Neural image data processing method and device

Abstract

The invention provides a neural image data processing method and device, and the method comprises the steps: obtaining a structure partition under a cerebral cortex corresponding to brain structure image data based on the brain structure image data and a brain structure segmentation model; obtaining a level set corresponding to the brain structure image data based on the brain structure image data, a white matter partition part included in a structure partition under the cerebral cortex corresponding to the brain structure image data and a cerebral cortex reconstruction model, reconstructing a cerebral cortex image based on the level set corresponding to the brain structure image data, and obtaining reconstructed cerebral cortex image data; and based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model, performing spherical registration of the cerebral cortex surface and the cerebral cortex surface of the first standard space to obtain a registered spherical surface of the cerebral cortex surface corresponding to the brain structure image data. According to the method and the device for processing the neural image data, the accuracy of processing the neural image data is improved.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging technology, and in particular to a method and device for processing neuroimaging data. Background Art

[0002] Neuroimaging is an important technology involving multiple disciplines. Brain structure and function data can be analyzed through neuroimaging. Through the processing of neuroimaging, we can clearly see the different regions, tissues and structures of the brain as well as the changes in blood flow in the brain. Based on the changes in blood flow, we can analyze the activity state of the brain under different task conditions.

[0003] At present, the neuroimaging data of some patients are quite different from those of normal people. For example, patients with brain damage, tumors, cerebral hemorrhage, patients with unclear consciousness, etc. Their brain tissue structure is different from that of normal people, which leads to the inability of the neuroimaging processing methods in the existing technology to perform correct structural segmentation, cortical reconstruction, and surface registration of the cerebral cortex; or the inability to perform correct spatial normalization. Therefore, how to propose a neuroimaging data processing method that can be applicable to the processing of patients' neuroimaging data and improve the success rate of neuroimaging data processing has become an important issue that needs to be solved in this field. Summary of the invention

[0004] In view of the problems in the prior art, the embodiments of the present invention provide a method and device for processing neuroimaging data, which can at least partially solve the problems in the prior art.

[0005] In a first aspect, the present invention provides a method for processing neuroimaging data, comprising:

[0006] Based on the brain structure image data and the brain structure segmentation model, obtaining the structural partition under the cerebral cortex corresponding to the brain structure image data; wherein the brain structure segmentation model is obtained by pre-training;

[0007] Based on the brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and the cerebral cortex reconstruction model, a level set corresponding to the brain structure image data is obtained, and a cerebral cortex image is reconstructed based on the level set corresponding to the brain structure image data to obtain reconstructed cerebral cortex image data corresponding to the brain structure image data; wherein the cerebral cortex reconstruction model is obtained by pre-training;

[0008] Based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model, the spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space is performed to obtain the spherical surface of the cerebral cortex surface corresponding to the brain structure image data after registration; wherein the cerebral cortex registration model is obtained by pre-training.

[0009] In a second aspect, the present invention provides a device for processing neuroimaging data, comprising:

[0010] A segmentation unit, used for obtaining the structural partition under the cerebral cortex corresponding to the brain structure image data based on the brain structure image data and the brain structure segmentation model; wherein the brain structure segmentation model is obtained by pre-training;

[0011] A reconstruction unit, configured to obtain a level set corresponding to the brain structure image data based on the brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and a cerebral cortex reconstruction model, and reconstruct the cerebral cortex image based on the level set corresponding to the brain structure image data to obtain reconstructed cerebral cortex image data corresponding to the brain structure image data; wherein the cerebral cortex reconstruction model is obtained by pre-training;

[0012] A registration unit is used to perform spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model, so as to obtain the spherical surface of the cerebral cortex surface corresponding to the brain structure image data after registration; wherein the cerebral cortex registration model is obtained by pre-training.

[0013] In a third aspect, the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the program to implement the method for processing neuroimaging data described in any one of the above embodiments.

[0014] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program / instruction, and when the computer program / instruction is executed by a processor, the method for processing neuroimaging data described in any of the above embodiments is implemented.

[0015] In a fifth aspect, the present invention provides a computer program product, comprising a computer program / instruction, which, when executed by a processor, implements the method for processing neuroimaging data described in any of the above embodiments.

[0016] The method and device for processing neural imaging data provided by the embodiment of the present invention obtain, based on brain structural imaging data and a brain structural segmentation model, the structural partition under the cerebral cortex corresponding to the brain structural imaging data; obtain the level set corresponding to the brain structural imaging data based on the brain structural imaging data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structural imaging data, and the cerebral cortex reconstruction model, and reconstruct the cerebral cortex image based on the level set corresponding to the brain structural imaging data to obtain the reconstructed cerebral cortex image data corresponding to the brain structural imaging data; perform spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structural imaging data and the cerebral cortex registration model to obtain the spherical surface of the cerebral cortex surface corresponding to the brain structural imaging data after registration, which has higher robustness and adaptability, effectively overcomes the shortcomings of traditional technology in processing neural imaging data of specific patients, and improves the accuracy of neural imaging data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0018] Figure 1 4 is a flowchart of a method for processing neuroimaging data provided by the first embodiment of the present invention.

[0019] Figure 2 4 is a flowchart of a method for processing neuroimaging data provided by the second embodiment of the present invention.

[0020] Figure 3 4 is a flowchart of a method for processing neuroimaging data provided in the third embodiment of the present invention.

[0021] Figure 4 4 is a flowchart of a method for processing neuroimaging data provided by a fourth embodiment of the present invention.

[0022] Figure 5 4 is a flowchart of a method for processing neuroimaging data provided in a fifth embodiment of the present invention.

[0023] Figure 6 4 is a flowchart of a method for processing neuroimaging data provided in a sixth embodiment of the present invention.

[0024] Figure 7 4 is a flowchart of a method for processing neuroimaging data provided by the seventh embodiment of the present invention.

[0025] Figure 8 4 is a schematic diagram of the structure of a neural image data processing device provided in the eighth embodiment of the present invention.

[0026] Fig. 9 4 is a schematic diagram of the structure of a neuroimaging data processing device provided in the ninth embodiment of the present invention.

[0027] Fig.10 4 is a schematic diagram of the structure of a neural imaging data processing device provided in the tenth embodiment of the present invention.

[0028] Fig.11 4 is a schematic diagram of the structure of a neural imaging data processing device provided by the eleventh embodiment of the present invention.

[0029] Fig.12 4 is a schematic diagram of the structure of a neural imaging data processing device provided in the twelfth embodiment of the present invention.

[0030] Fig.13 4 is a schematic diagram of the structure of a neural imaging data processing device provided in the thirteenth embodiment of the present invention.

[0031] Fig.14 4 is a schematic diagram of the structure of a neural image data processing device provided in the fourteenth embodiment of the present invention.

[0032] Fig.15 It is a schematic diagram of the physical structure of an electronic device provided in the fifteenth embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention. It should be noted that, in the absence of conflict, the embodiments in this application and the features in the embodiments can be combined with each other arbitrarily.

[0034] The acquisition, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of laws and regulations. The user information in the embodiments of this application is obtained through legal and compliant channels, and the acquisition, storage, use, and processing of user information are authorized and agreed by the customer.

[0035] In order to facilitate understanding of the technical solution provided by this application, the relevant contents of the technical solution of this application are first explained below.

[0036] Neuroimaging data is data that reflects brain structure and function information obtained through various neuroimaging technologies, mainly including structural magnetic resonance imaging (sMRI) data, functional magnetic resonance imaging (fMRI) data, etc.

[0037] In the field of medical imaging, especially in the analysis of brain structural imaging data, traditional processing techniques often show significant limitations when facing specific patient groups. In particular, for patients with brain damage, tumors, cerebral hemorrhage, and unconsciousness, the brain structural imaging data generated by them are significantly different from those of normal people. These differences are not only reflected in the overall structure of brain tissue, but also in the contrast, noise level, and morphology of brain neural images. These factors have caused difficulties in the existing technology in key processing such as structural segmentation, cortical reconstruction, and spatial standard normalization, resulting in a significant reduction in the accuracy and reliability of the processing results.

[0038] In order to meet the above challenges, the present invention introduces a variety of processing algorithms based on deep learning, including brain structure segmentation algorithm, cerebral cortex reconstruction algorithm, cerebral cortex registration algorithm and spatial normalization algorithm. Through the design and performance optimization of these algorithms, the success rate and accuracy of preprocessing of patient brain structure imaging data are greatly improved, thus providing a more accurate and reliable basis for clinical decision-making.

[0039] The following takes the server as an example to illustrate the specific implementation process of the method for processing neural image data provided by the embodiment of the present invention. It is understandable that the execution subject of the method for processing neural image data provided by the embodiment of the present invention is not limited to the server.

[0040] Figure 1 is a flowchart of a method for processing neuroimaging data provided by the first embodiment of the present invention. Figure 1 As shown, the method for processing neuroimaging data provided by the embodiment of the present invention includes:

[0041] S101, based on brain structure image data and a brain structure segmentation model, obtaining a subcortical structural partition corresponding to the brain structure image data; wherein the brain structure segmentation model is obtained by pre-training;

[0042] Specifically, brain structure image data can be obtained by magnetic resonance imaging (MRI) imaging equipment or CT imaging equipment. For example, brain structure image data includes T1 weighted data, T2 weighted data, diffusion tensor imaging, CT and other modal data. The server obtains the structural partitions under the cerebral cortex and the structural partitions outside the cerebral cortex corresponding to the brain structure image data based on the brain structure image data and the brain structure segmentation model. In order to compare and analyze the data on the same scale, the brain structure image data is normalized so that the intensity of the brain structure image data is converted to between 0 and 1, and the normalized brain structure image data is input into the brain structure segmentation model, and the structural partitions under the cerebral cortex and the structural partitions outside the cerebral cortex corresponding to the brain structure image data are output. The structural partitions under the cerebral cortex corresponding to the brain structure image data include white matter partitions. The brain structure segmentation model can distinguish the cerebral cortex and subcortical structures in the brain structure image data. Among them, the brain structure image data can use T1 weighted (T1-weighted) data.

[0043] Traditional structural segmentation algorithms often rely on some assumptions, such as the shape, size and local features of the structure. These assumptions are usually valid when dealing with normal brain structures, but when faced with patients with abnormal brain structures (such as patients with brain damage or tumors), these assumptions will fail, resulting in inaccurate segmentation. For example, the damaged area may be mistakenly labeled as normal brain tissue, while the actual normal area may be ignored, thus affecting subsequent analysis. Unlike traditional generative algorithms, the brain structure segmentation model of the present application adopts a non-generative model, so that abnormal brain tissue (such as lesions) can be accurately labeled as non-brain tissue, so that normal brain tissue is retained. The brain structure segmentation model of the present application not only improves the recognition rate of abnormal tissue, but also reduces the probability of mislabeling, ensuring the accuracy of subsequent processing.

[0044] S102, based on the brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and the cerebral cortex reconstruction model, obtain the level set corresponding to the brain structure image data, and reconstruct the cerebral cortex image based on the level set corresponding to the brain structure image data to obtain the reconstructed cerebral cortex image data corresponding to the brain structure image data; wherein the cerebral cortex reconstruction model is obtained by pre-training;

[0045] Specifically, the server can obtain the level set corresponding to the brain structure image data based on the brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and the cerebral cortex reconstruction model. In order to compare and analyze the data on the same scale, the brain structure image data is normalized so that the intensity of the brain structure image data is converted to between 0 and 1, and the normalized brain structure image data and the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data are input into the cerebral cortex reconstruction model, and the level set corresponding to the brain structure image data can be output. Wherein, the cerebral cortex reconstruction model is obtained by pre-training.

[0046] The server reconstructs the cerebral cortex image according to the level set corresponding to the brain structure image data, and obtains the reconstructed cerebral cortex image data corresponding to the brain structure image data. The specific process of reconstructing the cerebral cortex image based on the level set corresponding to the brain structure image data is a prior art and will not be repeated here. The traditional cerebral cortex segmentation algorithm is very slow, and the input of the cerebral cortex reconstruction model described in the present application is obtained based on the output of the aforementioned brain structure segmentation model, which can quickly reconstruct the cerebral cortex, so the reconstruction of the cerebral cortex of the present application is faster than the reconstruction of the cerebral cortex using the traditional cerebral cortex segmentation algorithm in terms of speed, and the present application improves the reconstruction efficiency of the cerebral cortex.

[0047] Furthermore, in order to obtain a more accurate level set, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data can be filled, and the filled white matter partition part and the normalized brain structure image data can be used as input data for the cerebral cortex reconstruction model.

[0048] Reconstruction of the cerebral cortex is an important step in understanding brain function. Traditional reconstruction algorithms are usually complex and highly dependent on the morphology of the input data. When there are structural abnormalities in the input data, these algorithms may require long iterative calculations, and the results are often unreliable. For some extreme cases, the reconstruction process may not even be completed, resulting in valuable clinical data that cannot be effectively utilized. The cerebral cortex reconstruction model of the present application can adapt to different brain structure morphologies and sizes. Compared with traditional algorithms, it can handle more complex topological structures and quickly reconstruct a reasonable cerebral cortex, greatly improving the processing speed and effect.

[0049] S103. Perform spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model to obtain the spherical surface of the cerebral cortex surface corresponding to the brain structure image data after registration; wherein the cerebral cortex registration model is obtained through pre-training.

[0050] Specifically, the server aligns the cerebral cortex surface with the cerebral cortex surface of the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model, and can obtain the spherical surface of the cerebral cortex surface after registration corresponding to the brain structure image data. The reconstructed cerebral cortex image data corresponding to the brain structure image data is inflated and spherical parameterized, and the original unregistered spherical surface and gyrus depth information corresponding to the brain structure image data can be obtained; the original unregistered spherical surface and gyrus depth information corresponding to the brain structure image data are sampled to the first standard space, and the spherical surface of the cerebral cortex surface sampled in the first standard space corresponding to the brain structure image data can be obtained; the spherical surface of the cerebral cortex surface in the first standard space, the spherical surface of the cerebral cortex surface sampled in the first standard space corresponding to the brain structure image data, and the template gyrus of the first standard space are sampled. The depth information is input into the cerebral cortex registration model, and the deformation matrix of the individual space registration to the first standard space can be obtained; according to the deformation matrix of the individual space registration to the first standard space, the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data and the spherical surface of the cerebral cortex surface in the first standard space, the deformation matrix of the original unregistered spherical space of the cerebral cortex surface can be obtained, and then the deformation matrix of the original unregistered spherical space of the cerebral cortex surface is applied to the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data, and the spherical surface of the cerebral cortex surface after registration corresponding to the brain structure image data can be obtained. Among them, the first standard space is a universal, standardized reference space, and the fsaverage standard space can be used. In the analysis of cerebral cortex image data, mapping the brain images of different individuals into the first standard space can realize comparison and statistical analysis between different individuals. The individual space is relative to the first standard space, and refers to the spatial structure unique to the brain of a specific individual. The cerebral cortex registration model is obtained by pre-training.

[0051] The cerebral cortex registration model of the present application can significantly improve the robustness of the model in registration between different individuals, avoid the deviation and distortion caused by a fixed template, and thus improve the accuracy of the registration results.

[0052] The method for processing neurological imaging data provided by the embodiment of the present invention obtains the structural partition under the cerebral cortex corresponding to the brain structural imaging data based on the brain structural imaging data and the brain structural segmentation model; obtains the level set corresponding to the brain structural imaging data based on the brain structural imaging data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structural imaging data, and the cerebral cortex reconstruction model, and reconstructs the cerebral cortex image based on the level set corresponding to the brain structural imaging data to obtain the reconstructed cerebral cortex image data corresponding to the brain structural imaging data; performs spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structural imaging data and the cerebral cortex registration model to obtain the spherical surface of the cerebral cortex surface corresponding to the brain structural imaging data after registration, which has higher robustness and adaptability, effectively overcomes the shortcomings of traditional technology in processing neurological imaging data of specific patients, and improves the accuracy of neurological imaging data processing.

[0053] Figure 2 is a flow chart of a method for processing neuroimaging data provided by a second embodiment of the present invention. Figure 2 As shown, on the basis of the above embodiments, further, the spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model is performed to obtain the registered spherical surface of the brain corresponding to the brain structure image data, including:

[0054] S201, performing expansion and spherical parameterization processing on the reconstructed cerebral cortex image data corresponding to the brain structure image data, to obtain the original unregistered spherical surface and sulcus depth information of the cerebral cortex surface corresponding to the brain structure image data;

[0055] Specifically, the server expands and spherical parameterizes the reconstructed cerebral cortex image data corresponding to the brain structure image data, and can obtain the original unregistered spherical surface and sulcus depth information of the cerebral cortex surface corresponding to the brain structure image data. By appropriately expanding the cerebral cortex, the surface details of the cortex can be more clearly displayed. The cerebral cortex has a complex sulcus structure. When the expansion process is not performed, some deeper sulci may be blocked and difficult to observe comprehensively. The expanded cortex can fully expand the sulci, and each area of ​​the cerebral cortex can be seen more intuitively, and the sulcus depth information can be obtained. Spherical parameterization refers to the process of gradually adjusting the shape of the cerebral cortex to an approximate sphere. Through spherical parameterization, the morphology of the cerebral cortex can be made more regular, which is convenient for subsequent analysis and processing.

[0056] S202, performing first standard space sampling according to the spherical surface of the cerebral cortex surface in the first standard space, the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data, and the original unregistered sulcus and gyrus depth information of the cerebral cortex surface corresponding to the brain structure image data, to obtain the sulcus and gyrus depth information of the cerebral cortex surface after the first standard space sampling corresponding to the brain structure image data;

[0057] Specifically, the server performs first standard space sampling based on the spherical surface of the cerebral cortex surface in the first standard space, the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data, and the original unregistered sulcus depth information of the cerebral cortex surface corresponding to the brain structure image data, and maps the original unregistered sulcus depth information of the cerebral cortex surface corresponding to the brain structure image data to the first standard space, so as to obtain the sulcus depth information of the cerebral cortex surface after sampling in the first standard space corresponding to the brain structure image data. Among them, the first standard space can adopt the fsaverage standard space.

[0058] S203, obtaining a deformation matrix of individual space registration to the first standard space based on the spherical surface of the cerebral cortex in the first standard space, the sulcus and gyrus depth information of the cerebral cortex surface after sampling in the first standard space corresponding to the brain structure image data, the sulcus and gyrus depth information of the template in the first standard space, and the cerebral cortex registration model.

[0059] Specifically, the server inputs the spherical surface of the cerebral cortex surface in the first standard space, the sulcus and gyrus depth information of the cerebral cortex surface after sampling in the first standard space corresponding to the brain structure image data, and the sulcus and gyrus depth information of the template in the first standard space into the cerebral cortex registration model, and can output the deformation matrix of the individual space registered to the first standard space.

[0060] S204, upsampling to obtain the deformation matrix of the original unregistered spherical space of the cerebral cortex surface according to the deformation matrix of the individual space registered to the first standard space, the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data, and the spherical surface of the cerebral cortex surface in the first standard space; obtaining the registered spherical surface of the cerebral cortex surface corresponding to the brain structure image data according to the deformation matrix of the original unregistered spherical space of the cerebral cortex surface and the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data.

[0061] Specifically, the server can upsample to obtain the deformation matrix of the original unregistered spherical space of the cerebral cortex surface based on the deformation matrix of the individual space registered to the first standard space, the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data, and the spherical surface of the cerebral cortex surface in the first standard space, and then apply the deformation matrix of the original unregistered spherical space of the cerebral cortex surface to the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data, so as to obtain the registered spherical surface of the cerebral cortex surface corresponding to the brain structure image data.

[0062] Figure 3 is a flowchart of a method for processing neuroimaging data provided by a third embodiment of the present invention. Figure 3 As shown, based on the above embodiments, further, the step of training the brain structure segmentation model includes:

[0063] S301, obtaining first training data according to first brain structure image sample data and cerebral cortical structure and subcutaneous structure partitions corresponding to the first brain structure image sample data;

[0064] Specifically, historical brain structure image data can be collected, and a first preset number of historical brain structure image data can be obtained as first brain structure image sample data. The first brain structure image sample data can be normalized, and the intensity of the first brain structure image sample data can be converted to between 0 and 1. The first brain structure image sample data can be annotated with a neuroimaging preprocessing tool to perform structural partitioning data annotation processing, and the cerebral cortex structural partition and subcortical structural partition of the first brain structure image sample data can be obtained. The normalized first brain structure image sample data, the cerebral cortex structural partition and subcortical structural partition of the first brain structure image sample data are used as the first training data.

[0065] S302: According to the first training data and the first original model, obtain the brain structure segmentation model through training.

[0066] Specifically, the server performs model training on the first original model through the first training data, and can train to obtain the brain structure segmentation model. Among them, the first original model can adopt FastSurferCNN, which is a segmentation model based on deep learning and uses the U-Net segmentation network. FastSurferCNN can quickly process a large amount of neuroimaging data. Compared with traditional methods, the segmentation and analysis of brain structures can be completed in a shorter time, which improves the efficiency of research and clinical applications.

[0067] Since the first training data includes the cerebral cortex structure partitions and the subcortical structure partitions corresponding to the first brain structure image sample data, and uses brain structure image data of different modalities and resolutions, the generalization ability and robustness of the brain structure segmentation model can be enhanced.

[0068] Figure 4 is a flowchart of a method for processing neuroimaging data provided by a fourth embodiment of the present invention. Figure 4 As shown, based on the above embodiments, further, the step of training the cerebral cortex reconstruction model includes:

[0069] S401, obtaining second training data according to the second brain structure image sample data, the white matter partition part in the structural partition under the cerebral cortex corresponding to the second brain structure image sample data, and the label corresponding to the white matter partition part;

[0070] Specifically, historical brain structure image data can be collected to obtain a second preset amount of historical brain structure image data as second brain structure image sample data, and the second brain structure image sample data can be normalized to convert the intensity of the second brain structure image sample data to between 0 and 1.

[0071] The conventional neuroimaging preprocessing tool can be used to perform structural partition data annotation processing on the second brain structure image sample data, and the subcortical structural partition corresponding to the second brain structure image sample data can be obtained, and the subcortical structural partition includes a white matter partition part and a label corresponding to the white matter partition part. In addition, after the brain structure segmentation model is obtained through training, the white matter partition part in the subcortical structural partition corresponding to the second brain structure image sample data can be obtained according to the second brain structure image sample data and the brain structure segmentation model. The normalized second brain structure image sample data, the white matter partition part in the subcortical structural partition corresponding to the second brain structure image sample data, and the label corresponding to the white matter partition part are used as the second training data.

[0072] S402: According to the second training data and the second original model, obtain the cerebral cortex reconstruction model through training.

[0073] Specifically, the server performs model training on the second original model through the second training data to obtain the cerebral cortex reconstruction model through training. The second original model may adopt a deep learning model with a 3D-U-Net structure.

[0074] Figure 5 is a flowchart of a method for processing neuroimaging data provided by a fifth embodiment of the present invention. Figure 5As shown, based on the above embodiments, further, the step of training the cerebral cortex registration model includes:

[0075] S501, obtaining cerebral cortex image sample data;

[0076] Specifically, historical cerebral cortex image data may be collected, and a third preset amount of cerebral cortex image data may be obtained therefrom as cerebral cortex image sample data.

[0077] S502, performing expansion and spherical parameterization processing on the cerebral cortex image sample data to obtain the original unregistered spherical surface and sulcus depth information of the cerebral cortex surface corresponding to the cerebral cortex image sample data;

[0078] Specifically, the server performs expansion and spherical parameterization processing on the cerebral cortex image sample data, and can obtain the original unregistered spherical surface of the cerebral cortex surface corresponding to the cerebral cortex image sample data and the original unregistered sulcus depth information of the cerebral cortex surface corresponding to the cerebral cortex image sample data.

[0079] S503, performing first standard space sampling according to the spherical surface of the cerebral cortex surface in the first standard space, the original unregistered spherical surface of the cerebral cortex surface corresponding to the cerebral cortex image sample data, and the original unregistered sulcus and gyrus depth information of the cerebral cortex surface corresponding to the cerebral cortex image sample data, to obtain the sulcus and gyrus depth information of the cerebral cortex surface after the first standard space sampling corresponding to the cerebral cortex image sample data; and using the spherical surface of the cerebral cortex surface in the first standard space, the template sulcus and gyrus depth information of the first standard space, the sulcus and gyrus depth information after the first standard space sampling corresponding to the cerebral cortex image sample data, and the corresponding partition label data as third training data;

[0080] Specifically, the server maps the original unregistered sulcus depth information of the cerebral cortex surface corresponding to the cerebral cortex image sample data to the first standard space, and can obtain the left and right brain sampling information after the first standard space sampling corresponding to the cerebral cortex image sample data. The server obtains the labels corresponding to the left and right brain sampling information after the standard space sampling corresponding to the cerebral cortex image sample data, and uses the spherical surface of the cerebral cortex surface in the first standard space, the template sulcus depth information of the first standard space, the sulcus depth information after the first standard space sampling corresponding to the cerebral cortex image sample data, and the corresponding partition label data as the third training data.

[0081] S504. According to the third training data and the third original model, the cerebral cortex registration model is obtained through training; wherein, in the process of obtaining the cerebral cortex registration model through training, the loss value corresponding to the third training data is calculated based on the improved loss function.

[0082] Specifically, the server performs model training on the third original model through the third training data to obtain the cerebral cortex reconstruction model. The third original model can adopt a spherical geometric neural network deep learning model based on a U-Net structure, such as S-GAT and Sphrical U-Net.

[0083] In the process of training to obtain the cerebral cortex registration model, the loss value corresponding to the third training data can be calculated based on the improved loss function. The improved loss function is used to reduce deformation and improve the registration accuracy. The improved loss function improves the regularization term and adds a supervision term relative to the prior art. By improving the calculation of the area loss and angle loss in the regularization term, it is helpful to reduce distortion. By adding the supervision term, the contrast loss between coordinates is increased, and the areas with large local deformation are reduced, so that the surface of the registration result is more uniform and the stability of the model is also improved.

[0084] On the basis of the above embodiments, further, the spherical surface of the cerebral cortex surface in the first standard space, the template sulcus and gyrus depth information in the first standard space, and the sulcus and gyrus depth information after sampling in the first standard space corresponding to the cerebral cortex image sample data are used as third training data to train the third original model, so as to obtain the cerebral cortex registration model through training.

[0085] At this time, the training adopts the unsupervised model training method. During the model training process, the loss function that can be used is

[0086] On the basis of the above embodiments, further, the improved loss function is:

[0087]

[0088] in, represents the loss value, represents the similarity term, represents the regularization term, represents the split term, represents the supervision item, express The corresponding weight, λ parc express The corresponding weight, λ comp express The corresponding weights; represents the area loss, represents the angle loss, Represents the distance loss, represents the folding loss, λ area express The corresponding weight, λangle express The corresponding weight, λ dist express The corresponding weight, λ fold Show The corresponding weights; N represents the number of vertices of non-rigid registration, M represents the deformed moving sphere, Φ represents the rotation tensor, C represents the contrast sphere; T represents the number of triangles, represents the original area of ​​the triangle, A t is the area of ​​the deformed triangle, t is a positive integer and t is less than or equal to T; Represents the angle value of the τth angle of the triangle, ∠t τ Represents the angle value of the τth angle of the deformed triangle, where τ is a positive integer and τ is less than or equal to 3.

[0089] Specifically, since the surface of the ball after registration by the algorithm in the prior art may be distorted in some local areas due to some extreme values, the area loss is improved to address the problem of distortion extreme values. and angle loss Expressed in the form of an index, it increases the focus on the maximum value and can reduce distortion. At the same time, the uniformity and accuracy of the registration sphere surface are optimized, and a supervision item is added. Increasing the contrast loss between coordinates reduces the areas with large local deformation, making the surface of the registration result more uniform and improving the stability of the model.

[0090] in, is the similarity term, and the mean square error is used to estimate the similarity of rigid and non-rigid registration.

[0091]

[0092] N represents the number of vertices for non-rigid registration, M represents the deformed moving sphere, Φ represents the rotation tensor, and F represents the fixed sphere.

[0093]

[0094] represents the area loss, represents the angle loss, Represents the distance loss, represents the folding loss, λ area express The corresponding weight, λ angle express The corresponding weight, λ dist express The corresponding weight, λ fold Show The corresponding weight.

[0095]

[0096] T represents the number of triangles, represents the original area of ​​the triangle, A t is the area of ​​the deformed triangle, t is a positive integer and t is less than or equal to T;

[0097]

[0098] Represents the angle value of the τth angle of the triangle, ∠t τ represents the angle value of the τth angle of the deformed triangle, where τ is a positive integer and τ is less than or equal to 3;

[0099]

[0100] d(·) represents the Euclidean distance, represents the original coordinates of the centroid of the 1-hop neighborhood, represents the original coordinates of the centroid of the 1-hop neighborhood, v′ C represents the updated coordinates of the centroid of the 1-hop neighborhood after deformation, v′ B represents the updated coordinates of the centroid of the 1-hop neighborhood after deformation, ∈ l Indicates the average side length between two points of a standard sphere in the current level.

[0101]

[0102] Δ t represents the current area of ​​the t-th triangle, represents the original area of ​​the t-th triangle, v i,j represents the edge vector from vertex i to vertex j of the t-th triangle, v i,k represents the edge vector from vertex i to vertex k of the t-th triangle, n represents the unit normal vector of the t-th triangle, and n points outward from the origin of the sphere.

[0103]

[0104] P represents the number of partitions, p is a positive integer and p is less than or equal to P. represents the pth partition on the moving sphere, F p represents the pth partition on the fixed sphere;

[0105]

[0106] N represents the number of vertices of non-rigid registration, M represents the deformed moving sphere, Φ represents the rotation tensor, and C represents the comparison sphere. C can be the spherical surface registered by the MSM_strain method.

[0107] Figure 6 is a flow chart of a method for processing neuroimaging data provided by a sixth embodiment of the present invention. Figure 6 Based on the above embodiments, the method for processing neuroimaging data provided by the embodiment of the present invention further includes:

[0108] S601, based on the individual's brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and the cerebral cortex reconstruction model, obtain a level set corresponding to the individual's brain structure image data, and reconstruct the cerebral cortex image based on the level set corresponding to the individual's brain structure image data to obtain reconstructed cerebral cortex image data corresponding to the individual's brain structure image data;

[0109] Specifically, the server can obtain a level set corresponding to the individual's brain structure image data based on the individual's brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the individual's brain structure image data, and the cerebral cortex reconstruction model. The server reconstructs the cerebral cortex image according to the level set corresponding to the individual's brain structure image data to obtain the reconstructed cerebral cortex image data corresponding to the individual's brain structure image data.

[0110] S602, obtaining a white matter surface under the cerebral cortex corresponding to the individual's brain structure image data based on the reconstructed cerebral cortex image data corresponding to the individual's brain structure image data;

[0111] Specifically, the server can obtain the white matter surface under the cerebral cortex corresponding to the individual's brain structure image data based on the reconstructed cerebral cortex image data corresponding to the individual's brain structure image data. The white matter surface under the cerebral cortex corresponding to the individual's brain structure image data is regarded as the white matter surface under the cerebral cortex corresponding to the individual's brain structure image data in the first individual space.

[0112] S603, spatially registering the brain structure image data of the individual in the first individual space, the white matter surface under the cerebral cortex corresponding to the brain structure image data of the individual in the first individual space, and the neural image data of the individual in the second individual space, to obtain volume data of the neural image data of the individual in the second individual space in the first individual space;

[0113] Specifically, the server spatially registers the neuroimaging data of the individual in the second individual space based on the brain structural imaging data of the individual in the first individual space, the white matter surface under the cerebral cortex corresponding to the brain structural imaging data of the individual in the first individual space, and can obtain the volume data of the neuroimaging data of the individual in the second individual space in the first individual space. The brain structural imaging data of the individual in the first individual space can be T1-weighted imaging (T1w) data, the neuroimaging data of the individual in the second space can be blood oxygen level-dependent functional magnetic resonance imaging (BOLD) data, etc., and the neuroimaging data of the individual in the second individual space can be structural neuroimaging data or neuroimaging data of any modality. Among them, the bbregister method can be used to register the volume data of the individual's other individual spaces to the brain structural imaging data of the first individual space, and obtain the volume data of the neuroimaging data of the individual in the other individual spaces in the first space. Neuroimaging data can be divided into brain structural imaging data and brain functional imaging data.

[0114] S604, based on the brain structure image data of the individual in the first individual space, the neuroimage data of the second standard template space and the brain space standardization model, obtaining a deformation field of the first individual space registered to the second standard template space of the individual; wherein the brain space standardization model is obtained by pre-training;

[0115] Specifically, the server inputs the brain structure image data of the individual in the first individual space and the neuroimage data of the second standard template space into the brain space standardization model, and can output the deformation field of the first individual space registered to the second standard template space. Among them, the brain neuroimage data of the standard template space can adopt the T1w data of the second standard template space. The second standard template space can adopt the MNI standard template space.

[0116] Furthermore, the brain structure image data of the individual in the first individual space and the neural image data in the second standard template space can be downsampled, and a suitable sampling rate can be set so that the sampling rate can reduce the occupation of video memory without affecting the accuracy.

[0117] S605, applying the deformation field of the first body space of the individual registered to the second standard template space to the volume data of the neural imaging data of the individual in the second body space of the individual, to obtain the volume data of the neural imaging data of the individual in the second body space on the second standard template space.

[0118] Specifically, the server applies the deformation field of the individual's first body space registered to the second standard template space to the volume data of the individual's neural imaging data in the second individual space on the individual's first body space, and can obtain the volume data of the individual's neural imaging data in the second individual space on the second standard template space.

[0119] The accuracy of registering individual data of different modalities to the data of the second standard template space through the above steps S601 to S605 is better than the result of directly using the brain space standardization model for cross-modal registration.

[0120] Figure 7 is a flowchart of a method for processing neuroimaging data provided by the seventh embodiment of the present invention. Figure 7 As shown, based on the above embodiments, further, training the brain space normalization model includes:

[0121] S701, obtaining brain structure image sample data and neural image sample data of a second standard template space as fourth training data;

[0122] Specifically, historical brain structure image data can be collected, and brain structure image data of a fourth preset number of individuals in the individual space can be obtained as brain structure image sample data based on the historical brain structure image data, and the neural image sample data of the second standard template space can be obtained. The brain structure image sample data and the neural image sample data of the second standard template space are used as the fourth training data. The brain structure image data can be the volume data of the individual in the weighted imaging space, which is selected according to actual needs and is not limited in the embodiments of the present invention.

[0123] S702: According to the fourth training data and the fourth original model, obtain the brain space normalization model through training.

[0124] Specifically, the server performs model training on the fourth original model through the fourth training data to obtain the brain space normalization model. The fourth original model may adopt a U-Net network structure model.

[0125] It is understandable that, based on the individual's neuroimaging data in the first individual space, the neuroimaging data in the second standard template space and the brain space standardization model, the deformation field of the first individual space registration to the second standard template space can be obtained, so that the neuroimaging data of any modality of the individual can be configured to the second standard template space. At this time, the brain space standardization model should be based on the training data composed of the neuroimaging sample data and the neuroimaging sample data of the second standard template space, and obtained through training.

[0126] The method for processing neuroimaging data provided by the embodiment of the present invention, with its higher robustness and adaptability, effectively overcomes the shortcomings of traditional technology in processing neuroimaging data of specific patients, and provides a more accurate and efficient preprocessing solution. The application of the method for processing neuroimaging data provided by the embodiment of the present invention can not only improve the accuracy of medical imaging analysis, but also provide a more solid scientific basis for clinical decision-making, and lay a better foundation for the diagnosis and treatment of patients.

[0127] Figure 8 is a schematic diagram of the structure of a neural image data processing device provided by an eighth embodiment of the present invention. Figure 8 As shown, the neural image data processing device provided by the embodiment of the present invention includes a segmentation unit 801, a reconstruction unit 802 and a registration unit 803, wherein:

[0128] The segmentation unit 801 is used to obtain the structural partition under the cerebral cortex corresponding to the brain structure image data based on the brain structure image data and the brain structure segmentation model; wherein the brain structure segmentation model is obtained by pre-training; the reconstruction unit 802 is used to obtain the level set corresponding to the brain structure image data based on the brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and the cerebral cortex reconstruction model, and reconstruct the cerebral cortex image based on the level set corresponding to the brain structure image data to obtain the reconstructed cerebral cortex image data corresponding to the brain structure image data; wherein the cerebral cortex reconstruction model is obtained by pre-training; the registration unit 803 is used to perform spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model to obtain the spherical surface of the cerebral cortex surface corresponding to the brain structure image data after registration; wherein the cerebral cortex registration model is obtained by pre-training.

[0129] Specifically, brain structure image data can be obtained by magnetic resonance imaging (MRI) imaging equipment or CT imaging equipment. For example, brain structure image data includes T1 weighted data, T2 weighted data, diffusion tensor imaging, CT and other modal data. The segmentation unit 801 obtains the structural partition under the cerebral cortex and the structural partition outside the cerebral cortex corresponding to the brain structure image data based on the brain structure image data and the brain structure segmentation model. In order to compare and analyze the data on the same scale, the brain structure image data is normalized so that the intensity of the brain structure image data is converted to between 0 and 1, and the normalized brain structure image data is input into the brain structure segmentation model, and the structural partition under the cerebral cortex and the structural partition outside the cerebral cortex corresponding to the brain structure image data are output. The structural partition under the cerebral cortex corresponding to the brain structure image data includes a white matter partition part. The brain structure segmentation model can distinguish the cerebral cortex and subcortical structure in the brain structure image data. Among them, the brain structure image data can use T1 weighted (T1-weighted) data.

[0130] The reconstruction unit 802 can obtain the level set corresponding to the brain structure image data based on the brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and the cerebral cortex reconstruction model. In order to compare and analyze the data on the same scale, the brain structure image data is normalized so that the intensity of the brain structure image data is converted to between 0 and 1, and the normalized brain structure image data and the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data are input into the cerebral cortex reconstruction model together, and the level set corresponding to the brain structure image data can be output. Among them, the cerebral cortex reconstruction model is obtained by pre-training. The reconstruction unit 802 reconstructs the cerebral cortex image according to the level set corresponding to the brain structure image data to obtain the reconstructed cerebral cortex image data corresponding to the brain structure image data.

[0131] The registration unit 803 performs registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model, and can obtain the spherical surface of the cerebral cortex surface after registration corresponding to the brain structure image data. The reconstructed cerebral cortex image data corresponding to the brain structure image data is inflated and spherical parameterized, and the original unregistered spherical surface and gyrus depth information corresponding to the brain structure image data can be obtained; the original unregistered spherical surface and gyrus depth information corresponding to the brain structure image data are sampled to the first standard space, and the spherical surface of the cerebral cortex surface sampled in the first standard space corresponding to the brain structure image data can be obtained; the spherical surface of the cerebral cortex surface in the first standard space, the spherical surface of the cerebral cortex surface sampled in the first standard space corresponding to the brain structure image data, and the template gyrus of the first standard space are sampled. The depth information is input into the cerebral cortex registration model, and the deformation matrix of the individual space registration to the first standard space can be obtained; according to the deformation matrix of the individual space registration to the first standard space, the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data and the spherical surface of the cerebral cortex surface in the first standard space, the deformation matrix of the original unregistered spherical space of the cerebral cortex surface can be obtained, and then the deformation matrix of the original unregistered spherical space of the cerebral cortex surface is applied to the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data, and the spherical surface of the cerebral cortex surface after registration corresponding to the brain structure image data can be obtained. Among them, the first standard space is a universal, standardized reference space, and the fsaverage standard space can be used. In the analysis of cerebral cortex image data, mapping the brain images of different individuals into the first standard space can realize comparison and statistical analysis between different individuals. The individual space is relative to the first standard space, and refers to the spatial structure unique to the brain of a specific individual. The cerebral cortex registration model is obtained by pre-training.

[0132] The neural imaging data processing device provided by the embodiment of the present invention obtains the structural partition under the cerebral cortex corresponding to the brain structural imaging data based on the brain structural imaging data and the brain structural segmentation model; obtains the level set corresponding to the brain structural imaging data based on the brain structural imaging data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structural imaging data, and the cerebral cortex reconstruction model, and reconstructs the cerebral cortex image based on the level set corresponding to the brain structural imaging data to obtain the reconstructed cerebral cortex image data corresponding to the brain structural imaging data; performs spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structural imaging data and the cerebral cortex registration model to obtain the spherical surface of the cerebral cortex surface corresponding to the brain structural imaging data after registration, which has higher robustness and adaptability, effectively overcomes the shortcomings of traditional technology in processing neural imaging data of specific patients, and improves the accuracy of neural imaging data processing.

[0133] Fig. 9 is a schematic diagram of the structure of a neural image data processing device provided by the ninth embodiment of the present invention. Fig. 9 As shown, on the basis of the above embodiments, further, the registration unit 803 includes a first obtaining subunit 8031, a standard space sampling subunit 8032, a second obtaining subunit 8033 and an individual space sampling subunit 8034, wherein:

[0134] The first acquisition subunit 8031 ​​is used to perform expansion and spherical parameterization processing on the reconstructed cerebral cortex image data corresponding to the brain structure image data, so as to obtain the original unregistered spherical surface and sulcus depth information of the cerebral cortex surface corresponding to the brain structure image data; the standard space sampling subunit 8032 is used to perform the first standard space sampling according to the spherical surface of the cerebral cortex in the first standard space, the original unregistered spherical surface of the brain corresponding to the brain structure image data, and the original unregistered sulcus depth information of the cerebral cortex corresponding to the brain structure image data, so as to obtain the sulcus depth information of the cerebral cortex surface after the first standard space sampling corresponding to the brain structure image data; the second acquisition subunit 8033 is used to perform the first standard space sampling according to the spherical surface of the cerebral cortex surface in the first standard space, the original unregistered sulcus depth information of the cerebral cortex corresponding to the brain structure image data The deformation matrix of the individual space registered to the first standard space is obtained by sampling the sulcus and gyrus depth information of the cerebral cortex surface after sampling in the first standard space, the template sulcus and gyrus depth information of the first standard space, and the cerebral cortex registration model; the individual space sampling subunit 8034 is used to upsample the deformation matrix of the original unregistered spherical space of the cerebral cortex surface according to the deformation matrix of the individual space registered to the first standard space, the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data, and the spherical surface of the cerebral cortex surface in the first standard space; the registered spherical surface of the cerebral cortex surface corresponding to the brain structure image data is obtained according to the deformation matrix of the original unregistered spherical space of the cerebral cortex surface and the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data.

[0135] Fig.10 is a schematic diagram of the structure of a neural image data processing device provided by the tenth embodiment of the present invention. Fig.10 As shown, based on the above embodiments, further, the neural image data processing device provided by the embodiment of the present invention further includes a first obtaining unit 804 and a first training unit 805, wherein:

[0136] The first acquisition unit 804 is used to obtain the first training data according to the first brain structure image sample data and the cerebral cortex structure and subcutaneous structure partitions corresponding to the first brain structure image sample data; the first training unit 805 is used to train and obtain the brain structure segmentation model according to the first training data and the first original model.

[0137] Fig.11 is a schematic diagram of the structure of a neural image data processing device provided by the eleventh embodiment of the present invention. Fig.11 As shown, based on the above embodiments, further, the apparatus for processing neural image data provided by the embodiment of the present invention further includes a second obtaining unit 806 and a second training unit 807, wherein:

[0138] The second acquisition unit 806 is used to obtain second training data based on the second brain structure image sample data, the white matter partition part in the structural partition under the cerebral cortex corresponding to the second brain structure image sample data, and the label corresponding to the white matter partition part; the second training unit 807 is used to train and obtain the cerebral cortex reconstruction model based on the second training data and the second original model.

[0139] Fig.12 is a schematic diagram of the structure of a neural image data processing device provided by the twelfth embodiment of the present invention. Fig.12 As shown, on the basis of the above embodiments, further, the apparatus for processing neuroimaging data provided by the embodiment of the present invention further includes a first acquisition unit 808, a third acquisition unit 809, a sample sampling unit 810 and a third training unit 811, wherein:

[0140] The first acquisition unit 808 is used to acquire cerebral cortex image sample data; the third acquisition unit 809 is used to perform expansion and spherical parameterization processing on the cerebral cortex image sample data to obtain the original unregistered spherical surface and sulcus depth information of the cerebral cortex surface corresponding to the cerebral cortex image sample data; the sample sampling unit 810 is used to perform first standard space sampling according to the spherical surface of the cerebral cortex surface in the first standard space, the original unregistered spherical surface of the cerebral cortex surface corresponding to the cerebral cortex image sample data, and the original unregistered sulcus depth information of the cerebral cortex surface corresponding to the cerebral cortex image sample data, to obtain the cerebral The sulcus depth information of the cerebral cortex surface after sampling in the first standard space corresponding to the cortical image sample data; and the spherical surface of the cerebral cortex surface in the first standard space, the template sulcus depth information in the first standard space, the sulcus depth information of the first standard space corresponding to the cerebral cortex image sample data after sampling, and the corresponding partition label data are used as the third training data; the third training unit 811 is used to train and obtain the cerebral cortex registration model according to the third training data and the third original model; wherein, in the process of training to obtain the cerebral cortex registration model, the loss value corresponding to the third training data is calculated based on the improved loss function.

[0141] Fig.13 is a schematic diagram of the structure of a neural image data processing device provided by the thirteenth embodiment of the present invention. Fig.13 As shown, on the basis of the above embodiments, further, the neural image data processing device provided in the embodiment of the present invention further includes a reconstruction unit 812, a fourth obtaining unit 813, an individual registration unit 814, a fifth obtaining unit 815 and a sixth obtaining unit 816, wherein:

[0142] The reconstruction unit 812 is used to obtain the level set corresponding to the brain structure image data of the individual based on the brain structure image data of the individual, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and the cerebral cortex reconstruction model, and reconstruct the cerebral cortex image based on the corresponding level set of the brain structure image data of the individual to obtain the reconstructed cerebral cortex image data corresponding to the brain structure image data of the individual; the fourth obtaining unit 813 is used to obtain the white matter surface under the cerebral cortex corresponding to the brain structure image data of the individual based on the reconstructed cerebral cortex image data corresponding to the brain structure image data of the individual; the individual registration unit 814 is used to obtain the white matter surface under the cerebral cortex corresponding to the brain structure image data of the individual in the first individual space, the white matter surface under the cerebral cortex corresponding to the brain structure image data of the individual in the first individual space, and the white matter surface under the cerebral cortex corresponding to the brain structure image data of the individual in the second individual space. The fifth obtaining unit 815 is used to obtain the deformation field of the first volume space of the individual registered to the second standard template space based on the brain structure image data of the individual in the first volume space, the neuroimaging data of the second standard template space and the brain space standardization model; wherein the brain space standardization model is obtained by pre-training; the sixth obtaining unit 816 is used to apply the deformation field of the first volume space of the individual registered to the second standard template space to the neuroimaging data of the individual in the second volume space on the volume data of the first volume space of the individual, and obtain the volume data of the neuroimaging data of the individual in the second volume space on the second standard template space.

[0143] Fig.14 is a schematic diagram of the structure of a neural image data processing device provided by the fourteenth embodiment of the present invention. Fig.14 As shown, based on the above embodiments, further, the apparatus for processing neural image data provided by the embodiment of the present invention further includes a second acquisition unit 817 and a fourth training unit 818, wherein:

[0144] The second acquisition unit 817 is used to acquire brain structure image sample data and neural image sample data of the second standard template space as fourth training data; the fourth training unit 818 is used to train and obtain the brain space standardization model according to the fourth training data and the fourth original model.

[0145] The embodiments of the apparatus provided in the embodiments of the present invention can be specifically used to execute the processing flow of the above-mentioned method embodiments, and its functions are not described in detail here, and reference can be made to the detailed description of the above-mentioned method embodiments.

[0146] Fig.15 is a schematic diagram of the physical structure of an electronic device provided in the fifteenth embodiment of the present invention, such as Fig.15As shown, the electronic device may include: a processor (processor) 1501, a communication interface (Communications Interface) 1502, a memory (memory) 1503 and a communication bus 1504, wherein the processor 1501, the communication interface 1502, and the memory 1503 communicate with each other through the communication bus 1504. The processor 1501 can call the logic instructions in the memory 1503 to execute the methods provided by the above-mentioned method embodiments, for example, including: based on the brain structure image data and the brain structure segmentation model, obtaining the structural partition under the cerebral cortex corresponding to the brain structure image data; wherein the brain structure segmentation model is obtained by pre-training; based on the brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and the cerebral cortex reconstruction model, obtaining the level set corresponding to the brain structure image data, and reconstructing the cerebral cortex image based on the level set corresponding to the brain structure image data to obtain the reconstructed cerebral cortex image data corresponding to the brain structure image data; wherein the cerebral cortex reconstruction model is obtained by pre-training; based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model, performing spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space, obtaining the spherical surface of the cerebral cortex surface corresponding to the brain structure image data after registration; wherein the cerebral cortex registration model is obtained by pre-training.

[0147] In addition, the logic instructions in the above-mentioned memory 1503 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0148] The present embodiment discloses a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the methods provided by the above-mentioned method embodiments, for example, including: based on brain structure image data and a brain structure segmentation model, obtaining a subcortical structural partition corresponding to the brain structure image data; wherein the brain structure segmentation model is obtained by pre-training; based on the brain structure image data, the subcortical structural partition corresponding to the brain structure image data includes a white matter partition part and A cerebral cortex reconstruction model is provided, and a level set corresponding to the brain structure image data is obtained, and a cerebral cortex image is reconstructed based on the level set corresponding to the brain structure image data to obtain the reconstructed cerebral cortex image data corresponding to the brain structure image data; wherein the cerebral cortex reconstruction model is obtained by pre-training; based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model, a spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space is performed to obtain the spherical surface of the cerebral cortex surface corresponding to the brain structure image data after registration; wherein the cerebral cortex registration model is obtained by pre-training.

[0149] The present embodiment provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program / instruction, wherein the computer program / instruction enables the computer to execute the methods provided by the above-mentioned method embodiments, for example, including: obtaining the structural partition under the cerebral cortex corresponding to the brain structure image data based on brain structure image data and a brain structure segmentation model; wherein the brain structure segmentation model is obtained by pre-training; obtaining the level set corresponding to the brain structure image data based on the brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and the cerebral cortex reconstruction model, and reconstructing the cerebral cortex image based on the level set corresponding to the brain structure image data to obtain the reconstructed cerebral cortex image data corresponding to the brain structure image data; wherein the cerebral cortex reconstruction model is obtained by pre-training; performing spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model to obtain the spherical surface of the cerebral cortex surface corresponding to the brain structure image data after registration; wherein the cerebral cortex registration model is obtained by pre-training.

[0150] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0151] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0152] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0153] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0154] In the description of this specification, the description with reference to the terms "one embodiment", "a specific embodiment", "some embodiments", "for example", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0155] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for processing neuroimaging data, characterized in that: include: Based on the brain structure image data and the brain structure segmentation model, obtaining the structural partition under the cerebral cortex corresponding to the brain structure image data; wherein the brain structure segmentation model is obtained by pre-training; Based on the brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and the cerebral cortex reconstruction model, a level set corresponding to the brain structure image data is obtained, and a cerebral cortex image is reconstructed based on the level set corresponding to the brain structure image data to obtain reconstructed cerebral cortex image data corresponding to the brain structure image data; wherein the cerebral cortex reconstruction model is obtained by pre-training; Based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model, the spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space is performed to obtain the spherical surface of the cerebral cortex surface corresponding to the brain structure image data after registration; wherein the cerebral cortex registration model is obtained by pre-training.

2. The method according to claim 1, characterized in that The spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model to obtain the registered spherical surface of the brain corresponding to the brain structure image data comprises: Performing expansion and spherical parameterization processing on the reconstructed cerebral cortex image data corresponding to the brain structure image data to obtain the original unregistered spherical surface and sulcus depth information of the cerebral cortex surface corresponding to the brain structure image data; Performing first standard space sampling according to the spherical surface of the cerebral cortex in the first standard space, the original unregistered spherical surface of the brain corresponding to the brain structure image data, and the original unregistered sulcus and gyrus depth information of the cerebral cortex corresponding to the brain structure image data, to obtain the sulcus and gyrus depth information of the cerebral cortex surface after the first standard space sampling corresponding to the brain structure image data; Obtaining a deformation matrix of individual space registration to the first standard space according to the spherical surface of the cerebral cortex in the first standard space, the sulcus depth information of the cerebral cortex surface after sampling in the first standard space corresponding to the brain structure image data, the sulcus depth information of the template in the first standard space, and the cerebral cortex registration model; According to the deformation matrix of the individual space registered to the first standard space, the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data, and the spherical surface of the cerebral cortex surface in the first standard space, upsampling is performed to obtain the deformation matrix of the original unregistered spherical space of the cerebral cortex surface; according to the deformation matrix of the original unregistered spherical space of the cerebral cortex surface and the original unregistered spherical surface of the cerebral cortex surface corresponding to the brain structure image data, the registered spherical surface of the cerebral cortex surface corresponding to the brain structure image data is obtained.

3. The method according to claim 1, characterized in that: The steps of training the brain structure segmentation model include: Obtaining first training data according to the first brain structure image sample data and the cerebral cortex structure and subcutaneous structure partitions corresponding to the first brain structure image sample data; The brain structure segmentation model is obtained through training according to the first training data and the first original model.

4. The method according to claim 1, characterized in that The steps of training the cerebral cortex reconstruction model include: Obtaining second training data according to the second brain structure image sample data, the white matter partition part in the structural partition under the cerebral cortex corresponding to the second brain structure image sample data, and the label corresponding to the white matter partition part; The cerebral cortex reconstruction model is obtained through training according to the second training data and the second original model.

5. The method according to claim 1, characterized in that The steps of training the cerebral cortex registration model include: Obtaining cerebral cortex image sample data; Performing expansion and spherical parameterization processing on the cerebral cortex image sample data to obtain original unregistered spherical surface and sulcus depth information of the cerebral cortex surface corresponding to the cerebral cortex image sample data; Performing first standard space sampling according to the spherical surface of the cerebral cortex surface in the first standard space, the original unregistered spherical surface of the cerebral cortex surface corresponding to the cerebral cortex image sample data, and the original unregistered sulcus and gyrus depth information of the cerebral cortex surface corresponding to the cerebral cortex image sample data, to obtain the sulcus and gyrus depth information of the cerebral cortex surface after the first standard space sampling corresponding to the cerebral cortex image sample data; and using the spherical surface of the cerebral cortex surface in the first standard space, the template sulcus and gyrus depth information of the first standard space, the sulcus and gyrus depth information after the first standard space sampling corresponding to the cerebral cortex image sample data, and the corresponding partition label data as third training data; The cerebral cortex registration model is obtained through training according to the third training data and the third original model; wherein, in the process of obtaining the cerebral cortex registration model through training, the loss value corresponding to the third training data is calculated based on the improved loss function.

6. The method according to claim 5, characterized in that The improved loss function is: in, represents the loss value, represents the similarity term, represents the regularization term, represents the split term, represents the supervision term, λ sim express The corresponding weight, λ parc express The corresponding weight, λ comp express The corresponding weights; represents the area loss, represents the angle loss, Represents the distance loss, represents the folding loss, λ area express The corresponding weight, λ angle express The corresponding weight, λ dist express The corresponding weight, λ fold Show The corresponding weights; N represents the number of vertices of non-rigid registration, M represents the deformed moving sphere, Φ represents the rotation tensor, C represents the contrast sphere; T represents the number of triangles, represents the original area of ​​the triangle, A t is the area of ​​the deformed triangle, t is a positive integer and t is less than or equal to T; Represents the angle value of the τth angle of the triangle, ∠t τ Represents the angle value of the τth angle of the deformed triangle, where τ is a positive integer and τ is less than or equal to 3.

7. The method according to any one of claims 1 to 6, characterized in that: Also includes: Based on the individual's brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and the cerebral cortex reconstruction model, a level set corresponding to the individual's brain structure image data is obtained, and a cerebral cortex image is reconstructed based on the level set corresponding to the individual's brain structure image data to obtain reconstructed cerebral cortex image data corresponding to the individual's brain structure image data; Based on the reconstructed cerebral cortex image data corresponding to the individual's brain structure image data, obtaining the white matter surface under the cerebral cortex corresponding to the individual's brain structure image data; Based on the brain structure image data of the individual in the first individual space, the white matter surface under the cerebral cortex corresponding to the brain structure image data of the individual in the first individual space, and the neuroimage data of the individual in the second individual space, spatial registration is performed to obtain the volume data of the neuroimage data of the individual in the second individual space in the first individual space; Based on the brain structure image data of the individual in the first individual space, the neuroimage data of the second standard template space and the brain space standardization model, a deformation field of the first individual space of the individual is registered to the second standard template space; wherein the brain space standardization model is obtained by pre-training; The deformation field of the first body space of the individual registered to the second standard template space is applied to the volume data of the neuroimaging data of the individual in the second body space on the first body space of the individual to obtain the volume data of the neuroimaging data of the individual in the second body space on the second standard template space.

8. The method according to claim 7, characterized in that Training the brain spatial normalization model includes: Acquire brain structure image sample data and neuroimage sample data of the second standard template space as fourth training data; The brain space normalization model is obtained through training according to the fourth training data and the fourth original model.

9. A device for processing neural image data, characterized in that: include: A segmentation unit, used for obtaining the structural partition under the cerebral cortex corresponding to the brain structure image data based on the brain structure image data and the brain structure segmentation model; wherein the brain structure segmentation model is obtained by pre-training; A reconstruction unit, configured to obtain a level set corresponding to the brain structure image data based on the brain structure image data, the white matter partition part included in the structural partition under the cerebral cortex corresponding to the brain structure image data, and a cerebral cortex reconstruction model, and reconstruct the cerebral cortex image based on the level set corresponding to the brain structure image data to obtain reconstructed cerebral cortex image data corresponding to the brain structure image data; wherein the cerebral cortex reconstruction model is obtained by pre-training; A registration unit is used to perform spherical registration of the cerebral cortex surface with the cerebral cortex surface in the first standard space based on the reconstructed cerebral cortex image data corresponding to the brain structure image data and the cerebral cortex registration model, so as to obtain the spherical surface of the cerebral cortex surface corresponding to the brain structure image data after registration; wherein the cerebral cortex registration model is obtained by pre-training.

10. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program / instruction, which implements the steps of the method according to any one of claims 1 to 8 when executed by a processor.

12. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images