A neurological prognosis evaluation system

Abstract

This invention discloses a neurological prognostic assessment system, comprising a data acquisition module, an image enhancement module, an image segmentation module, a neurological prognostic assessment model building module, and a neurological prognostic assessment module. This invention belongs to the field of prognostic assessment technology, specifically referring to a neurological prognostic assessment system. This scheme utilizes fuzzy matrix transformation and a neighborhood average method to more finely adjust the overall brightness of the image; recursively applies the sigmoid membership function to enhance sharpness and adjust contrast; introduces fuzzy clustering methods for more accurate image segmentation; uses the Xie-Beni function to evaluate and optimize the clustering results; introduces a residual connection structure; adds basic block modules after convolutional layers for extracting and transforming image features; simplifies fully connected layers; and designs activation functions that effectively mitigate gradient problems that may occur in deep networks. Weight normalization is introduced, which helps improve the accuracy and stability of model training.

Description

Technical Field

[0001] This invention relates to the field of prognostic assessment technology, specifically to a neurological prognostic assessment system. Background Technology

[0002] A neurological prognostic assessment system is a system that combines deep learning and medical image processing technologies to assess and predict the neurological prognosis of patients. The main goal of this system is to provide accurate prognostic assessment grades by analyzing brain CT scan images to assist clinicians in decision-making. However, general neurological prognostic assessment systems suffer from several problems: poor original image visual quality; over-enhancement or loss of detail introduced by common image enhancement methods; inability to handle complex backgrounds and fine structures, and inability to preserve internal image structure and features during processing; and overfitting, computational burden, inability to handle large-scale data, and poor model robustness. Summary of the Invention

[0003] To address the aforementioned issues and overcome the shortcomings of existing technologies, this invention provides a neurological prognostic assessment system. Addressing the problems of poor original image visual quality and over-enhancement or detail loss introduced by general image enhancement methods in conventional neurological prognostic assessment systems, this solution utilizes fuzzy matrix transformation and a neighborhood average-based method to more finely adjust the overall brightness of the image. It recursively applies a sigmoid membership function to enhance sharpness and adjust contrast, using a non-linear approach to more precisely control pixel value changes. This enhances contrast while maintaining image naturalness and detail, resulting in a clearer and smoother image. Inverse spatial threshold transformation better preserves the overall grayscale dynamic range of the image, avoiding the over-contrast and grayscale loss that may occur with traditional histogram equalization. Addressing the inability of conventional neurological prognostic assessment systems to handle complex backgrounds and fine structures, and to preserve internal image structure and features during processing, this solution introduces a fuzzy clustering method for more accurate image segmentation. Fuzzy clustering considers the distance between pixels and cluster centers, as well as membership degrees, enabling more flexible handling of complex image structures and background noise. The Xie-Beni function is used to process the clustering results. The system evaluates and optimizes the data to maintain the structural integrity and accuracy of the segmented images. Cluster centers are updated based on cluster weights and Mahalanobis distance to better reflect the contribution of different pixels to the cluster centers. Evaluation weights are introduced to balance cluster compactness and separability, allowing for adjustments to the accuracy and clarity of clustering results based on application scenarios and needs. Segmented images are grouped by cluster centers to provide a more accurate input dataset. Addressing the issues of overfitting, excessive computational burden, inability to handle large-scale data, and poor robustness in general neurological prognostic assessment systems, this solution introduces residual connections to effectively solve the vanishing and exploding gradient problems. Basic block modules are added after convolutional layers for extracting and transforming image features. Residual connections maintain a reasonable number of parameters, avoiding overfitting and excessive computational burden. Simplifying fully connected layers reduces the risk of overfitting while accelerating model computation. The designed activation function effectively mitigates gradient problems that may occur in deep networks, improving training stability and convergence speed. The introduction of weight normalization techniques, based on local response normalization layers, helps improve the accuracy and stability of model training.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides a neurological prognostic assessment system, including a data acquisition module, an image enhancement module, an image segmentation module, a neurological prognostic assessment model establishment module, and a neurological prognostic assessment module;

[0005] The data acquisition module collects historical prognostic assessment data;

[0006] The image enhancement module uses fuzzy matrix transformation to adjust the overall brightness of the image based on the neighborhood average method; it recursively applies the S-shaped membership function to enhance sharpness and adjust contrast, and performs inverse spatial threshold transformation.

[0007] The image segmentation module introduces fuzzy clustering to segment the image; fuzzy clustering considers the distance between pixels and cluster centers as well as membership degree; the Xie-Beni function is used to evaluate and optimize the clustering results; the cluster centers are updated based on cluster weights and Mahalanobis distance; and evaluation weights are introduced to balance cluster compactness and separability.

[0008] The neurology prognostic assessment model module introduces a residual connection structure on the basis of a convolutional neural network; adds a basic block module after the convolutional layer; simplifies the fully connected layer and designs the activation function; and introduces a weight normalization technique.

[0009] The neurological prognostic assessment module is based on the established neurological prognostic assessment model to perform prognostic assessments on brain CT scan images.

[0010] Furthermore, in the data acquisition module, the historical prognostic assessment data includes historical brain CT scan image data and prognostic assessment levels.

[0011] Furthermore, the image enhancement module specifically includes the following:

[0012] The formula for converting a grayscale image into a blur matrix is ​​as follows:

[0013] ;

[0014] In the formula, X ij These are the elements of the transformed fuzzy matrix; h and w are the height and width of the grayscale image, respectively; i and j are the indices of the height and width, and also the number of rows and columns of the matrix; It is the average value of the neighborhood of the corresponding pixel in the original grayscale image; x ij It is the grayscale value of the original image;

[0015] To enhance image contrast, an S-shaped membership function is introduced to adjust the image contrast. The formula used is as follows:

[0016] ;

[0017] ;

[0018] In the formula, These are the pixel values ​​after initial contrast enhancement; s1(·) is the value of the first sigmoid membership function pair. Further processing of the value;

[0019] To enhance image sharpness, an S-shaped membership function is recursively applied. The formula used is as follows:

[0020] ;

[0021] In the formula, s r (q ij ) and s r-1 (q ij The r-th and (r-1)-th times the sigmoid membership function is used on the pairs. Further processing of the value;

[0022] The fuzzy values ​​are converted to spatial domain values ​​through inverse transform, using the following formula:

[0023] ;

[0024] In the formula, Ky is the transformed spatial domain value, i.e., the final pixel value; It is the fuzzy value after the recursion ends; α is a parameter used to control the transformation.

[0025] Furthermore, the image segmentation module specifically includes the following steps:

[0026] Initialize the cluster centers, maximum number of iterations, and fuzziness index;

[0027] Pixel assignment involves calculating the membership value of each pixel and assigning pixels based on these membership values. The formula for calculating the membership value is as follows:

[0028] ;

[0029] In the formula, It is the membership value of the J-th pixel to the I-th cluster center; x J It is the grayscale value of the Jth pixel; v I and v k , respectively, are the gray values ​​of the i-th and k-th cluster centers; c is the number of cluster centers; m is the fuzziness index;

[0030] Cluster optimization uses the Xie-Beni function to maintain the compactness and separation strength of the clusters. The formula used is as follows:

[0031] ;

[0032] In the formula, x Ib (U) is the clustering optimization performed on the clustering result U using the Xie-Beni function; v p It is the p-th cluster center; n is the total number of pixels;

[0033] The cluster centers are updated based on the cluster weights, using the following formula:

[0034] ;

[0035] ;

[0036] In the formula, w IJ It is the weight of the Jth pixel to the Ith cluster center; and It is the normalized membership degree; It is the Mahalanobis distance from the i-th cluster center to the j-th pixel; It is the Mahalanobis distance from the k-th cluster center to the J-th pixel; x I It is the grayscale value of the i-th pixel;

[0037] Cluster evaluation, based on membership degree, performs fuzzy cluster assignment and uses membership degree values ​​to evaluate the effectiveness of clustering, as shown below:

[0038] ;

[0039] In the formula, K is the cluster evaluation value; β is the evaluation weight;

[0040] Clustering determination is performed with a pre-set evaluation threshold. If the current cluster evaluation value is higher than the evaluation threshold, the process proceeds to segmentation; if the maximum number of iterations is reached, the process proceeds to initialization; otherwise, the process proceeds to pixel allocation.

[0041] Segmentation involves grouping pixels according to their cluster centers to obtain different segmented images; the segmented images are then selected to form an image dataset, which is further divided into a training set and a test set.

[0042] Furthermore, the neurological prognostic assessment model building module constructs a convolutional neural network, where the input layer receives segmented image data and the output layer outputs the predicted prognostic assessment level; it also optimizes the network structure; trains the model based on the training set and verifies the model's performance based on the test set; if the prediction accuracy of the trained model on the test set reaches a threshold, the neurological prognostic assessment model is considered established; the network structure optimization specifically includes the following:

[0043] Residual connections, introduced as a structure, allow skip connections and help solve the vanishing and exploding gradient problems; this structure allows data to be passed directly between network layers, which helps train deeper networks and improves accuracy.

[0044] Basic block modules are added after convolutional layers to extract and transform image features. Basic block modules contain convolutional layers and activation functions, and maintain a reasonable number of parameters through residual connections, avoiding the problems of overfitting and excessive computational burden.

[0045] Simplify the fully connected layer by reducing three fully connected layers into one; this reduces the number of model parameters and complexity while improving training and inference efficiency.

[0046] Activation function design, the activation function is expressed as This is to mitigate gradient problems that may occur in deep networks and improve training stability and convergence speed; where x is the input to the activation function. It is an exponential function; γ is the adjustment parameter;

[0047] Weight normalization, introduced on the basis of local response normalization layer, helps to improve the accuracy and stability of model training.

[0048] Furthermore, the neurological prognostic assessment module, based on the established neurological prognostic assessment model, acquires brain CT scan images in real time, processes them through the image enhancement module and image segmentation module, and inputs them into the neurological prognostic assessment model. The prognostic assessment level predicted by the neurological prognostic assessment model is used as the neurological prognostic assessment result corresponding to the image.

[0049] The beneficial effects achieved by the present invention using the above solution are as follows:

[0050] (1) In view of the problems of poor visual quality of original images in general neurological prognostic assessment systems and the over-enhancement or loss of detail introduced by general image enhancement methods, this scheme uses fuzzy matrix transformation and neighborhood average method to adjust the overall brightness of the image more finely; recursively applying sigmoid membership function to enhance sharpness and adjust contrast, and controlling the change of pixel value more finely through nonlinear means, which can improve contrast while maintaining the naturalness and detail of the image, making the image clearer and smoother; based on spatial threshold inverse transformation, the overall gray-scale dynamic range of the image is better preserved, avoiding the over-contrast and gray-scale loss that may occur with traditional histogram equalization.

[0051] (2) To address the problem that general neurological prognostic assessment systems cannot handle complex backgrounds and fine structures in images, and cannot preserve the internal structure and features of images during processing, this solution introduces a fuzzy clustering method to segment images more accurately. Fuzzy clustering considers the distance between pixels and cluster centers as well as membership, and can handle complex image structures and background noise more flexibly. The Xie-Beni function is used to evaluate and optimize the clustering results, maintaining the structural integrity and accuracy of the segmented images. The cluster centers are updated based on cluster weights and Mahalanobis distance to better reflect the contribution of different pixels to the cluster centers. Evaluation weights are introduced to balance cluster compactness and separability. This method can adjust the accuracy and clarity of the clustering results according to the application scenario and requirements. The segmented images are grouped according to cluster centers to provide a more accurate input dataset.

[0052] (3) In view of the problems of overfitting, excessive computational burden, inability to handle large-scale data, and poor robustness of general neurological prognostic assessment systems, this scheme introduces residual connection structure to effectively solve the problems of gradient vanishing and gradient explosion; basic block modules are added after convolutional layers to extract and transform image features; reasonable number of parameters are maintained through residual connections to avoid overfitting and excessive computational burden; simplified fully connected layers reduce the risk of overfitting and speed up the calculation of the model; the designed activation function can effectively alleviate the gradient problem that may occur in deep networks and improve the stability and convergence speed of training; the introduction of weight normalization technology on the basis of local response normalization layer helps to improve the accuracy and stability of model training. Attached Figure Description

[0053] Figure 1 A schematic diagram of a neurological prognostic assessment system provided by the present invention;

[0054] Figure 2 This is a flowchart illustrating the image enhancement module.

[0055] Figure 3 This is a flowchart illustrating the image segmentation module.

[0056] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0057] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0058] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0059] Example 1, see Figure 1 The present invention provides a neurological prognostic assessment system, including a data acquisition module, an image enhancement module, an image segmentation module, a neurological prognostic assessment model establishment module, and a neurological prognostic assessment module;

[0060] The data acquisition module collects historical prognostic assessment data and sends the data to the image enhancement module.

[0061] The image enhancement module receives data sent by the data acquisition module, uses fuzzy matrix transformation, and adjusts the overall brightness of the image based on the neighborhood average method; it recursively applies the S-shaped membership function to enhance sharpness and adjust contrast, and performs inverse spatial threshold transformation; and then sends the data to the image segmentation module.

[0062] The image segmentation module receives data sent by the image enhancement module, introduces fuzzy clustering to segment the image; fuzzy clustering considers the distance between pixels and cluster centers as well as membership degree; the Xie-Beni function is used to evaluate and optimize the clustering results; the cluster centers are updated based on cluster weights and Mahalanobis distance; evaluation weights are introduced to balance cluster compactness and separability; and the data is sent to the neurology prognostic assessment model building module.

[0063] The neurology prognostic assessment model building module receives data sent by the image segmentation module, introduces a residual connection structure on the basis of a convolutional neural network, adds a basic block module after the convolutional layer, simplifies the fully connected layer and designs the activation function, introduces weight normalization technology, and sends the data to the neurology prognostic assessment module.

[0064] The neurological prognostic assessment module receives data sent by the neurological prognostic assessment model establishment module and performs prognostic assessment of brain CT scan images based on the established neurological prognostic assessment model.

[0065] Example 2, see Figure 1 This embodiment is based on the above embodiment. In the data acquisition module, the historical prognostic assessment data includes historical brain CT scan image data and prognostic assessment level.

[0066] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment, and the image enhancement module specifically includes the following:

[0067] The formula for converting a grayscale image into a blur matrix is ​​as follows:

[0068] ;

[0069] In the formula, X ij These are the elements of the transformed fuzzy matrix; h and w are the height and width of the grayscale image, respectively; i and j are the indices of the height and width, and also the number of rows and columns of the matrix; It is the average value of the neighborhood of the corresponding pixel in the original grayscale image; x ij It is the grayscale value of the original image;

[0070] To enhance image contrast, an S-shaped membership function is introduced to adjust the image contrast. The formula used is as follows:

[0071] ;

[0072] ;

[0073] In the formula, These are the pixel values ​​after initial contrast enhancement; s1(·) is the value of the first sigmoid membership function pair. Further processing of the value;

[0074] To enhance image sharpness, an S-shaped membership function is recursively applied. The formula used is as follows:

[0075] ;

[0076] In the formula, s r (q ij ) and s r-1 (q ij The r-th and (r-1)-th times the sigmoid membership function is used on the pairs. Further processing of the value;

[0077] The fuzzy values ​​are converted to spatial domain values ​​through inverse transform, using the following formula:

[0078] ;

[0079] In the formula, Ky is the transformed spatial domain value, i.e., the final pixel value; It is the fuzzy value after the recursion ends; α is a parameter used to control the transformation.

[0080] By performing the above operations, this solution addresses the issues of poor original image visual quality in general neurological prognostic assessment systems and the over-enhancement or detail loss introduced by general image enhancement methods. It utilizes fuzzy matrix transformation and a neighborhood average-based method to more finely adjust the overall brightness of the image. Recursively applying the sigmoid membership function enhances sharpness and adjusts contrast, using a non-linear approach to more precisely control pixel value changes. This improves contrast while maintaining image naturalness and detail, resulting in a clearer and smoother image. Furthermore, the inverse spatial threshold transformation better preserves the overall grayscale dynamic range of the image, avoiding the over-contrast and grayscale loss that may occur with traditional histogram equalization.

[0081] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the image segmentation module specifically includes the following steps:

[0082] Initialize the cluster centers, maximum number of iterations, and fuzziness index;

[0083] Pixel assignment involves calculating the membership value of each pixel and assigning pixels based on these membership values. The formula for calculating the membership value is as follows:

[0084] ;

[0085] In the formula, It is the membership value of the J-th pixel to the I-th cluster center; x J It is the grayscale value of the Jth pixel; v I and v k , respectively, are the gray values ​​of the i-th and k-th cluster centers; c is the number of cluster centers; m is the fuzziness index;

[0086] Cluster optimization uses the Xie-Beni function to maintain the compactness and separation strength of the clusters. The formula used is as follows:

[0087] ;

[0088] In the formula, x Ib (U) is the clustering optimization performed on the clustering result U using the Xie-Beni function; v pIt is the p-th cluster center; n is the total number of pixels;

[0089] The cluster centers are updated based on the cluster weights, using the following formula:

[0090] ;

[0091] ;

[0092] In the formula, w IJ It is the weight of the Jth pixel to the Ith cluster center; and It is the normalized membership degree; It is the Mahalanobis distance from the i-th cluster center to the j-th pixel; It is the Mahalanobis distance from the k-th cluster center to the J-th pixel; x I It is the grayscale value of the i-th pixel;

[0093] Cluster evaluation, based on membership degree, performs fuzzy cluster assignment and uses membership degree values ​​to evaluate the effectiveness of clustering, as shown below:

[0094] ;

[0095] In the formula, K is the cluster evaluation value; β is the evaluation weight;

[0096] Clustering determination is performed with a pre-set evaluation threshold. If the current cluster evaluation value is higher than the evaluation threshold, the process proceeds to segmentation; if the maximum number of iterations is reached, the process proceeds to initialization; otherwise, the process proceeds to pixel allocation.

[0097] Segmentation involves grouping pixels according to their cluster centers to obtain different segmented images; the segmented images are then selected to form an image dataset, which is further divided into a training set and a test set.

[0098] By performing the above operations, this solution addresses the problem that general neurological prognostic assessment systems cannot handle complex backgrounds and fine structures in images, and cannot preserve the internal structure and features of images during processing. It introduces a fuzzy clustering method for more accurate image segmentation. Fuzzy clustering considers the distance between pixels and cluster centers, as well as membership, enabling more flexible handling of complex image structures and background noise. The Xie-Beni function is used to evaluate and optimize the clustering results, maintaining the structural integrity and accuracy of the segmented images. Cluster centers are updated based on cluster weights and Mahalanobis distance, better reflecting the contribution of different pixels to the cluster centers. Evaluation weights are introduced to balance cluster compactness and separability; this method can adjust the accuracy and clarity of the clustering results according to the application scenario and requirements. The segmented images are grouped by cluster centers, providing a more accurate input dataset.

[0099] Example 5, see Figure 1 This embodiment is based on the above embodiment. The neurological prognostic assessment model building module constructs a convolutional neural network, which receives segmented image data as input and outputs the predicted prognostic assessment level as output. The network structure is optimized; the model is trained based on the training set and its performance is verified based on the test set. If the prediction accuracy of the trained model on the test set reaches a threshold, the neurological prognostic assessment model is successfully built. The network structure optimization specifically includes the following:

[0100] Residual connections, introduced as a structure, allow skip connections and help solve the vanishing and exploding gradient problems; this structure allows data to be passed directly between network layers, which helps train deeper networks and improves accuracy.

[0101] Basic block modules are added after convolutional layers to extract and transform image features. Basic block modules contain convolutional layers and activation functions, and maintain a reasonable number of parameters through residual connections, avoiding the problems of overfitting and excessive computational burden.

[0102] Simplify the fully connected layer by reducing three fully connected layers into one; this reduces the number of model parameters and complexity while improving training and inference efficiency.

[0103] Activation function design, the activation function is expressed as This is to mitigate gradient problems that may occur in deep networks and improve training stability and convergence speed; where x is the input to the activation function. It is an exponential function; γ is the adjustment parameter;

[0104] Weight normalization, introduced on the basis of local response normalization layer, helps to improve the accuracy and stability of model training.

[0105] By performing the above operations, this solution addresses the problems of overfitting, excessive computational burden, inability to handle large-scale data, and poor robustness in general neurological prognostic assessment systems. It introduces a residual connection structure to effectively solve the vanishing and exploding gradient problems. Basic block modules are added after convolutional layers for extracting and transforming image features. Residual connections maintain a reasonable number of parameters, avoiding overfitting and excessive computational burden. Simplifying fully connected layers reduces the risk of overfitting while accelerating model computation. The designed activation function effectively mitigates gradient problems that may occur in deep networks, improving training stability and convergence speed. The introduction of weight normalization techniques, based on local response normalization layers, helps improve the accuracy and stability of model training.

[0106] Example 6, see Figure 1 This embodiment is based on the above embodiment. The neurological prognostic assessment module acquires brain CT scan images in real time based on the established neurological prognostic assessment model. After being processed by the image enhancement module and the image segmentation module, the images are input into the neurological prognostic assessment model. The prognostic assessment level predicted by the neurological prognostic assessment model is used as the neurological prognostic assessment result corresponding to the image.

[0107] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0108] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0109] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A neurological prognostic assessment system, characterized in that: The system includes a data acquisition module, an image enhancement module, an image segmentation module, a neurological prognostic assessment model establishment module, and a neurological prognostic assessment module. The data acquisition module collects historical prognostic assessment data and sends the data to the image enhancement module. The image enhancement module uses fuzzy matrix transformation to adjust the overall brightness of the image based on the neighborhood average method; it recursively applies the S-shaped membership function to enhance sharpness and adjust contrast, and performs inverse spatial threshold transformation; and then sends the data to the image segmentation module. The image segmentation module introduces fuzzy clustering to segment the image; fuzzy clustering considers the distance between pixels and cluster centers as well as membership degree; the Xie-Beni function is used to evaluate and optimize the clustering results; the cluster centers are updated based on cluster weights and Mahalanobis distance; evaluation weights are introduced to balance cluster compactness and separability; and the data is sent to the neurology prognostic assessment model building module. The neurology prognostic assessment model module introduces a residual connection structure on the basis of a convolutional neural network; adds a basic block module after the convolutional layer; simplifies the fully connected layer and designs the activation function; introduces weight normalization technology; and sends the data to the neurology prognostic assessment module. The neurological prognostic assessment module is based on the established neurological prognostic assessment model to perform prognostic assessments on brain CT scan images.

2. The neurological prognostic assessment system according to claim 1, characterized in that: The neurological prognostic assessment model building module constructs a convolutional neural network. The input layer receives segmented image data, and the output layer outputs the predicted prognostic assessment level. The network structure is optimized. The model is trained on a training set and its performance is validated on a test set. If the prediction accuracy of the trained model on the test set reaches a threshold, the neurological prognostic assessment model is considered complete. The network structure optimization specifically includes the following: Residual connections, introduced as a structure, allow skip connections; this structure allows data to be passed directly between network layers. Basic block modules are added after convolutional layers to extract and transform image features; Basic block modules contain convolutional layers and activation functions, connected via residuals; Simplify the fully connected layer by reducing three fully connected layers to one. Activation function design, the activation function is expressed as Where x is the input to the activation function; It is an exponential function; γ is the adjustment parameter; Weight normalization is introduced, and weight normalization is applied on the basis of the local response normalization layer.

3. The neurological prognostic assessment system according to claim 1, characterized in that: The image segmentation module specifically includes the following: Initialize the cluster centers, maximum number of iterations, and fuzziness index; Pixel assignment involves calculating the membership value of each pixel and assigning pixels based on these membership values. The formula for calculating the membership value is as follows: ； In the formula, It is the membership value of the J-th pixel to the I-th cluster center; x J It is the grayscale value of the Jth pixel; v I and v k , respectively, are the gray values ​​of the i-th and k-th cluster centers; c is the number of cluster centers; m is the fuzziness index; Cluster optimization uses the Xie-Beni function to maintain the compactness and separation strength of the clusters. The formula used is as follows: ； In the formula, x Ib (U) is the clustering optimization performed on the clustering result U using the Xie-Beni function; v p It is the p-th cluster center; where n is the total number of pixels; The cluster centers are updated based on the cluster weights, using the following formula: ； ； In the formula, w IJ It is the weight of the Jth pixel to the Ith cluster center; and It is the normalized membership degree; It is the Mahalanobis distance from the i-th cluster center to the j-th pixel; It is the Mahalanobis distance from the k-th cluster center to the J-th pixel; x I It is the grayscale value of the i-th pixel; Cluster evaluation, based on membership degree, performs fuzzy cluster assignment and uses membership degree values ​​to evaluate the effectiveness of clustering, as shown below: ； In the formula, K is the cluster evaluation value; β is the evaluation weight; Clustering judgment is performed with a pre-set evaluation threshold. If the current cluster evaluation value is higher than the evaluation threshold, the process will proceed to segmentation. If the maximum number of iterations is reached, proceed to initialization; otherwise, proceed to pixel allocation. Segmentation involves grouping pixels according to their cluster centers to obtain different segmented images; the segmented images are then selected to form an image dataset, which is further divided into a training set and a test set.

4. The neurological prognostic assessment system according to claim 1, characterized in that: The image enhancement module specifically includes the following: The formula for converting a grayscale image into a blur matrix is ​​as follows: ； In the formula, X ij These are the elements of the transformed fuzzy matrix; h and w are the height and width of the grayscale image, respectively; i and j are the indices of the height and width, and also the number of rows and columns of the matrix; It is the average value of the neighborhood of the corresponding pixel in the original grayscale image; x ij It is the grayscale value of the original image; To enhance image contrast, an S-shaped membership function is introduced to adjust the image contrast. The formula used is as follows: ； ； In the formula, These are the pixel values ​​after initial contrast enhancement; s1(·) is the value of the first sigmoid membership function pair. Further processing of the value; To enhance image sharpness, an S-shaped membership function is recursively applied. The formula used is as follows: ； In the formula, s r (q ij ) and s r-1 (q ij The r-th and (r-1)-th times respectively use the sigmoid membership function pairs. Further processing of the value; The fuzzy values ​​are converted to spatial domain values ​​through inverse transform, using the following formula: ； In the formula, Ky is the transformed spatial domain value, i.e., the final pixel value; It is the fuzzy value after the recursion ends; α is a parameter used to control the transformation.

5. A neurological prognostic assessment system according to claim 1, characterized in that: In the data acquisition module, the historical prognostic assessment data includes historical brain CT scan image data and prognostic assessment level.

6. The neurological prognostic assessment system according to claim 1, characterized in that: The neurological prognostic assessment module is based on the established neurological prognostic assessment model. It acquires brain CT scan images in real time, processes them through the image enhancement module and the image segmentation module, and then inputs them into the neurological prognostic assessment model. The prognostic assessment level predicted by the neurological prognostic assessment model is used as the neurological prognostic assessment result corresponding to the image.

Images