification networks and previous AD classification studies, our framework achieved the highest accuracy of 91.1%, the highest sensitivity of 92.5%, and the highest F1 score of 92.2%. In addition, in order to test the generalization performance of the model, we used the dataset of an external center for external validation. The quantitative results are shown in Figure 2. Fig.12 As shown in Figure 3, our framework still achieves the highest accuracy of 84.0%, the highest sensitivity of 88.8%, and the highest F1 score of 83.6%.

[0169] Finally, we show grad-CAM to visualize whether the feature-guided module focuses the network’s attention from the surrounding tissue to the tubular structure of the aorta. Fig.13 As shown in Figure 1, we show grad-CAM in the coronal and sagittal planes before and after feature guidance to visually show the effect of the module. In the sagittal plane, before adding feature guidance, the model's attention was on the aortic arch, but the attention strength was not enough. After adding feature guidance, the model's attention in the aortic tubular structure area was significantly enhanced. In the coronal plane, before using feature guidance, the model's attention area had some deviations from the lumen area. After adding feature guidance, the model's attention was correctly focused on the aortic tubular structure area. Obviously, after adding feature guidance, the network's attention is more focused on the tubular structure of the aorta.

[0170] 1) This paper proposes a deep learning-based framework that can effectively process and fuse medical image data of multiple modalities, achieving efficient use and comprehensive analysis of data. This innovation ensures that the system has higher accuracy and robustness when processing multimodal data.

[0171] 2) Establish a standard process for preprocessing multi-center aortic NCE-CT and CE-CT imaging data

[0172] 3) The present invention introduces a feature guidance module, which can dynamically select and optimize features according to the characteristics of the data and task requirements, thereby improving the performance and adaptability of the model.

[0173] 4) The present invention introduces adaptive feature fusion, which achieves effective screening and combination of features by learning the importance weights of features.

[0174] In summary, the present invention can achieve end-to-end accurate detection of aortic dissection disease through NCE-CT images. The current image processing algorithms and feature extraction methods can only extract low-dimensional geometric information of the aorta. Our model can guide the tubular features of the aorta and combine adaptive feature fusion to make the model focus on the inside of the aorta, and can solve the problem of geometric information complexity presented by the aortic geometric structure. In addition, in the multi-task model framework designed by us, the standardized preprocessing of data collected from multiple centers completes the automatic segmentation of the aortic mask, reducing the diagnostic errors caused by new or rare lesions.

[0175] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0176] In general, various example embodiments of the present disclosure may be implemented in hardware or dedicated circuits, software, firmware, logic, or any combination thereof. Certain aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor, or other computing device. When various aspects of the disclosed embodiments are illustrated or described as block diagrams, flow charts, or using some other graphical representation, it will be understood that the blocks, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuits or logic, general purpose hardware or controllers or other computing devices, or some combination thereof as non-limiting examples.

[0177] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0178] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0179] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0180] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0181] The exemplary embodiments of the present disclosure described in detail above are merely illustrative and not restrictive. It should be understood by those skilled in the art that various modifications and combinations may be made to these embodiments or their features without departing from the principles and spirit of the present disclosure, and such modifications should fall within the scope of the present disclosure.

Claims

1. A method for constructing a prediction model for aortic dissection based on multimodal features, characterized in that: The method comprises: S1: Obtain NCE-CT images of aortic dissection in the training set samples; S2: training based on NCE-CT images to obtain aortic tubular features; performing aortic segmentation based on NCE-CT images to obtain aortic segmentation features; the aortic tubular features are obtained based on a tubular feature prediction model, and the method for constructing the tubular feature prediction model includes: S201: obtaining CE-CT images of aortic dissection paired with NCE-CT images in the training set samples; S202: extracting high-scale features and low-scale features of the aorta in the NCE-CT images using a residual network; S203: obtaining an offset based on the high-scale features; guiding the position offset of the low-scale features based on the offset, and inputting the value of the low-scale features after the position offset and the offset into The multi-head self-attention network is fused to obtain tubular feature data; tubular prompts are obtained, and the tubular feature data, tubular prompts, and aorta segmentation features are input into a cross-attention network with a multi-head attention network to obtain fused tubular features; the fused tubular features are added element by element to the tubular feature data, and a student model classification result is output based on the result obtained by the element-by-element addition; S204: the pre-trained large model is adjusted based on the NCE-CT image to obtain a teacher model, and a corresponding teacher model classification result is obtained; based on the teacher model classification result and the student model classification result, the student model is optimized by using knowledge distillation, and the constructed student model is the tubular feature prediction model; S3: Obtain segmentation hints, input the aorta segmentation features, segmentation hints, and aorta tubular features into a cross-attention network with a multi-head attention network to obtain fused segmentation features; S4: Input the fused segmentation features into the multi-head self-attention network, add the result obtained after the input to the aorta segmentation feature element by element, and obtain a prediction model showing the segmentation result.

2. The method for constructing a multimodal feature aortic dissection prediction model according to claim 1, characterized in that: The method for obtaining the offset offset includes: S2031: sequentially inputting high-scale features into the convolution layer and the normalization layer to obtain a query vector Q to capture key features inside the aorta; S2032: mapping the query vector Q to obtain the offset offset.

3. The method for constructing a multimodal feature aortic dissection prediction model according to claim 2, characterized in that: The formula for obtaining the query vector in S2031 is: ; Among them, Q is the query vector, Fhigh is the high-scale feature, conv is the convolution layer, and LayerNorm is the normalization layer.

4. The method for constructing a multimodal feature aortic dissection prediction model according to claim 2, characterized in that: The formula for obtaining the offset in S2032 is: ; Among them, offset is the offset, Conv is the convolution layer, LayerNorm is the normalization layer, and gelu is the activation function.

5. The method for constructing a multimodal feature-based aortic dissection prediction model according to claim 1, characterized in that: The value of the low-scale feature after position shift in S203 includes the deformed key K and value V. The deformed key K and value V and the query vector Q are input into the multi-head self-attention network for fusion to obtain the tubular feature data F. Defor .

6. The method for constructing a multimodal feature-based aortic dissection prediction model according to claim 1, characterized in that: Between S2 and S3, the method further includes S21: reshaping, encoding and processing the aortic tubular feature and the aortic segmentation feature to obtain a processed aortic tubular feature and a processed aortic segmentation feature; executing S3, replacing the aortic tubular feature in S3 with the processed aortic tubular feature, and replacing the aortic segmentation feature in S3 with the processed aortic segmentation feature.

7. The method for constructing a multimodal feature-based aortic dissection prediction model according to claim 6, characterized in that: The formula for obtaining the tubular features of the aorta after processing in S21 is: ; Among them, Encls is the transformation encoder, Fcls is the aortic tubular feature, and Xcls is the processed aortic tubular feature.

8. The method for constructing a multimodal feature-based aortic dissection prediction model according to claim 6, characterized in that: The formula for obtaining the segmentation features of the aorta after processing in S21 is: ; Among them, Enseg is the transform encoder, Fseg is the aorta segmentation feature, and Xseg is the processed aorta segmentation feature.

9. The method for constructing a multimodal feature-based aortic dissection prediction model according to claim 1, characterized in that: The formula for obtaining the fused tubular feature is: ; Among them, Ycls is the fused tubular feature, Xcls is the aorta tubular feature, is a tubular hint, Xseg is the aorta segmentation feature, and Enshare is a transform encoder; the acquisition formula of the fused segmentation feature is: ; Among them, Yseg is the fusion segmentation feature, Xseg is the aorta segmentation feature, is the segmentation hint, Xcls is the aorta tubular feature, and Enshare is the transformation encoder.

10. The method for constructing a multimodal feature-based aortic dissection prediction model according to claim 1, characterized in that: The method for acquiring the aorta segmentation feature includes: performing aorta segmentation on the NCE-CT image by TotalSegmentator to obtain mask data; and downsampling the mask data by conventional U-Net to obtain a high-level segmentation feature map.

11. The method for constructing a multimodal feature-based aortic dissection prediction model according to claim 1, characterized in that: After S1 and before S2, the method further includes: performing standardized preprocessing on the paired NCE-CT and CE-CT images to obtain standardized NCE-CT data and standardized CE-CT data.

12. The method for constructing a multimodal feature-based aortic dissection prediction model according to claim 11, characterized in that: The standardized preprocessing includes: extracting the ROI area of ​​the aorta in the image to obtain an aortic mask; increasing the proportion of the aorta in the image based on the aortic mask to obtain an image with an increased aortic proportion; downsampling, CT threshold adjustment, and data enhancement of the image with an increased aortic proportion to obtain the standardized NCE-CT data and standardized CE-CT data.

13. The method for constructing a multimodal feature-based aortic dissection prediction model according to claim 12, characterized in that: The method for increasing the proportion of the aorta in the image includes: cropping the circumscribed matrix of the image based on the aorta mask, and increasing the proportion of the aorta mask in the image.

14. A method for detecting aortic dissection, characterized in that: The method comprises: Acquire the NCE-CT image to be tested; The NCE-CT image is input into the prediction model described in any one of claims 1 to 13 to obtain a segmentation result of the aorta.

15. The method for detecting aortic dissection according to claim 14, characterized in that: The method further includes: obtaining a classification result of whether a layer is contained based on the segmentation result.

16. A computer device, characterized in that: The device comprises: a memory and a processor; the memory is used to store a computer program; the processor executes the computer program to implement the steps of the method according to any one of claims 1 to 15.

17. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 15 are implemented.

18. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 15 are implemented.

Images