Method, System, Medium and Device for Processing Clinical Data of Alzheimer's Disease

Abstract

The present invention provides a method, system, medium and device for processing clinical data of Alzheimer's disease, including obtaining three-dimensional image data containing only brain parenchyma data; resampling according to a first sampling rule to obtain resampled image data; cropping to obtain final image data with blank slices and zero element values at the image edges removed; normalizing to obtain normalized image data; using a neural network based on U-net to extract features from the processed data to obtain output features; fusing the output features to obtain global features; classifying the global features for Alzheimer's disease to obtain an Alzheimer's disease classification model; calculating the influence degree of each part in the image on the diagnostic result of the Alzheimer's disease classification model by means of backpropagation, and displaying it through heat visualization. The present invention is used for classifying clinical data of Alzheimer's disease and visualizing the degree of attention of the model to each part of the brain region during the diagnosis process.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly to a method, system, medium and device for processing clinical data of Alzheimer's disease. Background Art

[0002] Alzheimer's disease (AD), also known as senile dementia, is a degenerative disease of the central nervous system. It has a latent onset and a chronic progressive course, and is the most common type of senile dementia. Its main manifestations include progressive memory impairment, cognitive dysfunction, personality changes, language disorders and other neuropsychiatric symptoms, which seriously affect social, occupational and life functions. The etiology and pathogenesis of AD have not been elucidated. The characteristic pathological changes are extracellular senile plaques formed by β-amyloid protein deposition, neurofibrillary tangles formed by hyperphosphorylation of tau protein in nerve cells, and neuron loss accompanied by gliosis, etc.

[0003] Convolutional neural networks have achieved great results in many fields, and this technology has been successfully introduced into the diagnosis of Alzheimer's disease. However, ordinary convolutional neural network models cannot diagnose well.

[0004] Therefore, it is hoped to solve the problems that the existing convolutional neural network models cannot classify Alzheimer's disease well, which limits their promotion and use in clinical practice, and cannot intuitively explain the classification results of Alzheimer's disease. Summary of the Invention

[0005] In view of the above-mentioned disadvantages of the prior art, the purpose of the present invention is to provide a method, system, medium and device for processing clinical data of Alzheimer's disease, so as to solve the problems that the existing convolutional neural network models in the prior art cannot classify Alzheimer's disease well, which limits their promotion and use in clinical practice, and cannot intuitively display the classification results of Alzheimer's disease.

[0006] To achieve the above and other related objectives, the present invention provides a method for processing clinical data of Alzheimer's disease, including the following steps: obtaining three-dimensional image data containing brain data from the ADNI clinical dataset based on a first preset rule; performing skull stripping on the three-dimensional image data to obtain brain three-dimensional image data containing only brain parenchyma; resampling the brain three-dimensional image data according to a first sampling rule to obtain resampled image data; cropping the resampled image data to obtain final image data with blank slices and image edges removed; normalizing the final image data to obtain normalized image data; using a neural network model based on U-net to extract features from the normalized image data to obtain output features; using global pooling operation to fuse the output features to obtain global features; using a classifier to classify the global features for Alzheimer's disease to obtain an Alzheimer's disease classification result; calculating the influence degree of each brain region in the three-dimensional image data on the Alzheimer's disease classification result by the method of backpropagation based on the Alzheimer's disease classification result, and displaying it through heat visualization.

[0007] In an embodiment of the present invention, the performing skull stripping on the three-dimensional image data to obtain brain three-dimensional image data containing only brain parenchyma includes:

[0008] Using a standard template to remove the skull-related parts in the three-dimensional image data, so as to obtain brain three-dimensional image data containing only brain parenchyma-related parts.

[0009] In an embodiment of the present invention, the resampling the brain three-dimensional image data according to a first sampling rule to obtain resampled image data includes: resampling through a first sampling formula or a second sampling formula to obtain resampled image data; the first sampling formula is:

[0010] P (x，y，z) =Q(2x, 2y, 2z)

[0011] Wherein, P (x，y，z) represents the pixel value at the coordinate (x, y, z) obtained by resampling through the first sampling formula, and Q(2x, 2y, 2z) represents the pixel value at the coordinate (2x, 2y, 2z) in the image of the three-dimensional image data; the second sampling formula is:

[0012] P (x，y，z) =Q(2x + 1, 2y + 1, 2z + 1)

[0013] Wherein, P (x，y，z)P(x, y, z) represents the pixel value at the coordinate (x, y, z) obtained by resampling through the second sampling formula, and Q(2x + 1, 2y + 1, 2z + 1) represents the pixel value at the coordinate (2x + 1, 2y + 1, 2z + 1) in the image of the three-dimensional brain image data.

[0014] In an embodiment of the present invention, the cropping the resampled image data to obtain the final image data with blank slices and image edges removed includes: taking the image center of the resampled image data as the origin and cropping with a cropping range of (64, 104, 80) unit pixels to obtain the image of the final image data, so as to obtain the corresponding final image data based on the image of the resampled image data.

[0015] In an embodiment of the present invention, the normalizing the final image data to obtain the normalized image data includes: normalizing the final image data based on the normalization formula to obtain the normalized image data; the normalization formula is;

[0016]

[0017] where P (x，y，z) represents the pixel value at the coordinate (x, y, z) in the image of the normalized image data, OP (x，y，z) represents the pixel value at the coordinate (x, y, z) in the image of the final image data, min(P) represents the minimum pixel value in the image of the final image data, and max(P) represents the maximum pixel value in the image of the final image data.

[0018] In an embodiment of the present invention, the using the U-net based neural network model to extract features and classify the normalized image data to obtain the output features includes: the U-net neural network model includes eight convolutional blocks, and the number of channels of each convolutional block is 4, 16, 32, 64, 64, 32, 16, 8 respectively. And each convolutional block includes a convolutional layer to extract features; using the rectified linear activation function to process the features; using the batch normalization function to normalize the features.

[0019] In an embodiment of the present invention, using the global pooling operation to fuse the output features to obtain the global features includes: fusing the output features through the global pooling operation formula to obtain the global features; the global pooling operation formula is:

[0020] F = Pooling(F layer |layer = 1, 2, 3)

[0021] where F represents the fused global feature, F layerIt represents the features extracted by the output layer of the $layer$, and Pooling(·) is the global pooling function.

[0022] In an embodiment of the present invention, using a classifier to classify the global features for Alzheimer's disease to obtain an Alzheimer's disease classification result includes: the classifier classifies the global features for Alzheimer's disease based on the global feature formula to obtain an Alzheimer's disease classification result; the global feature formula includes:

[0023]

[0024] where $C$ is the number of disease categories, and $C$ k is the $k$-th category. Common categories are: when $k = 1$, it is normal (NC); when $k = 2$, it is early mild cognitive impairment (EMCI); when $k = 3$, it is late mild cognitive impairment (LMCI); and when $k = 4$, it is Alzheimer's disease (AD); $S$ k represents the score for classifying the image into the $k$-th category, and $S$ i represents the score for classifying the image into the $i$-th category.

[0025] To achieve the above object, the present invention also provides a processing system for clinical data of Alzheimer's disease, including: an acquisition module, a skull stripping module, a sampling module, a cropping module, a normalization module, an extraction module, a fusion module, a classification module, and a visualization module; the acquisition module is used to obtain three-dimensional image data containing only brain parenchyma data from the ADNI clinical dataset based on a first preset rule; the skull stripping module performs skull stripping on the three-dimensional image data to obtain brain three-dimensional image data containing only brain parenchyma; the sampling module is used to resample the brain three-dimensional image data according to a first sampling rule to obtain resampled image data; the cropping module is used to crop the resampled image data to obtain final image data with blank slices and image edges removed; the normalization module is used to normalize the final image data to obtain normalized image data; the extraction module is used to extract features from the normalized image data using a neural network model based on U-net to obtain output features; the fusion module is used to fuse the output features using global pooling operations to obtain global features; the classification module is used to classify the global features for Alzheimer's disease using a classifier to obtain an Alzheimer's disease classification result; the visualization module is used to calculate the influence degree of each brain region in the three-dimensional image data on the Alzheimer's disease classification result by the method of backpropagation based on the Alzheimer's disease classification result, and display it in a heatmap visualization manner.

[0026] To achieve the above object, the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, any of the above-mentioned methods for processing clinical data of Alzheimer's disease is implemented.

[0027] To achieve the above object, the present invention further provides a device for processing clinical data of Alzheimer's disease, including: a processor and a memory; the memory is used for storing a computer program; the processor is connected to the memory and is used for executing the computer program stored in the memory, so that the device for processing clinical data of Alzheimer's disease executes any of the above-mentioned methods for processing clinical data of Alzheimer's disease.

[0028] As described above, a method, system, medium and device for processing clinical data of Alzheimer's disease according to the present invention have the following beneficial effects: being used for classifying the clinical data of Alzheimer's disease and showing the influence degree of each brain region in the three-dimensional brain image data on the classification result of Alzheimer's disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It shows a flowchart of the method for processing clinical data of Alzheimer's disease according to the present invention in an embodiment;

[0030] Figure 2 It shows a schematic structural diagram of the system for processing clinical data of Alzheimer's disease according to the present invention in an embodiment;

[0031] Figure 3 It shows a schematic structural diagram of the device for processing clinical data of Alzheimer's disease according to the present invention in an embodiment.

[0032] DESCRIPTION OF REFERENCE NUMERALS

[0033] 21 Acquisition module

[0034] 22 Skull stripping module

[0035] 23 Sampling module

[0036] 24 Cropping module

[0037] 25 Normalization module

[0038] 26 Extraction module

[0039] 27 Fusion module

[0040] 28 Classification module

[0041] 29 Visualization module

[0042] 31 Processor

[0043] 32 Memory Detailed implementation manners

[0044] The following uses specific specific examples to illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0045] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Therefore, only the components related to the present invention are shown in the diagrams, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and ratio of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0046] The method, system, medium, and device for processing clinical data of Alzheimer's disease of the present invention are used to classify the clinical data of Alzheimer's disease and realize the visualization of Alzheimer's disease categories.

[0047] As Figure 1 shown, in one embodiment, the method for processing clinical data of Alzheimer's disease of the present invention includes the following steps:

[0048] Step S11: Obtain three-dimensional image data containing brain data from the ADNI clinical data set based on a first preset rule.

[0049] Specifically, the ADNI (Alzheimer's Disease Neuroimaging Initiative) clinical data includes clinical information about each subject, including recruitment, demographics, physical examinations, and cognitive assessment data. The entire set of clinical data can be downloaded in batches as a comma-separated values (CSV) file. The ADNI clinical data is collected and managed by the Alzheimer's Therapeutic Research Institute (ATRI).

[0050] Specifically, the first preset rule is to obtain three-dimensional image data containing brain data from the publicly available MRI (Magnetic Resonance Imaging) T1 (weighted image: longitudinal relaxation) data of the ADNI data set. This removes other noise data and facilitates subsequent feature extraction and judgment.

[0051] Step S12: Perform skull stripping on the three-dimensional image data to obtain three-dimensional brain image data containing only brain parenchyma.

[0052] To simplify the learning complexity of the model, three-dimensional brain image data containing only brain parenchyma data is obtained by removing the skull.

[0053] Use a standard template to remove the skull-related parts in the 3D image data, so as to obtain the brain 3D image data containing only the brain-related parts. The standard template refers to the template that comes with 3Dslicer for stripping the scalp and skull on CT or MIR images, such as atlasbrainmask: brain mask.

[0054] Step S13: resample the brain three-dimensional image data according to a first sampling rule to obtain resampled image data.

[0055] Specifically, resampling the three-dimensional image data according to the first sampling rule to obtain resampled image data includes: resampling the processed image data to obtain image data with a smaller size, a larger number but a slightly lower precision, to facilitate subsequent processing. Specifically:

[0056] Resampling is performed by using the first sampling formula or the second sampling formula to obtain resampled image data;

[0057] The first sampling formula is:

[0058] P (x，y，z) =Q(2x, 2y, 2z)

[0059] Among them, P (x，y，z) represents the pixel value at the coordinate (x, y, z) of the resampled image data obtained by resampling by the first sampling formula, and Q(2x, 2y, 2z) represents the pixel value at the coordinate (2x, 2y, 2z) in the image of the three-dimensional image data;

[0060] In addition, another batch of resampled image data can be obtained using the second sampling formula. The second sampling formula is:

[0061] P (x，y，z) =Q(2x+1,2y+1,2z+1)

[0062] Among them, P (x，y，z) The pixel value at the coordinate (x, y, z) of the resampled image data obtained by resampling through the second sampling formula, Q(2x+1, 2y+1, 2z+1) represents the pixel value at the coordinate (2x+1, 2y+1, 2z+1) in the image of the three-dimensional image data. By adding better data preprocessing methods, the model has better generalization ability. The ADNI clinical data set was preprocessed through a series of methods such as pixel sampling, cropping and standardization, so that the subtle features in the original image can participate in the final model classification process.

[0063] Step S13: cropping the resampled image data to obtain final image data with blank slices and image edges removed.

[0064] Specifically, it includes: taking the image center of the resampled image data as the origin and using (64, 104, 80) unit pixels as the cropping range to crop the image of the final image data, so as to obtain the corresponding final image data based on the image of the final image data. Since, then considering the characteristics of the U-net neural network, (64 in length, 104 in width, 80 in height) unit pixels are used as the cropping range, which is the standard range for cropping the valid image data with the image center of the resampled image data as the origin. Thus, valid data (non-zero pixel values) are obtained, and blank slices are pixel values of zero, in order to obtain the final image data with less redundant information. The ADNI clinical dataset was preprocessed through a series of methods such as pixel sampling, cropping, and normalization, so that the subtle features in the original image can participate in the final model classification process.

[0065] Step S14: Normalize the final image data to obtain the normalized image data.

[0066] Specifically, the normalization of the final image data to obtain the normalized image data includes:

[0067] Normalize the final image data based on the normalization formula to obtain the normalized image data;

[0068] The normalization formula is;

[0069]

[0070] where P (x，y，z) represents the pixel value at the coordinate (x, y, z) in the image of the normalized image data, OP (x，y，z) represents the pixel value at the coordinate (x, y, z) in the image of the final image data, min(P) represents the minimum pixel value in the image of the final image data, and max(P) represents the maximum pixel value in the image of the final image data. In this way, the pixel value data of all images of the final image data are normalized to the range of [0 - 1], which is convenient for subsequent processing and calculation. The ADNI clinical dataset was preprocessed through a series of methods such as pixel sampling, cropping, and normalization, so that the subtle features in the original image can participate in the final model classification process.

[0071] Step S15: Use the U-net-based neural network to extract features from the normalized image data to obtain output features.

[0072] Specifically, the neural network model based on U-net (fully convolutional neural network) includes: the U-net neural network model contains eight convolutional blocks, and the number of channels of each convolutional block is 4, 16, 32, 64, 64, 32, 16, 8 respectively. And each convolutional block contains a convolutional layer to extract features; uses the rectified linear activation function to process the features; and uses the batch normalization function to normalize the features.

[0073] Specifically, the U-net neural network model is used to extract and automatically classify / diagnose the normalized image data. The U-net neural network model contains eight convolutional blocks, and the number of channels of each convolutional block is successively 4, 16, 32, 64, 64, 32, 16, 8 respectively. And each convolutional block contains a convolutional layer to extract features, then uses the rectified linear activation function (ReLU) to process the features, and then uses the batch normalization function (BN) to normalize the features. And pooling layers are set after the first, second, and third convolutional blocks to scale the features, and the pooling layer can use max pooling or average pooling. After the sixth and seventh convolutional blocks, upsampling or deconvolution methods are used to amplify the output features. At the same time, skip connections are used to connect the output features after the second convolutional layer and after the seventh convolutional layer, after the third convolutional layer and after the sixth convolutional layer, and after the fourth convolutional layer and after the fifth convolutional layer. Combining with the U-net neural network to learn MRI image features can be used in the diagnosis of Alzheimer's disease and can effectively learn the data features related to the disease diagnosis. The U-net neural network model is introduced to directly transfer the features learned in the shallow layer to the deep layer.

[0074] Step S16: Use global pooling operation to fuse the output features to obtain global features.

[0075] Specifically, using global pooling operation to fuse the output features to obtain global features includes: fusing the output features through the global pooling operation formula to obtain global features; the global pooling operation formula is: F = Pooling(F layer |layer = 1, 2, 3), where F represents the fused global features, F layer represents the features extracted by the layer-th output layer, and Pooling(·) is the global pooling function, which can be implemented using max pooling, average pooling, or other methods that can further fuse the features. There are three layers in the layer layer of the U-net neural network model, namely the sup1 layer, the sup2 layer, and the sup3 layer, which are the first layer, the second layer, and the third layer of the layer-th output layer respectively. Using the method of global max pooling greatly reduces the resource requirements during the model training process.

[0076] Step S17: Use a classifier to classify the global features for Alzheimer's disease to obtain an Alzheimer's disease classification result.

[0077] Specifically, the classifier classifies the global features for Alzheimer's disease based on the global feature formula to obtain an Alzheimer's disease classification result;

[0078] The global feature formula includes:

[0079]

[0080] where C is the number of disease categories, and C k represents the k-th category in C. The common categories are: when k = 1, it is normal (NC); when k = 2, it is early mild cognitive impairment (EMCI); when k = 3, it is late mild cognitive impairment (LMCI); and when k = 4, it is Alzheimer's disease (AD); S k represents the score for classifying the image into the k-th category, and S k indicates that classifying the image into the k-th category is determined by the global feature F. S i represents the score for classifying the image into the i-th category. Accurate classification of Alzheimer's disease is achieved, which can be applied to assist doctors in clinically diagnosing the condition of Alzheimer's disease.

[0081] Step S18: Based on the Alzheimer's disease classification result, use the backpropagation method to calculate the degree of influence of each brain region in the three-dimensional image data on the Alzheimer's disease classification result, and display it through a heat visualization method.

[0082] Specifically, use the backpropagation method to calculate the degree of influence of each part of the image on the model diagnosis result, and finally visualize its heat (i.e., the visualization of heatmaps of class activation in an image). Through the heatmap, understand the classification of each part of the image in the image classification problem, and at the same time, the position of the object in the image can be located). Thus, intuitively display the Alzheimer's disease classification result, and provide more clinical diagnosis and treatment support for junior clinicians.

[0083] Such as Figure 2As shown, in one embodiment, a processing system for clinical data of Alzheimer's disease according to the present invention includes: an acquisition module 21, a skull stripping module 22, a sampling module 23, a cropping module 24, a normalization module 25, an extraction module 26, a fusion module 27, a classification module 28, and a visualization module 29; the acquisition module 21 is configured to obtain three-dimensional image data containing only brain parenchyma data from the ADNI clinical dataset based on a first preset rule; the skull stripping module 22 performs skull stripping on the three-dimensional image data to obtain three-dimensional brain image data containing only brain parenchyma; the sampling module 23 is configured to resample the three-dimensional brain image data according to a first sampling rule to obtain resampled image data; the cropping module 24 is configured to crop the resampled image data to obtain final image data with blank slices and image edges removed; the normalization module 25 is configured to normalize the final image data to obtain normalized image data; the extraction module 26 is configured to use a neural network model based on U-net to extract features from the normalized image data to obtain output features; the fusion module 27 is configured to use global pooling operation to fuse the output features to obtain global features; the classification module 28 is configured to use a classifier to classify the global features for Alzheimer's disease to obtain an Alzheimer's disease classification result; the visualization module 29 is configured to calculate the influence degree of each brain region in the three-dimensional image data on the Alzheimer's disease classification result by backpropagation based on the Alzheimer's disease classification result, and display it in a heat visualization manner.

[0084] It should be noted that the structures and principles of the acquisition module 21, the skull stripping module 22, the sampling module 23, the cropping module 24, the normalization module 25, the extraction module 26, the fusion module 27, the classification module 28, and the visualization module 29; correspond one by one to the steps in the above-mentioned method for processing clinical data of Alzheimer's disease, so they will not be elaborated here.

[0085] It should be noted that it should be understood that the division of each module of the above system is only a division of logical functions. In actual implementation, it can be fully or partially integrated into a physical entity, or physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also all be implemented in the form of hardware; or some modules can be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the x module can be a separately established processing element, or can be integrated in a certain chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and the function of the above x module can be called and executed by a certain processing element of the above device. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together or can be independently implemented. The processing element mentioned here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed by the integrated logic circuit in the processor element or the instructions in the form of software.

[0086] For example, the above-mentioned modules can be one or more integrated circuits configured to implement the above method, such as: one or more application specific integrated circuits (ASICs), or, one or more microprocessor units (MPUs), or, one or more field programmable gate arrays (FPGAs), etc. Again, when a certain module above is implemented in the form of a processing element scheduling program code, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processors that can call program code. Again, these modules can be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0087] In an embodiment of the present invention, the present invention further includes a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the method for processing clinical data of Alzheimer's disease described above.

[0088] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to computer programs. The aforementioned computer program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above method embodiments; and the aforementioned storage medium includes: various media such as ROM, RAM, magnetic disks, or optical discs that can store program code.

[0089] As Figure 3 shown, in one embodiment, the processing device for clinical data of Alzheimer's disease according to the present invention includes: a processor 31 and a memory 32; the memory 32 is used for storing a computer program; the processor 31 is connected to the memory 32 and is used for executing the computer program stored in the memory 32, so that the processing device for clinical data of Alzheimer's disease executes any one of the processing methods for clinical data of Alzheimer's disease.

[0090] Specifically, the memory 32 includes: various media such as ROM, RAM, magnetic disks, USB flash drives, memory cards or optical discs that can store program codes.

[0091] Preferably, the processor 31 may be a general-purpose processor, including a central processing unit (Central Processing Unit, abbreviated as CPU), a network processor (Network Processor, abbreviated as NP), etc.; it may also be a digital signal processor (Digital Signal Processor, abbreviated as DSP), an application specific integrated circuit (Application Specific Integrated Circuit, abbreviated as ASIC), a field programmable gate array (Field Programmable Gate Array, abbreviated as FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0092] In summary, the processing method, system, medium and device for clinical data of Alzheimer's disease according to the present invention are used for classifying the clinical data of Alzheimer's disease and realizing the visualization of Alzheimer's disease categories. Therefore, the present invention effectively overcomes various disadvantages in the prior art and has high industrial utilization value.

[0093] The above embodiments are only illustrative of the principles and effects of the present invention and are not used to limit the present invention. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes completed by those with ordinary knowledge in the technical field without departing from the spirit and technical ideas disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. A method for processing clinical data of Alzheimer's disease, characterized in that, It includes the following steps: Obtain three-dimensional image data containing brain data from the ADNI clinical dataset based on the first preset rule; Perform skull stripping on the three-dimensional image data to obtain three-dimensional brain image data containing only brain parenchyma; Resample the three-dimensional brain image data according to the first sampling rule to obtain resampled image data; Among them, the resampling of the three-dimensional brain image data according to the first sampling rule to obtain the resampled image data includes: Perform resampling through the first sampling formula or the second sampling formula to obtain resampled image data; The first sampling formula is: P( x,y,z ) = Q(2x, 2y, 2z) The second sampling formula is: P( x,y,z ) = Q(2x + 1, 2y + 1, 2z + 1) Among them, P (x,y,z) represents the pixel value at the coordinate (x, y, z) obtained by resampling through the first sampling formula or the second sampling formula, and Q(2x, 2y, 2z), Q(2x + 1, 2y + 1, 2z + 1) represent the pixel values at the coordinates (2x, 2y, 2z), (2x + 1, 2y + 1, 2z + 1) in the image of the three-dimensional brain image data; Crop the resampled image data to obtain the final image data after removing blank slices and image edges; Normalize the final image data to obtain the normalized image data; Use a neural network model based on U-net to extract features from the normalized image data to obtain output features; Use global pooling operation to fuse the output features to obtain global features; Use a classifier to classify the global features for Alzheimer's disease to obtain the Alzheimer's disease classification result; Based on the Alzheimer's disease classification result, calculate the influence degree of each brain region in the three-dimensional image data on the Alzheimer's disease classification result through backpropagation, and display it through heat visualization.

2. The method for processing clinical data of Alzheimer's disease according to claim 1, wherein The performing skull stripping on the three-dimensional image data to obtain three-dimensional brain image data containing only brain parenchyma includes: Use a standard template to remove the skull-related parts in the three-dimensional image data, so as to obtain three-dimensional brain image data containing only brain parenchyma-related parts.

3. The method for processing clinical data of Alzheimer's disease according to claim 1, characterized in that The cropping the resampled image data to obtain the final image data after removing blank slices and image edges includes: Take the image center of the resampled image data as the origin and crop with a cropping range of (64, 104, 80) unit pixels to obtain the image of the final image data, so as to obtain the corresponding final image data based on the image of the resampled image data.

4. The method for processing clinical data of Alzheimer's disease according to claim 1, wherein, The normalizing the final image data to obtain the normalized image data includes: Normalize the final image data based on the normalization formula to obtain the normalized image data; The normalization formula is; Among them, P (x,y,z) represents the pixel value at the coordinate (x, y, z) in the image of the normalized image data, and OP (x,y,z) represents the pixel value at the coordinate (x, y, z) in the image of the final image data. min(P) represents the minimum pixel value in the image of the final image data, and max(P) represents the maximum pixel value in the image of the final image data.

5. The method for processing clinical data of Alzheimer's disease according to claim 1, wherein The using a neural network model based on U-net to extract features and classify the normalized image data to obtain output features includes: The U-net neural network model contains eight convolutional blocks, and the number of channels of each convolutional block is 4, 16, 32, 64, 64, 32, 16, 8 respectively, and each convolutional block contains a convolutional layer to extract features; Process the features using a rectified linear activation function; Normalize the features using a batch normalization function.

6. The method for processing clinical data of Alzheimer's disease according to claim 1, wherein The using global pooling operation to fuse the output features to obtain global features includes: Fuse the output features through the global pooling operation formula to obtain global features; The global pooling operation formula is: F = Pooling(F layer | layer = 1, 2, 3) Among them, F represents the fused global feature, and F layer represents the feature extracted by the output layer of the layer, and Pooling(·) is the global pooling function.

7. The method for processing clinical data of Alzheimer's disease according to claim 1, wherein The using a classifier to classify the global features for Alzheimer's disease to obtain the Alzheimer's disease classification result includes: The classifier classifies the global features based on the global feature formula to obtain the Alzheimer's disease classification result; The global feature formula includes: where C is the number of categories of the disease conditions, C k is the k-th category. Common categories are: normal when k = 1, early cognitive impairment when k = 2, late cognitive impairment when k = 3, and Alzheimer's disease when k = 4; S k represents the score for classifying the image into the k-th category, S i represents the score for classifying the image into the i-th category.

8. A processing system for clinical data of Alzheimer's disease, characterized in that, including: an acquisition module, a skull stripping module, a sampling module, a cropping module, a normalization module, an extraction module, a fusion module, a classification module, and a visualization module; The acquisition module is used to obtain three-dimensional image data containing only brain parenchyma data from the ADNI clinical dataset based on a first preset rule; The skull stripping module strips the skull from the three-dimensional image data to obtain brain three-dimensional image data containing only brain parenchyma; The sampling module is used to resample the brain three-dimensional image data according to a first sampling rule to obtain resampled image data; wherein, resampling the brain three-dimensional image data according to the first sampling rule to obtain the resampled image data includes: resampling through a first sampling formula or a second sampling formula to obtain resampled image data; The first sampling formula is: P( x,y,z ) = Q(2x, 2y, 2z) The second sampling formula is: P( x,y,z ) = Q(2x + 1, 2y + 1, 2z + 1) Among them, P (x,y,z) represents the pixel value at the coordinate (x, y, z) obtained by resampling through the first resampling formula or the second resampling formula, and Q(2x, 2y, 2z) and Q(2x + 1, 2y + 1, 2z + 1) represent the pixel values at the coordinates (2x, 2y, 2z) and (2x + 1, 2y + 1, 2z + 1) in the image of the three-dimensional brain image data; The cropping module is used to crop the resampled image data to obtain final image data with blank slices and image edges removed; The normalization module is used to normalize the final image data to obtain normalized image data; The extraction module is used to extract features from the normalized image data using a neural network model based on U-net to obtain output features; The fusion module is used to fuse the output features using global pooling operation to obtain global features; The classification module is used to classify the global features using a classifier to obtain the Alzheimer's disease classification result; The visualization module is used to calculate the influence degree of each brain region in the three-dimensional image data on the Alzheimer's disease classification result by the method of backpropagation based on the Alzheimer's disease classification result, and display it through heat visualization; 9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is executed by a processor to implement the method for processing clinical data of Alzheimer's disease according to any one of claims 1 to 7.

10. A processing device for clinical data of Alzheimer's disease, characterized in that, including: a processor and a memory; The memory is used to store a computer program; The processor is connected to the memory and is used to execute the computer program stored in the memory, so that the device for processing clinical data of Alzheimer's disease executes the method for processing clinical data of Alzheimer's disease according to any one of claims 1 to 7.

Images