Hippocampus area extraction method and device, electronic equipment and storage medium
By combining the ViTDet and UNeXt models with the point-spreading method, the problem of feature capture in hippocampal image segmentation was solved, achieving efficient and accurate hippocampal area extraction and assisting in the early diagnosis of Alzheimer's disease.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- DONGFANG HOSPITAL BEIJING UNIV OF CHINESE MEDICINE
- Filing Date
- 2023-01-17
- Publication Date
- 2026-07-31
AI Technical Summary
Existing hippocampal image segmentation methods struggle to effectively capture hidden and complex features, resulting in low computational efficiency and impacting the efficiency and effectiveness of hippocampal extraction.
A combined object detection and segmentation network using ViTDet and UNeXt models was employed, along with a point-scattering method to calculate the hippocampal area. By training on brain imaging data from multiple patients at different times, hippocampal region features were extracted and the area was calculated.
It improves the efficiency and accuracy of hippocampal extraction, enabling the capture of complex and hidden features, assisting in the early diagnosis of Alzheimer's disease, reducing the workload of doctors, and improving diagnostic efficiency.
Smart Images
Figure CN116205965B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of medical image processing technology, and in particular to a method, apparatus, electronic device, and storage medium for extracting hippocampal body area. Background Technology
[0002] The hippocampus is a brain region associated with memory. Studies have shown that hippocampal atrophy occurs in Alzheimer's patients, resulting in cognitive impairment. With the gradual maturation of artificial intelligence technology, especially the significant progress made in computer vision, using AI-assisted diagnostic techniques for hippocampal image-based diagnosis and screening is a feasible approach for screening and diagnosing Alzheimer's disease. Traditional hippocampal image segmentation methods identify and diagnose patients' medical images, but these images are differentiated and hierarchical, making diagnosis based on doctors' experience time-consuming and labor-intensive, reducing diagnostic efficiency. Related technologies use learning networks based on convolutional neural networks for hippocampal detection and segmentation; however, these methods directly process medical images with complex noise, increasing the difficulty of feature extraction, making it difficult to capture hidden or complex features effectively, and resulting in low computational efficiency, affecting the efficiency and effectiveness of hippocampal image extraction. Summary of the Invention
[0003] This invention provides a method, apparatus, electronic device, and storage medium for extracting hippocampal body area, which addresses the shortcomings of existing hippocampal body area extraction methods, such as difficulty in capturing hidden or complex features and low computational efficiency.
[0004] This invention provides a method for extracting hippocampal body area, comprising:
[0005] The image to be extracted is input into the target detection and segmentation network to obtain image data labeled with the hippocampal region. The target detection and segmentation network is trained based on multiple brain images of patients at different times and the hippocampal region label data corresponding to each brain image.
[0006] Extract the area of the hippocampus from the image data that identifies the hippocampal region.
[0007] This invention provides a method for extracting hippocampal area, wherein the target detection and segmentation network includes a ViTDet model and a UNeXt model, and the step of inputting the image to be extracted into the target detection and segmentation network to obtain image data marked with hippocampal regions includes:
[0008] The image to be extracted is input into the UNeXt model to obtain multi-dimensional features of the hippocampus region;
[0009] The multi-dimensional features of the hippocampal region are input into the ViTDet model to obtain image data with the hippocampal region identified.
[0010] This invention provides a method for extracting hippocampal area, wherein the UNeXt model includes an encoder and a decoder, and the step of inputting the image to be extracted into the UNeXt model to obtain multi-dimensional features of the hippocampal region includes:
[0011] The image to be extracted is input into the encoder, and the encoder performs a downsampling operation on the image to be extracted to extract the feature code corresponding to the hippocampal region.
[0012] The feature encoding corresponding to the hippocampal region is input into the decoder, and the decoder upsamples the feature encoding corresponding to the hippocampal region to obtain multi-dimensional features of the hippocampal region.
[0013] The present invention provides a method for extracting hippocampal body area, wherein the encoder includes at least one encoding convolution module and at least one encoding sensing module, the decoder includes at least one decoding sensing module and at least one decoding convolution module, and further includes:
[0014] The at least one encoding convolution module is used to perform convolution and pooling on the image to be extracted to obtain the extracted features in one step.
[0015] The at least one coding-aware module is used to perform residual convolution on the image to be extracted to obtain secondary extraction features, and to perform feature fusion on the primary extraction features and the secondary extraction features to obtain fused image features;
[0016] The at least one decoding perception module is used to upsample the secondary extracted features based on the fused image features to obtain secondary upsampled features;
[0017] The at least one decoding convolution module is used to upsample the first extracted features based on the fused image features to obtain a first upsampled feature, and to obtain multi-dimensional features of the hippocampal region based on the first upsampled feature and the second upsampled feature.
[0018] This invention provides a method for extracting the hippocampal body area, wherein the multi-dimensional features of the hippocampal body region are input into the ViTDet model to obtain image data identifying the hippocampal body region, including:
[0019] The multi-dimensional features of the hippocampus region are input into the pre-trained model of the ViTDet model to generate a predicted hippocampus region.
[0020] The pre-trained model of the ViTDet model is fine-tuned based on the difference between the predicted hippocampal region and the hippocampal region label data corresponding to each brain image.
[0021] Image data identifying the hippocampal region was obtained based on the finely tuned ViTDet model.
[0022] The present invention provides a method for extracting the area of the hippocampus, wherein extracting the area of the hippocampus from image data marked with hippocampal regions includes:
[0023] Multiple random points are uniformly generated in the image data that identifies the hippocampal region;
[0024] Obtain the number of random points in the hippocampal region and calculate the proportion of random points falling into the hippocampal region;
[0025] The area of the hippocampus is extracted based on the proportion of random points falling into the hippocampal region and the area of the image marked with the hippocampal region.
[0026] The present invention provides a method for extracting hippocampal body area, further comprising:
[0027] Record and compare the changes in hippocampal body area of the same patient at different times, and use the changes in hippocampal body area as an auxiliary basis for diagnosis.
[0028] The present invention also provides a hippocampal body area extraction device, comprising:
[0029] The acquisition module is used to input the image to be extracted into the target detection and segmentation network to acquire image data labeled with the hippocampal region. The target detection and segmentation network is trained based on multiple brain images of patients at different times and the hippocampal region label data corresponding to each brain image.
[0030] An extraction module is used to extract the area of the hippocampus from the image data marked with the hippocampal region.
[0031] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the hippocampal area extraction method described in any of the preceding claims.
[0032] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the hippocampal area extraction method described in any of the preceding claims.
[0033] This invention provides a method, apparatus, electronic device, and storage medium for extracting hippocampal area. The method involves inputting the image to be extracted into a target detection and segmentation network to obtain image data labeled with hippocampal regions. The target detection and segmentation network is trained based on multiple brain images of patients at different times and the corresponding hippocampal region label data for each brain image. Extracting the hippocampal area from the image data labeled with hippocampal regions allows for focusing on the extraction area in the brain image, capturing hidden and complex features, reducing model computation data, and improving extraction efficiency and effectiveness. Attached Figure Description
[0034] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0035] Figure 1 This is one of the flowcharts of the hippocampal body area extraction method provided by the present invention;
[0036] Figure 2 This is the second flowchart of the hippocampal body area extraction method provided by the present invention;
[0037] Figure 3 This is the third flowchart of the hippocampal body area extraction method provided by the present invention;
[0038] Figure 4 This is a schematic diagram of the hippocampal body area extraction device provided by the present invention;
[0039] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation
[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0041] Figure 1 The flowchart of the hippocampal body area extraction method provided in the embodiments of the present invention is as follows: Figure 1 As shown, the hippocampal body area extraction method provided in this embodiment of the invention includes:
[0042] Step 101: Input the image to be extracted into the target detection and segmentation network to obtain image data labeled with the hippocampal region. The target detection and segmentation network is trained based on multiple brain images of patients at different times and the corresponding hippocampal region label data for each brain image.
[0043] In this embodiment of the invention, brain images of 500 Alzheimer's patients at different times are used as training data for the target detection and segmentation network, and the hippocampus region of each image is labeled for subsequent training of the target detection and segmentation network model for entity recognition and segmentation.
[0044] Step 102: Extract the area of the hippocampus from the image data marked with the hippocampal region.
[0045] In this embodiment of the invention, hippocampal image target detection is performed using the Transformer architecture. The overall Transformer architecture can be divided into four parts: input part, output part, encoder part, and decoder part. The division of labor and cooperation among these four parts can improve the training efficiency and feature extraction effect of the target detection and segmentation network.
[0046] In some embodiments of the present invention, the hippocampal body area extraction method further includes:
[0047] Record and compare the changes in hippocampal body area of the same patient at different times, and use the changes in hippocampal body area as an auxiliary basis for diagnosis.
[0048] The method of using hippocampal surface area calculation for Alzheimer's disease auxiliary diagnosis can perform relatively more accurate and efficient detection of more complex and hidden hippocampal images, thereby significantly improving the diagnostic efficiency of Alzheimer's disease images.
[0049] Traditional Alzheimer's disease diagnosis often relies on clinical manifestations, neurological scales, biomarker tests, and medical imaging. However, this diagnostic process typically fails to ensure timely detection and treatment, is prone to overlooking imaging findings, and lacks the ability to compare changes in imaging before and after the disease. Related technologies utilize learning networks based on convolutional neural networks for hippocampal detection and segmentation. However, these methods directly process medical images with complex noise, increasing the difficulty of feature extraction, making it difficult to capture hidden or complex features effectively, and also suffer from low computational efficiency, impacting the efficiency and effectiveness of hippocampal extraction.
[0050] This invention provides a method for extracting the area of the hippocampus. The method involves inputting the image to be extracted into a target detection and segmentation network to obtain image data labeled with hippocampal regions. The target detection and segmentation network is trained based on multiple brain images of patients at different times and the corresponding hippocampal region label data for each brain image. Extracting the area of the hippocampus from the image data labeled with hippocampal regions allows for focusing on the extraction area in the brain image, capturing hidden and complex features, reducing model computation data, and improving extraction efficiency and effectiveness.
[0051] Based on any of the above embodiments, the target detection and segmentation network includes the ViTDet model and the UNeXt model, such as... Figure 2 As shown, the image to be extracted is input into the target detection and segmentation network to obtain image data marked with hippocampal regions, including:
[0052] Step 201: Input the image to be extracted into the UNeXt model to obtain multi-dimensional features of the hippocampus region;
[0053] In this embodiment of the invention, the UNeXt model is an architecture based on convolution and multilayer perceptron (MLP). The UNeXt model includes an encoder and a decoder. The image to be extracted is input into the UNeXt model to obtain multi-dimensional features of the hippocampal region, including:
[0054] Step 2011: Input the image to be extracted into the encoder, and use the encoder to perform downsampling operation on the image to be extracted to extract the feature code corresponding to the hippocampal region;
[0055] Step 2012: Input the feature encoding corresponding to the hippocampal region into the decoder, and upsample the feature encoding corresponding to the hippocampal region through the decoder to obtain multi-dimensional features of the hippocampal region.
[0056] In an embodiment of the present invention, the encoder includes at least one encoding convolutional module and at least one encoding sensing module, and the decoder includes at least one decoding sensing module and at least one decoding convolutional module. In a specific embodiment of the present invention, the encoder includes three convolutional blocks and two tokenized MLP blocks; the decoder includes two tokenized MLP blocks and three convolutional blocks, and the modules in the encoder and decoder are symmetrically arranged.
[0057] In some embodiments of the invention, inputting the image to be extracted into the UNeXt model to obtain multi-dimensional features of the hippocampus region further includes:
[0058] At least one coded convolutional module is used to perform convolution and pooling on the image to be extracted, so as to obtain the extracted features in one step.
[0059] At least one coding-aware module is used to perform residual convolution on the image to be extracted to obtain secondary extraction features, and to fuse the primary extraction features and the secondary extraction features to obtain fused image features;
[0060] At least one decoding and sensing module is used to upsample the secondary extracted features based on the fused image features to obtain secondary upsampled features;
[0061] At least one decoding convolutional module is used to upsample the extracted features based on the fused image features to obtain the first upsampled features, and to obtain the multi-dimensional features of the hippocampal region based on the first upsampled features and the second upsampled features.
[0062] By using the UNeXt model to extract features and segment images from the image to be extracted, the network detection efficiency can be improved.
[0063] Step 202: Input the multi-dimensional features of the hippocampus region into the ViTDet model to obtain image data with the hippocampus region marked.
[0064] In this embodiment of the invention, the ViTDet model is a detection model proposed by the Meta AI team based on the native ViT model as the backbone network. In some embodiments of the invention, multi-dimensional features of the hippocampus region are input into the ViTDet model to obtain image data identifying the hippocampus region, including:
[0065] Step 2021: Input the multi-dimensional features of the hippocampus region into the pre-trained model of the ViTDet model to generate the predicted hippocampus region;
[0066] Step 2022: Fine-tune the pre-trained model of the ViTDet model based on the difference between the predicted hippocampal region and the hippocampal region label data corresponding to each brain image;
[0067] Step 2023: Obtain image data with hippocampal regions identified based on the fine-tuned ViTDet model.
[0068] In this embodiment of the invention, the pre-trained model is a model that has already been trained on a dataset, which can be obtained from an existing model library. The model is trained on hippocampal images, and its parameters are fine-tuned. Fine-tuning methods include, for example, truncating the last layer (softmax layer) of the pre-trained network and replacing it with a new softmax layer relevant to hippocampal region recognition. For instance, a pre-trained network on ImageNet has a softmax layer with 1000 categories. If our task is to classify 10 categories, the new softmax layer of the network will consist of 10 categories instead of 1000. The pre-trained weights are then run on the network. Alternatively, a smaller learning rate is used to train the network. Since the pre-trained weights are expected to be quite good compared to randomly initialized weights, the initial learning rate is made 10 times smaller than the initial learning rate used for training from scratch. If the dataset size is too small, only the last layer is trained; if the dataset size is moderate, the weights of the first few layers of the pre-trained network are frozen, as these weights are kept unchanged since the first few layers capture common features relevant to hippocampal region recognition, such as curves and edges. The finely tuned ViTDet model has the ability to detect hippocampal images.
[0069] In this embodiment of the invention, by fusing and training the ViTDet model and the UNeXt model, the final object detection and segmentation network is able to obtain a segmented hippocampal image from a single brain image.
[0070] Based on any of the above embodiments, such as Figure 3 As shown, the area of the hippocampus is extracted from image data identifying the hippocampal region, including:
[0071] Step 301: Generate multiple random points uniformly in the image data that are marked with hippocampal regions;
[0072] Step 302: Obtain the number of random points in the hippocampal region and calculate the proportion of random points falling into the hippocampal region;
[0073] Step 303: Extract the area of the hippocampus based on the proportion of random points falling into the hippocampal region and the area of the image marked with the hippocampal region.
[0074] By uniformly generating multiple random points in image data identifying the hippocampal region, the area of the irregular hippocampus in a two-dimensional plane is calculated using a point-scattering method. For example, the area of the original brain image is S0, which includes an irregular region—the hippocampus. M points are uniformly and randomly scattered across the brain image, with N points falling within the irregular hippocampal region. Based on this, we can estimate the area S of the hippocampus.
[0075]
[0076] As M approaches infinity
[0077]
[0078] The method of calculating the area of the hippocampus by scattering dots solves the problem that doctors have difficulty detecting subtle differences in brain images, thus improving the accuracy of recognition.
[0079] In some embodiments of the present invention, tests were conducted on a test set of 2,000 labeled images. After inference and manual verification, the accuracy rate reached 92.3%, which meets the requirements of business scenarios.
[0080] This invention provides a hippocampal area extraction method that combines hippocampal image target detection using a Transformer architecture with hippocampal area analysis to provide auxiliary evidence for the diagnosis of Alzheimer's disease. This method enables more accurate and rapid detection and screening of patient images, reduces personnel costs, and addresses the issues of insufficient specialist doctors and low early Alzheimer's screening rates. It assists clinicians in making decisions and providing timely treatment for patients. Furthermore, the method integrates ViTDet and UNeXt to improve recognition accuracy, and combines the dot-mapping method to calculate area, solving the problem of doctors not easily detecting subtle differences in brain images and improving recognition accuracy.
[0081] The hippocampal body area extraction device provided by the present invention is described below. The hippocampal body area extraction device described below can be referred to in correspondence with the hippocampal body area extraction method described above.
[0082] Figure 4 This is a schematic diagram of the hippocampal body area extraction device provided in an embodiment of the present invention, as shown below. Figure 4 As shown, the hippocampal body area extraction device provided in this embodiment of the invention includes:
[0083] The acquisition module 401 is used to input the image to be extracted into the target detection and segmentation network to acquire image data labeled with the hippocampal region. The target detection and segmentation network is trained based on multiple brain images of patients at different times and the hippocampal region label data corresponding to each brain image.
[0084] Extraction module 402 is used to extract the area of the hippocampus from image data that identifies the hippocampal region.
[0085] The hippocampal area extraction device provided in this invention obtains image data labeled with hippocampal regions by inputting the image to be extracted into a target detection and segmentation network. The target detection and segmentation network is trained based on multiple brain images of patients at different times and the corresponding hippocampal region label data for each brain image. Extracting the hippocampal area from the image data labeled with hippocampal regions can focus on the extraction area, capture hidden and complex features, reduce model computation data, and improve extraction efficiency and extraction effect.
[0086] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a hippocampal area extraction method, which includes: inputting the image to be extracted into a target detection and segmentation network to obtain image data labeled with hippocampal regions; the target detection and segmentation network is trained based on multiple brain images of patients at different times and the corresponding hippocampal region label data for each brain image; and extracting the area of the hippocampus from the image data labeled with hippocampal regions.
[0087] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0088] On the other hand, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the hippocampal area extraction method provided by the above methods. The method includes: inputting the image to be extracted into a target detection and segmentation network to obtain image data labeled with hippocampal regions, wherein the target detection and segmentation network is trained based on multiple brain images of patients at different times and hippocampal region label data corresponding to each brain image; and extracting the area of the hippocampus from the image data labeled with hippocampal regions.
[0089] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0090] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.
[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A hippocampus area extraction method characterized by, include: The image to be extracted is input into the target detection and segmentation network to obtain image data labeled with the hippocampal region. The target detection and segmentation network is trained based on multiple brain images of patients at different times and the hippocampal region label data corresponding to each brain image. Extract the area of the hippocampus from the image data marked with the hippocampal region; The target detection and segmentation network includes the ViTDet model and the UNeXt model. The step of inputting the image to be extracted into the target detection and segmentation network to obtain image data marked with hippocampal regions includes: The image to be extracted is input into the UNeXt model to obtain multi-dimensional features of the hippocampus region; The multi-dimensional features of the hippocampus region are input into the ViTDet model to obtain image data with the hippocampus region identified. The step of inputting the multi-dimensional features of the hippocampal region into the ViTDet model to obtain image data identifying the hippocampal region includes: The multi-dimensional features of the hippocampus region are input into the pre-trained model of the ViTDet model to generate a predicted hippocampus region. The pre-trained model of the ViTDet model is fine-tuned based on the difference between the predicted hippocampal region and the hippocampal region label data corresponding to each brain image. Image data with hippocampal regions were obtained based on the finely tuned ViTDet model; Extracting the area of the hippocampus from the image data marked with the hippocampal region includes: Multiple random points are uniformly generated in the image data that identifies the hippocampal region; Obtain the number of random points in the hippocampal region and calculate the proportion of random points falling into the hippocampal region; The area of the hippocampus is extracted based on the proportion of random points falling into the hippocampal region and the area of the image with the hippocampal region marked. The area of the two-dimensional plane of the irregular hippocampus is calculated by uniformly generating multiple random points in the image data with the hippocampal region marked, using a point-scattering method.
2. The hippocampus area extraction method according to claim 1, characterized in that, The UNeXt model includes an encoder and a decoder. The process of inputting the image to be extracted into the UNeXt model to obtain multi-dimensional features of the hippocampal region includes: The image to be extracted is input into the encoder, and the encoder performs a downsampling operation on the image to be extracted to extract the feature code corresponding to the hippocampal region. The feature encoding corresponding to the hippocampal region is input into the decoder, and the decoder upsamples the feature encoding corresponding to the hippocampal region to obtain multi-dimensional features of the hippocampal region.
3. The method for extracting hippocampal body area according to claim 2, characterized in that, The encoder includes at least one encoding convolutional module and at least one encoding perceptron module, and the decoder includes at least one decoding perceptron module and at least one decoding convolutional module, and further includes: The at least one encoding convolution module is used to perform convolution and pooling on the image to be extracted to obtain the extracted features in one step. The at least one coding-aware module is used to perform residual convolution on the image to be extracted to obtain secondary extraction features, and to perform feature fusion on the primary extraction features and the secondary extraction features to obtain fused image features; The at least one decoding perception module is used to upsample the secondary extracted features based on the fused image features to obtain secondary upsampled features; The at least one decoding convolution module is used to upsample the first extracted features based on the fused image features to obtain a first upsampled feature, and to obtain multi-dimensional features of the hippocampal region based on the first upsampled feature and the second upsampled feature.
4. The method for extracting hippocampal body area according to claim 1, characterized in that, Also includes: Record and compare the changes in hippocampal body area of the same patient at different times, and use the changes in hippocampal body area as an auxiliary basis for diagnosis.
5. A device for extracting the body area of a hippocampus, characterized in that, include: The acquisition module is used to input the image to be extracted into the target detection and segmentation network to acquire image data labeled with the hippocampal region. The target detection and segmentation network is trained based on multiple brain images of patients at different times and the hippocampal region label data corresponding to each brain image. The target detection and segmentation network includes the ViTDet model and the UNeXt model. The step of inputting the image to be extracted into the target detection and segmentation network to obtain image data marked with hippocampal regions includes: The image to be extracted is input into the UNeXt model to obtain multi-dimensional features of the hippocampus region; The multi-dimensional features of the hippocampus region are input into the ViTDet model to obtain image data with the hippocampus region identified. The step of inputting the multi-dimensional features of the hippocampal region into the ViTDet model to obtain image data identifying the hippocampal region includes: The multi-dimensional features of the hippocampus region are input into the pre-trained model of the ViTDet model to generate a predicted hippocampus region. The pre-trained model of the ViTDet model is fine-tuned based on the difference between the predicted hippocampal region and the hippocampal region label data corresponding to each brain image. Image data with hippocampal regions were obtained based on the finely tuned ViTDet model; The extraction module is used to extract the area of the hippocampus from the image data marked with the hippocampal region. Extracting the area of the hippocampus from the image data marked with the hippocampal region includes: Multiple random points are uniformly generated in the image data that identifies the hippocampal region; Obtain the number of random points in the hippocampal region and calculate the proportion of random points falling into the hippocampal region; The area of the hippocampus is extracted based on the proportion of random points falling into the hippocampal region and the area of the image with the hippocampal region marked. The area of the two-dimensional plane of the irregular hippocampus is calculated by uniformly generating multiple random points in the image data with the hippocampal region marked, using a point-scattering method.
6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the hippocampal area extraction method as described in any one of claims 1 to 4.
7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the hippocampal area extraction method as described in any one of claims 1 to 4.