Post-stroke dysphagia risk scoring method, device, equipment, medium and product
By constructing a two-path model for swallowing dysphagia risk score, using the patient's brain magnetic resonance image and stroke infarction lesions segmentation map for feature extraction and fusion, the problem of early prediction and evaluation accuracy of swallowing dysphagia risk score after stroke is solved, and more accurate risk prediction and diagnostic assistance is achieved.
Patent Information
- Application Number
- CN202411747664.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-29
- Publication Date
- 2025-03-25
AI Technical Summary
The prior art has problems with insufficient early prediction ability and low evaluation accuracy in the risk score of dysphagia after stroke.
A dual-path model for swallowing dysphagia risk scores including two backbone networks, feature fusion modules and risk prediction head modules was used to extract and fusion features of the patient's brain magnetic resonance image and stroke infarction lesions segmentation map to predict the probability of patients' poststroke dysphagia.
It has achieved an improved accuracy in predicting and evaluating the risk of poststroke dysphagia, which can provide doctors with more accurate diagnostic assistance and rehabilitation intervention suggestions.
Smart Images
Figure CN119694561A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of post-stroke dysphagia, and in particular to a post-stroke dysphagia risk scoring method, device, equipment, medium and product. Background Art
[0002] Dysphagia is a major complication after acute stroke, affecting the majority of patients. Advances in the diagnosis of dysphagia after stroke will enable management based on individual patterns of functional impairment and risk factors for susceptibility to complications. Advances in dysphagia rehabilitation are essential to reduce mortality and improve quality of life for patients after stroke.
[0003] There are two main technologies related to the risk scoring of post-stroke dysphagia. One is to rely on traditional clinical screening and instrument testing. However, screening depends on the operation of clinicians and the symptoms of patients, and cannot predict risks in advance, resulting in some patients being unable to receive effective treatment in a timely manner, and the real-time performance is poor. The second is to use the NIHSS model, POST-NIHSSO model and Predictive Swallowing Score model to score. These models can make early predictions, but the prediction accuracy is not high.
[0004] In summary, there is a need for a risk scoring method for post-stroke dysphagia that can both predict in advance and improve assessment accuracy. Summary of the invention
[0005] The purpose of this application is to provide a method, device, equipment, medium and product for scoring the risk of post-stroke dysphagia, which can both predict in advance and improve the accuracy of assessment.
[0006] To achieve the above objectives, this application provides the following solutions:
[0007] In a first aspect, the present application provides a method for scoring the risk of post-stroke dysphagia, comprising:
[0008] Obtaining magnetic resonance images of the patient's brain and stroke infarction lesion segmentation maps;
[0009] Inputting the magnetic resonance image of the patient's brain and the stroke infarction lesion segmentation map into the dysphagia risk scoring model to obtain the probability of the patient suffering from post-stroke dysphagia; the dysphagia risk scoring model is obtained by training the dysphagia risk scoring dual-path model; the dysphagia risk scoring dual-path model comprises: two backbone networks, a feature fusion module connected to the output ends of the two backbone networks, and a risk prediction head module connected to the output end of the feature fusion module; the backbone network comprises: an initialization layer and a residual module connected in sequence; the residual module comprises N residual blocks connected by residuals; the residual block comprises M convolution modules connected by residuals, and the convolution module comprises a plurality of first convolution layers connected in sequence;
[0010] A score of post-stroke dysphagia occurring in the patient is obtained according to the probability of the patient experiencing post-stroke dysphagia.
[0011] Optionally, the input end of the first convolution module in the residual block is connected to the input end of the second convolution module, which is the input end of the residual block; the output end of the mth convolution module is connected to the input end of the m+1th convolution module and the output end of the m+1th convolution module, and the output end of the last convolution module is the output end of the residual module; 1≤m<M.
[0012] Optionally, the input end of the first residual block in the residual module is the input end of the residual module, the output end of the nth residual block is connected to the input end of the n+1th residual block through two parallel channels; the output end of the Nth residual block is the output end of the residual module, 1≤n<N.
[0013] Optionally, the residual module also includes: N-1 second convolutional layers; the output end of the nth residual block is directly connected to the input end of the n+1th residual block, and the output end of the nth residual block is also connected to the input end of the n+1th residual block through the nth second convolutional layer.
[0014] Optionally, the process of training the dysphagia risk score dual-path model is:
[0015] A post-stroke dysphagia dataset is obtained; the post-stroke dysphagia dataset includes multiple samples, each sample includes a true label of the sample patient and a brain magnetic resonance image and a stroke infarction lesion segmentation map of the sample patient; the true label is whether post-stroke dysphagia occurs or whether post-stroke dysphagia does not occur;
[0016] Taking minimizing the loss function as the goal, the dysphagia risk score dual-path model is trained according to the post-stroke dysphagia dataset to obtain a dysphagia risk score model.
[0017] Optionally, the initialization layer includes a third convolutional layer, a batch normalization layer and an activation layer connected in sequence.
[0018] In a second aspect, the present application provides a post-stroke dysphagia risk scoring device, comprising:
[0019] An acquisition module, used for acquiring a magnetic resonance image of the patient's brain and a stroke infarction lesion segmentation map;
[0020] A prediction module is used to input the magnetic resonance image of the patient's brain and the stroke infarction lesion segmentation map into a dysphagia risk scoring model to obtain the probability of the patient suffering from post-stroke dysphagia; the dysphagia risk scoring model is obtained by training the dysphagia risk scoring dual-path model; the dysphagia risk scoring dual-path model includes: two backbone networks, a feature fusion module connected to the output ends of the two backbone networks, and a risk prediction head module connected to the output end of the feature fusion module; the backbone network includes: an initialization layer and a residual module connected in sequence; the residual module includes N residual blocks connected by residuals; the residual block includes M convolution modules connected by residuals, and the convolution module includes a plurality of first convolution layers connected in sequence;
[0021] A scoring module is used to obtain a score of post-stroke dysphagia in the patient based on the probability of the patient suffering from post-stroke dysphagia.
[0022] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described post-stroke dysphagia risk scoring methods.
[0023] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described post-stroke dysphagia risk scoring methods.
[0024] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned post-stroke dysphagia risk scoring methods.
[0025] According to the specific embodiments provided in this application, this application has the following technical effects:
[0026] The present application provides a method, device, equipment, medium and product for scoring risk of post-stroke dysphagia, wherein the magnetic resonance image of the patient's brain and the stroke infarction lesion segmentation map are input into a dysphagia risk scoring model to obtain the probability of the patient suffering from post-stroke dysphagia, and the score of the patient suffering from post-stroke dysphagia is obtained according to the probability of the patient suffering from post-stroke dysphagia. The model is used for prediction, and it can be predicted in advance. The dual-path model for dysphagia risk scoring includes: two backbone networks, a feature fusion module connected to the output ends of the two backbone networks, and a risk prediction head module connected to the output end of the feature fusion module; the backbone networks are connected to the output ends of the two backbone networks, and the risk prediction head module is ... The backbone network includes: an initialization layer and a residual module connected in sequence; the residual module includes N residual blocks connected by residual connections; the residual block includes M convolution modules connected by residual connections, and the convolution module includes multiple first convolution layers connected in sequence. In order to improve the model's feature extraction capability for medical images, the convolution modules are connected by using residual connections to ensure that the model can extract sufficient information for risk prediction. By constructing two backbone networks, feature extraction of brain image features and segmentation mask features is achieved, thereby enhancing the overall feature extraction capability of the model, facilitating the improvement of the accuracy of the final risk score, and improving the accuracy of the assessment. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments 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 creative work.
[0028] Figure 1 A structural diagram of a dual-path model for dysphagia risk scoring provided in an embodiment of the present application;
[0029] Figure 2 A diagram of the backbone model structure provided in one embodiment of the present application;
[0030] Figure 3 A residual block structure diagram provided in an embodiment of the present application;
[0031] Figure 4 A block diagram of a method for scoring risk of post-stroke dysphagia provided in one embodiment of the present application;
[0032] Figure 5 A flowchart of a method for scoring risk of post-stroke dysphagia provided in one embodiment of the present application;
[0033] Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION
[0034] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0035] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.
[0036] From the perspective of imaging, using the characteristic information contained in the brain images of stroke patients to predict the risk of swallowing disorders can help doctors identify high-risk patients earlier, so as to assist doctors in making more accurate decisions and rehabilitation intervention suggestions. Based on this, in an exemplary embodiment, Figure 4 As shown, a method for scoring the risk of post-stroke dysphagia is provided, including constructing a post-stroke dysphagia dataset, constructing a dual-path model for dysphagia risk scoring, using the model to predict the risk of dysphagia, and obtaining a dysphagia risk score, such as Figure 5 As shown, the specific steps include:
[0037] Step 201: Obtain a magnetic resonance image of the patient's brain and a stroke infarction lesion segmentation map.
[0038] Step 202: Inputting the magnetic resonance image of the patient's brain and the stroke infarction lesion segmentation map into a dysphagia risk scoring model to obtain the probability of the patient suffering from post-stroke dysphagia; the dysphagia risk scoring model is obtained by training a dysphagia risk scoring dual-path model; Figure 1 As shown in FIG, the dual-path model for dysphagia risk scoring includes: two backbone networks (one of which is an imaging backbone network for inputting magnetic resonance images to obtain brain imaging features (image features), and the other is a segmentation backbone network for inputting stroke infarction lesion segmentation maps to obtain segmentation mask features (mask features)), a feature fusion module connected to the output ends of the two backbone networks, and a risk prediction head module connected to the output end of the feature fusion module; Figure 2 As shown, the backbone network includes: an initialization layer and a residual module connected in sequence; the residual module includes N residual blocks (ResBlock) connected by residuals; the residual block includes M convolution modules connected by residuals, and the convolution module includes multiple first convolution layers connected in sequence (the convolution kernel size is 3, the step size is 2, and the padding is 1). The structures of the two backbone network models are consistent, and they are used to extract features from magnetic resonance images and stroke infarction lesion segmentation maps respectively.
[0039] Step 203: Obtaining a score of post-stroke dysphagia for the patient according to the probability of post-stroke dysphagia for the patient.
[0040] Implementing the above steps can both make predictions in advance and improve the accuracy of assessments.
[0041] In an exemplary embodiment, Figure 3 As shown, the input end of the first convolution module in the residual block is connected to the input end of the second convolution module, which is the input end of the residual block; the output end of the mth convolution module is connected to the input end of the m+1th convolution module and the output end of the m+1th convolution module, and the output end of the last convolution module is the output end of the residual module; 1≤m<M. When the convolution module includes two convolution layers, and M=2, the input end of the first first convolution layer is connected to the output end of the second first convolution layer as the input end of the residual block; the output end of the first first convolution layer is connected to the input end of the second first convolution layer; the output end of the second first convolution layer is connected to the input end of the third first convolution layer and the output end of the fourth first convolution layer; the output end of the third first convolution layer is connected to the input end of the fourth first convolution layer; the output end of the fourth first convolution layer serves as the output end of the residual block.
[0042] In an exemplary embodiment, the input end of the first residual block in the residual module is the input end of the residual module, the output end of the nth residual block is connected to the input end of the n+1th residual block through two parallel channels; the output end of the Nth residual block is the output end of the residual module, 1≤n<N. When N=4, the input end of the first residual block is the input end of the residual module; the output end of the first residual block is connected to the input end of the second residual block through two parallel channels; the output end of the second residual block is connected to the input end of the third residual block through two parallel channels; the output end of the third residual block is connected to the input end of the fourth residual block through two parallel channels; the output end of the fourth residual block is the output end of the residual module.
[0043] In an exemplary embodiment, the residual module further includes: N-1 second convolutional layers (the convolution kernel size is 1); the output of the nth residual block is directly connected to the input of the n+1th residual block, and the output of the nth residual block is also connected to the input of the n+1th residual block through the nth second convolutional layer. When N=4, the output of the first residual block is directly connected to the input of the second residual block, and the output of the first residual block is also connected to the input of the second residual block through the first second convolutional layer; the output of the second residual block is directly connected to the input of the third residual block, and the output of the second residual block is also connected to the input of the third residual block through the second second convolutional layer; the output of the third residual block is directly connected to the input of the fourth residual block, and the output of the third residual block is also connected to the input of the fourth residual block through the third second convolutional layer. In order to deal with the problem of mismatch of feature dimensions between residual blocks, a convolutional layer of the second convolutional layer is added between them to achieve dimensional adaptation.
[0044] In an exemplary embodiment, the process of training the dysphagia risk score dual-path model is as follows:
[0045] A post-stroke dysphagia dataset is obtained; the post-stroke dysphagia dataset includes multiple samples, each sample includes a true label of the sample patient and a brain magnetic resonance image and a stroke infarction lesion segmentation map of the sample patient; the true label is whether post-stroke dysphagia occurs or whether post-stroke dysphagia does not occur.
[0046] Taking minimizing the loss function as the goal, the dysphagia risk score dual-path model is trained according to the post-stroke dysphagia dataset to obtain a dysphagia risk score model.
[0047] In an exemplary embodiment, the initialization layer includes a third convolutional layer, a batch normalization layer BN, and an activation layer ReLU which are connected in sequence.
[0048] In an exemplary embodiment, before obtaining the post-stroke dysphagia dataset, the method further includes: obtaining an original dataset and preprocessing the original dataset.
[0049] The original data set includes sample imaging data of stroke patients with different post-stroke dysphagia conditions, including sample diffusion magnetic resonance images and stroke infarction lesion segmentation maps.
[0050] The image data in the original dataset were preprocessed, and the preprocessed images were used to construct a post-stroke dysphagia dataset.
[0051] The preprocessing operation includes: image cropping-supplementing operation and image data normalization. Specifically, the image cropping-supplementing operation aims to regularize the input image to a size of 256*256*24, perform a cropping operation on the original image that exceeds the target size, and pad the original image that is smaller than the target size to the target size, and supplement the voxel value to 0; the image data normalization operation is to perform Gaussian normalization on the voxel value in the image.
[0052] In an exemplary embodiment, the feature fusion module is used to align and fuse the features extracted by each backbone network to obtain a fused feature. The operation formula is as follows:
[0053]
[0054] in, and The brain image features and segmentation mask features extracted by the two backbone networks are respectively, and Softmax() represents the function operation. represents matrix multiplication and · represents element-wise multiplication.
[0055] In an exemplary embodiment, the risk prediction head is composed of an average pooling layer and a fully connected layer connected in sequence, and the output dimension is 1, which represents the final predicted dysphagia risk probability.
[0056] In an exemplary embodiment, the loss function used in training is a cross entropy loss function, and its calculation formula is:
[0057]
[0058] Among them, i represents the sample number in the post-stroke dysphagia dataset, y i represents the true label value of the i-th sample, It represents the predicted value (probability) obtained by inputting the i-th sample into the dual-path model of dysphagia risk score. Finally, the network is trained by reverse gradient propagation and gradient descent algorithms until the network converges, and the model with the highest prediction accuracy is selected to save the parameters to obtain the final dysphagia risk score model.
[0059] In an exemplary embodiment, the score of the patient suffering from post-stroke dysphagia is obtained according to the probability of the patient suffering from post-stroke dysphagia:
[0060] The obtained probability is converted into a percentage system to obtain the final score. The higher the score, the greater the risk of swallowing disorders, which assists doctors in making the diagnosis.
[0061] This application builds a swallowing disorder risk scoring model based on a deep learning model to quickly predict the swallowing disorder risk of post-stroke patients and further assist doctors in diagnosis. This method can automatically extract and predict features based on the rich information in the patient's brain images to obtain the swallowing disorder risk score of the current patient.
[0062] This application has the following technical effects:
[0063] 1. The feature fusion module proposed in this application uses a class attention mechanism The alignment and fusion of multimodal data features are achieved to provide more refined image features for the risk score prediction head.
[0064] 2. This application uses a deep learning algorithm to automatically score the risk of dysphagia in stroke patients' brain images, avoiding the traditional method that relies on the subjective judgment of clinicians and providing more accurate and timely risk predictions.
[0065] 3. This application maximizes the use of multimodal information of patients by combining brain imaging features with segmentation mask features, improving the accuracy and predictive ability of risk scoring. In addition, the model can provide each patient with a personalized risk prediction of dysphagia, thereby helping doctors adjust treatment plans and optimize intervention measures according to the patient's specific situation.
[0066] 4. Compared with traditional instrumental examination methods, this application can achieve rapid and real-time prediction of dysphagia risk, effectively reduce dependence on expensive equipment and complex examination procedures, and can be applied to the early screening and diagnosis of various post-stroke dysphagia patients to improve medical efficiency.
[0067] Based on the same inventive concept, the embodiment of the present application also provides a post-stroke dysphagia risk scoring device for implementing the post-stroke dysphagia risk scoring method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more post-stroke dysphagia risk scoring device embodiments provided below can refer to the limitations of the post-stroke dysphagia risk scoring method above, and will not be repeated here.
[0068] In an exemplary embodiment, a post-stroke dysphagia risk scoring device is provided, comprising:
[0069] The acquisition module is used to acquire the magnetic resonance image of the patient's brain and the stroke infarction lesion segmentation map.
[0070] A prediction module is used to input the magnetic resonance image of the patient's brain and the stroke infarction lesion segmentation map into a dysphagia risk scoring model to obtain the probability of the patient suffering from post-stroke dysphagia; the dysphagia risk scoring model is obtained by training the dysphagia risk scoring dual-path model; the dysphagia risk scoring dual-path model includes: two backbone networks, a feature fusion module connected to the output ends of the two backbone networks, and a risk prediction head module connected to the output end of the feature fusion module; the backbone network includes: an initialization layer and a residual module connected in sequence; the residual module includes N residual blocks connected by residuals; the residual block includes M convolution modules connected by residuals, and the convolution module includes multiple first convolution layers connected in sequence.
[0071] A scoring module is used to obtain a score of post-stroke dysphagia in the patient based on the probability of the patient suffering from post-stroke dysphagia.
[0072] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store post-stroke dysphagia risk scoring data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a post-stroke dysphagia risk scoring method is implemented.
[0073] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the above-mentioned method embodiments when executing the computer program.
[0074] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the above-mentioned method embodiments are implemented.
[0075] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the above-mentioned method embodiments are implemented.
[0076] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0077] In this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.
[0078] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).
[0079] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.
[0080] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0081] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.
Claims
1. A method for scoring the risk of post-stroke dysphagia, characterized in that: The post-stroke dysphagia risk scoring method includes: Obtaining magnetic resonance images of the patient's brain and stroke infarction lesion segmentation maps; Inputting the magnetic resonance image of the patient's brain and the stroke infarction lesion segmentation map into the dysphagia risk scoring model to obtain the probability of the patient suffering from post-stroke dysphagia; the dysphagia risk scoring model is obtained by training the dysphagia risk scoring dual-path model; the dysphagia risk scoring dual-path model comprises: two backbone networks, a feature fusion module connected to the output ends of the two backbone networks, and a risk prediction head module connected to the output end of the feature fusion module; the backbone network comprises: an initialization layer and a residual module connected in sequence; the residual module comprises N residual blocks connected by residuals; the residual block comprises M convolution modules connected by residuals, and the convolution module comprises a plurality of first convolution layers connected in sequence; A score of post-stroke dysphagia occurring in the patient is obtained according to the probability of the patient experiencing post-stroke dysphagia.
2. The method for scoring risk of post-stroke dysphagia according to claim 1, characterized in that: The input end of the first convolution module in the residual block is connected to the input end of the second convolution module, which is the input end of the residual block; the output end of the mth convolution module is connected to the input end of the m+1th convolution module and the output end of the m+1th convolution module, and the output end of the last convolution module is the output end of the residual module; 1≤m<M.
3. The method for scoring risk of post-stroke dysphagia according to claim 1, characterized in that: The input end of the first residual block in the residual module is the input end of the residual module, the output end of the nth residual block is connected to the input end of the n+1th residual block through two parallel channels; the output end of the Nth residual block is the output end of the residual module, 1≤n<N.
4. The method for scoring risk of post-stroke dysphagia according to claim 3, characterized in that: The residual module also includes: N-1 second convolutional layers; the output end of the nth residual block is directly connected to the input end of the n+1th residual block, and the output end of the nth residual block is also connected to the input end of the n+1th residual block through the nth second convolutional layer.
5. The method for scoring risk of post-stroke dysphagia according to claim 1, characterized in that: The process of training the dual-path model for dysphagia risk scoring is as follows: A post-stroke dysphagia dataset is obtained; the post-stroke dysphagia dataset includes multiple samples, each sample includes a true label of the sample patient and a brain magnetic resonance image and a stroke infarction lesion segmentation map of the sample patient; the true label is whether post-stroke dysphagia occurs or whether post-stroke dysphagia does not occur; Taking minimizing the loss function as the goal, the dysphagia risk score dual-path model is trained according to the post-stroke dysphagia dataset to obtain a dysphagia risk score model.
6. The method for scoring risk of post-stroke dysphagia according to claim 1, characterized in that: The initialization layer includes a third convolutional layer, a batch normalization layer, and an activation layer which are connected in sequence.
7. A post-stroke dysphagia risk scoring device, characterized in that: The post-stroke dysphagia risk scoring device comprises: An acquisition module, used for acquiring a magnetic resonance image of the patient's brain and a stroke infarction lesion segmentation map; A prediction module is used to input the magnetic resonance image of the patient's brain and the stroke infarction lesion segmentation map into a dysphagia risk scoring model to obtain the probability of the patient suffering from post-stroke dysphagia; the dysphagia risk scoring model is obtained by training the dysphagia risk scoring dual-path model; the dysphagia risk scoring dual-path model includes: two backbone networks, a feature fusion module connected to the output ends of the two backbone networks, and a risk prediction head module connected to the output end of the feature fusion module; the backbone network includes: an initialization layer and a residual module connected in sequence; the residual module includes N residual blocks connected by residuals; the residual block includes M convolution modules connected by residuals, and the convolution module includes a plurality of first convolution layers connected in sequence; A scoring module is used to obtain a score of post-stroke dysphagia in the patient based on the probability of the patient suffering from post-stroke dysphagia.
8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the post-stroke dysphagia risk scoring method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for scoring the risk of post-stroke dysphagia according to any one of claims 1 to 6 is implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for scoring the risk of post-stroke dysphagia according to any one of claims 1 to 6 is implemented.