Neuroelectrophysiological modeling method and system for stroke patients based on image completion

Abstract

The present invention provides a method and system for neuroelectrophysiological modeling of stroke patients based on image completion, the method comprising the steps of automatic segmentation of lesion areas, structural completion of lesion areas, fine segmentation of multi-tissues in the whole brain, construction of a multi-layer head model, EEG positive problem modeling, and modeling accuracy verification. The present invention achieves high-precision conversion of MRI images of stroke patients to individualized multi-layer head models by constructing an integrated system for stroke lesion identification, structural image completion, fine segmentation of multi-tissues, and conductivity parameter modeling. Compared with the traditional five-layer template method, this method can not only identify and repair lesion areas, improve structural integrity and electromagnetic simulation reliability, but also refine tissue division, construct an eight-layer head model including lesions, gray and white matter, and set conductivity parameters in a targeted manner, significantly improving the accuracy of power source positioning and neural regulation simulation, and providing basic support for personalized diagnosis and treatment of stroke patients.

Description

Technical Field

[0001] The present invention relates to the field of medical image processing and neuroelectrophysiology technology, and in particular to a method and system for neuroelectrophysiological modeling of stroke patients based on image completion. Background Art

[0002] Neuroelectrophysiological modeling is a key component in brain source localization and non-invasive neuromodulation. Its core is to construct a head model that reflects individual differences in tissue structure and electromagnetic properties. Currently, head models are often segmented and modeled based on magnetic resonance imaging images. A five-layer structure is often used: scalp, skull, cerebrospinal fluid, gray matter, and white matter. Each layer of tissue is assigned a fixed conductivity value to solve forward and inverse EEG problems. This modeling approach is relatively mature in healthy individuals and can support routine EEG signal analysis and neural stimulation simulation.

[0003] However, stroke patients often experience significant changes in brain tissue structure, including focal necrosis, brain atrophy, and liquefaction foci. These structural abnormalities not only alter the anatomical distribution but also cause a shift in the electrical conductivity characteristics of the lesion area, significantly impacting modeling accuracy. If head models are still constructed based on healthy templates or fixed tissue segmentation structures, they can easily lead to distortions in electric field simulation and source localization results, compromising precise neuromodulation and personalized treatment for stroke patients.

[0004] For example, existing technology (Malaysian patent application MYPI2013702521) discloses a method for constructing a head model and localizing brain sources based on combined EEG and fMRI analysis. This method uses the finite element method to establish a standard head model and combines it with fMRI to verify the source localization results. However, this solution is designed for healthy individuals and employs a standard five-layer structure. It does not consider anatomical changes and conductivity abnormalities in the stroke lesion area, lacks the ability to process images of structural defects, and is unsuitable for the personalized modeling needs of patients with structural brain diseases such as stroke.

[0005] Existing technology (Chinese Patent ZL202110488900.5) also discloses a method for training EEG traceability models. This method focuses on constructing a dataset mapping the relationship between neural activity sources and head electrodes, used to optimize the performance of traceability network models. This method, which generates a lead matrix based on a healthy template, does not address the modeling mechanism of stroke lesions, nor does it address image structure restoration and individual tissue conductivity modeling. Therefore, it has significant deficiencies in constructing individual head models and modeling electrophysiological abnormalities in stroke patients.

[0006] Therefore, a neuroelectrophysiological modeling method for stroke patients is urgently needed to solve the problems of difficulty in identifying structural abnormalities, inaccurate conductivity distribution, and rough tissue classification in modeling the head model of stroke patients. Summary of the Invention

[0007] In order to achieve the above-mentioned purpose and other advantages of the present invention, the first purpose of the present invention is to provide a method for neuroelectrophysiological modeling of stroke patients based on image completion, comprising the following steps:

[0008] Acquiring T1-weighted magnetic resonance imaging data of a stroke patient, and performing lesion area segmentation on the T1-weighted magnetic resonance imaging data;

[0009] Performing image completion on the lesion area to generate a structurally continuous brain tissue image;

[0010] performing high-precision structural segmentation on the brain tissue in the brain tissue image to identify multiple types of brain tissue;

[0011] The lesion labels obtained by lesion region segmentation are integrated with multiple brain tissue labels to construct a multi-tissue label map including the lesion. Multiple conductivity gradients are set for the lesion region based on tissue status, and standard conductivity values ​​are set for other tissues to generate a multi-layer personalized head model.

[0012] An individualized cortical gray matter surface model is constructed, and the electrode coordinate system is aligned with the individual MRI space through affine transformation to achieve precise alignment. The forward problem is solved to calculate the conduction matrix between the electrodes and the cortical source under the individualized head model to provide a basic physical model.

[0013] Furthermore, the step of segmenting the T1-weighted magnetic resonance imaging data into lesion areas includes:

[0014] The T1-weighted magnetic resonance imaging data is segmented into three-dimensional lesion areas based on the nn-UNet framework, and a three-dimensional label map is output to mark the location and range of the lesion.

[0015] Furthermore, the step of completing the image of the lesion area includes:

[0016] Based on the pix2pix conditional generative adversarial network, the image of the lesion area is completed to generate a structurally continuous brain tissue image.

[0017] Furthermore, the adversarial network adopts a U-Net structure with skip connections as a generator, and cooperates with the PatchGAN local discriminator to improve texture consistency. It uses ATLAS public stroke image data to construct a training set of healthy-pseudo-lesion image pairs, constructs degraded images by simulating lesion occlusion masks, and combines the L1 reconstruction loss with the perceptual consistency loss function to guide the generator to learn the structural prior of healthy tissue and achieve reasonable anatomical repair of the lesion area.

[0018] Furthermore, the step of performing high-precision structural segmentation on the brain tissue in the brain tissue image includes:

[0019] The brain tissue image after image completion is used as input, and the SPM12 and Freesurfer tool chains integrated in the Fieldtrip platform are used to perform high-precision structural segmentation of the brain tissue.

[0020] Furthermore, the multiple types of brain tissue identified include seven types of tissue: scalp, skull, cerebrospinal fluid, gray matter, white matter, cerebellar gray matter and cerebellar white matter.

[0021] Furthermore, after the step of performing high-precision structural segmentation on the brain tissue in the brain tissue image and identifying multiple types of brain tissue, the step further includes:

[0022] Based on the structural segmentation results, a seven-category nn-UNet segmentation model is constructed to achieve fast and automatic segmentation of new samples.

[0023] Furthermore, the step of fusing the lesion labels obtained by lesion region segmentation with the multi-class brain tissue labels to construct a multi-class tissue label map including the lesions includes:

[0024] The lesion labels are fused with seven types of brain tissue labels to construct an eight-type tissue label map including lesions.

[0025] Furthermore, the step of setting multiple conductivity gradients for the lesion area according to tissue status and setting standard conductivity values ​​for other tissues to generate a multi-layered individualized head model includes:

[0026] The mri2mesh module of the SimNIBS platform was called to generate a multi-layer finite element mesh, and the conductivity was assigned in combination with the Fieldtrip toolkit;

[0027] For the lesion area, three conductivity gradients were set according to the tissue state to simulate tissue necrosis, liquefaction or congestion;

[0028] For other tissues, the Fieldtrip recommended standard conductivity values ​​were used;

[0029] Finally, an eight-layer individualized head model with complete structure and reasonable conductivity distribution is generated.

[0030] Furthermore, the steps of constructing an individualized cortical gray matter surface model, performing affine transformation registration between the electrode coordinate system and the individual MRI space to achieve precise alignment, solving a forward problem, and calculating the conduction matrix between the electrodes and the cortical source under the individualized head model to provide a basic physical model include:

[0031] We constructed a personalized cortical gray matter surface model based on Freesurfer and the Human Connectome Project MRI data processing pipeline and simplified it to support source model construction.

[0032] The Fieldtrip template electrode coordinate system is aligned with the individual MRI space through affine transformation and fine-tuned to achieve precise alignment;

[0033] The Fieldtrip-SimBio module is used to solve the forward problem and calculate the conduction matrix between the electrodes and the cortical sources under the individual head model, providing a basic physical model for subsequent source localization or neural regulation simulation.

[0034] Furthermore, the modeling accuracy verification step is also included:

[0035] In the cortical gray matter surface model, dipoles are set as electrical activity sources, and their neighborhood expansion is simulated to construct spatially continuous activation areas, generating corresponding simulated EEG signals.

[0036] The inverse problem of the simulated EEG signals was solved using the reweighted generalized total variation fusion group sparsity constraint algorithm. The model accuracy was evaluated using the dipole localization error and the normalized root mean square error by comparing the source reconstruction results with the original activation areas. The advantages of the multi-layer individualized head model in the source localization task were quantified.

[0037] The second object of the present invention is to provide a neuroelectrophysiological modeling system for stroke patients based on image completion, which applies the above method and includes a lesion area automatic segmentation module, a lesion area structure completion module, a whole-brain multi-tissue fine segmentation module, a multi-layer head model construction module, and an EEG positive problem modeling module; wherein,

[0038] The lesion area automatic segmentation module is used to obtain T1-weighted magnetic resonance imaging data of a stroke patient and perform lesion area segmentation on the T1-weighted magnetic resonance imaging data;

[0039] The lesion region structure completion module is used to complete the image of the lesion region and generate a brain tissue image with continuous structure;

[0040] The whole-brain multi-tissue fine segmentation module is used to perform high-precision structural segmentation of brain tissue in the brain tissue image and identify multiple types of brain tissue;

[0041] The multi-layer head model construction module is used to fuse the lesion labels obtained by lesion area segmentation with multiple types of brain tissue labels to construct a multi-type tissue label map including the lesion. It sets multiple conductivity gradients for the lesion area based on the tissue state and sets standard conductivity values ​​for other tissues to generate a multi-layer personalized head model.

[0042] The EEG forward problem modeling module is used to construct an individualized cortical gray matter surface model, perform affine transformation registration between the electrode coordinate system and the individual MRI space to achieve precise alignment, and solve the forward problem to calculate the conduction matrix between the electrodes and the cortical source under the individualized head model to provide a basic physical model.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] This invention achieves high-precision conversion of stroke patient MRI images into personalized multi-layer head models by constructing an integrated system for stroke lesion identification, structural image completion, multi-tissue fine segmentation, and conductivity parameter modeling. Compared with the traditional five-layer template method, this method not only identifies and repairs lesion areas, improving structural integrity and electromagnetic simulation reliability, but also refines tissue segmentation to construct an eight-layer head model that includes lesions, gray matter, and white matter. It also sets targeted conductivity parameters, significantly improving the accuracy of power source positioning and neural regulation simulation, providing fundamental support for personalized diagnosis and treatment of stroke patients.

[0045] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and to implement it according to the contents of the description, the following preferred embodiments of the present invention are described in detail with reference to the accompanying drawings. The specific implementation methods of the present invention are given in detail by the following embodiments and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0047] Figure 1 Process of neuroelectrophysiological modeling method for stroke patients based on image completion Figure 1 ;

[0048] Figure 2 Process of neuroelectrophysiological modeling method for stroke patients based on image completion Figure 2 ;

[0049] Figure 3 is the original image and the segmentation result;

[0050] Figure 4 Input the image, the restoration result and the original image to the network;

[0051] Figure 5 Construct a flow chart for the eight-layer head model;

[0052] Figure 6 Modeling a flow chart for EEG positive problems;

[0053] Figure 7 Verify the flow chart for modeling accuracy;

[0054] Figure 8 The impact of the inverse problem source location index on the head model for stroke lesion repair;

[0055] Figure 9 Schematic diagram of the neurophysiological modeling system for stroke patients based on image completion;

[0056] Figure 10 A schematic diagram of a computer device;

[0057] Figure 11 A schematic diagram of a computer-readable storage medium. DETAILED DESCRIPTION

[0058] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. It should be noted that, without conflict, the embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0059] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

[0060] The figure numbers in this application are only used to distinguish the various steps in the scheme and are not used to limit the execution order of the various steps. The specific execution order is subject to the description in the specification.

[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0062] Most current mainstream modeling tools construct template head models based on healthy individuals, using fixed tissue segmentation structures and conductivity parameters. These tools lack mechanisms for identifying, repairing, and modeling stroke lesion regions. In particular, in MRI scans of stroke patients, the lesion region suffers from severe structural loss, making it difficult for traditional segmentation methods to accurately label the corresponding tissue types, leading to significant misclassification of gray matter and white matter regions. Furthermore, commonly used tools do not assign independent conductivity parameters to the lesion region, but instead use the default values ​​for healthy tissue, ignoring the unique electrophysiological properties of stroke tissue. This ultimately results in insufficient accuracy and individual adaptability of the entire head model, limiting its application value in stroke neuroelectrical regulation.

[0063] This invention aims to address the challenges of difficult structural abnormality identification, inaccurate conductivity assignment, and crude tissue classification in stroke patient head modeling. By introducing an image completion mechanism and a multi-tissue joint modeling strategy, an integrated neuroelectrophysiological modeling method and system is constructed to support stroke lesion identification, structural repair, and multi-layer conductivity modeling. This method enables the precise generation of individualized multi-layer head models based on brain images, improving the accuracy and clinical adaptability of brain source localization and electrical stimulation simulation. Example 1

[0064] A neurophysiological modeling method for stroke patients based on image completion, such as Figure 1-Figure 2 As shown, the following steps are included:

[0065] S100, acquiring T1-weighted magnetic resonance imaging data of a stroke patient, and performing lesion area segmentation on the T1-weighted magnetic resonance imaging data;

[0066] In some embodiments, the step of segmenting the T1-weighted magnetic resonance imaging data into lesion areas comprises:

[0067] The nn-UNet framework automatically segmented the T1-weighted MRI data into three-dimensional lesion regions. This network, employing an encoder-decoder architecture and incorporating adaptive preprocessing and dynamic network topology, automatically optimized normalization, data augmentation, and hyperparameter settings to address the heterogeneity of stroke lesions in morphology, boundaries, and scale. This improved the accuracy of lesion recognition and output a three-dimensional label map, annotating the lesion's location and extent. Figure 3 (a) is the original MRI image, Figure 3 In (b), the red color is the true lesion mask, and the green color is the network output segmentation result.

[0068] S200, performing image completion on the lesion area to generate a structurally continuous brain tissue image;

[0069] In some embodiments, the step of completing the image of the lesion area includes:

[0070] A pix2pix conditional generative adversarial network (cGAN) was used to complete the image of the lesion region, generating structurally continuous brain tissue images. The model used a U-Net architecture with skip connections as the generator, combined with a PatchGAN local discriminator to improve texture consistency. Given the difficulty in obtaining supervision of true restoration pairs, a training set of "healthy-pseudo-lesion" image pairs was constructed using ATLAS public stroke imaging data. Degraded images were constructed by simulating lesion occlusion masks. Combining the L1 reconstruction loss with a perceptual consistency loss function, the generator was guided to learn the structural priors of healthy tissue, achieving anatomically sound restoration of the lesion region. Figure 4(a) is the network input image, Figure 4 (b) is the repair result. Figure 4 (c) is the original image.

[0071] S300, performing high-precision structural segmentation on the brain tissue in the brain tissue image to identify multiple types of brain tissue;

[0072] In some embodiments, the step of performing high-precision structural segmentation on the brain tissue in the brain tissue image includes:

[0073] Using the completed MRI image as input, the Fieldtrip platform's integrated SPM12 and Freesurfer toolchains perform high-precision brain tissue segmentation, identifying seven tissue types: scalp, skull, cerebrospinal fluid, gray matter, white matter, cerebellar gray matter, and cerebellar white matter. Subsequently, a seven-class nn-UNet segmentation model is constructed based on the segmentation results, enabling rapid and automatic segmentation of subsequent new samples, improving the system's versatility and automation.

[0074] S400, fusing the lesion labels obtained by lesion region segmentation with multiple brain tissue labels to construct a multi-tissue label map including the lesion, setting multiple conductivity gradients for the lesion region based on tissue status, and setting standard conductivity values ​​for other tissues to generate a multi-layer personalized head model;

[0075] In some embodiments, as Figure 5 As shown, the steps of fusing the lesion labels obtained by lesion region segmentation with the multi-class brain tissue labels to construct a multi-class tissue label map including the lesion include:

[0076] S410 , fusing the lesion labels obtained in step S100 with the seven types of brain tissue labels in step S300 to construct an eight-type tissue label map including lesions.

[0077] In some embodiments, the step of setting multiple conductivity gradients for the lesion area according to tissue status and setting standard conductivity values ​​for other tissues to generate a multi-layer personalized head model includes:

[0078] S420, call the mri2mesh module of the SimNIBS platform to generate a multi-layer finite element mesh, and use the Fieldtrip toolkit to assign conductivity;

[0079] S430: For the lesion area, three conductivity gradients (e.g., 0.1 S / m, 0.8087 S / m, 1.71 S / m) were set according to the tissue state to simulate tissue necrosis, liquefaction, or congestion.

[0080] S440, For other tissues, use the Fieldtrip recommended standard conductivity value;

[0081] S450, finally generates an eight-layer individualized head model with complete structure and reasonable conductivity distribution.

[0082] S500: Construct an individualized cortical gray matter surface model, perform affine transformation registration between the electrode coordinate system and the individual MRI space to achieve precise alignment, solve the forward problem, and calculate the conduction matrix between the electrodes and the cortical source under the individualized head model to provide a basic physical model.

[0083] In some embodiments, as Figure 6 As shown, the steps of constructing an individualized cortical gray matter surface model, performing affine transformation registration between the electrode coordinate system and the individual MRI space to achieve precise alignment, solving the forward problem, and calculating the conduction matrix between the electrodes and the cortical source under the individualized head model to provide a basic physical model include:

[0084] S510. Build an individualized cortical gray matter surface model based on Freesurfer and the Human Connectome Project MRI Data Processing Pipeline (HCP pipeline), and simplify it to support source model construction. The Human Connectome Project MRI Data Processing Pipeline provides a standardized neuroimaging analysis toolchain for MRI data preprocessing, cortical reconstruction, and feature extraction. S520. Align the Fieldtrip template electrode coordinate system with the individual MRI space using an affine transformation. For example, perform a simulated transformation and registration of the template electrode coordinate system file containing 64 standard electrode position information in the Fieldtrip toolbox with the individual MRI space, and manually fine-tune it for precise alignment.

[0085] S530. Use the Fieldtrip-SimBio module to solve the forward problem and calculate the conduction matrix between the electrodes and the cortical source under the individual head model, providing a basic physical model for subsequent source localization or neural regulation simulation.

[0086] In order to verify the applicability of the constructed head model in brain power source localization, simulation experiments are used to evaluate the accuracy. Figure 7 As shown, it also includes the modeling accuracy verification step:

[0087] S600, setting a dipole as an electrical activity source in the cortical gray matter surface model, and simulating its neighborhood expansion to construct a spatially continuous activation area, and generating a corresponding simulated EEG signal;

[0088] S700, using the reweighted generalized total variation fusion group sparse constraint algorithm to solve the inverse problem of the simulated EEG signal, by comparing the source reconstruction results with the original activation area, using indicators such as dipole localization error (DLE) and normalized root mean square error (NRMSE) to evaluate the model accuracy, quantifying the advantages of the multi-layer individualized head model in the source localization task. Figure 8 It can be seen that the eight-layer finite element model of the repaired lesion has a significant advantage in the source localization task.

[0089] This example provides a neuroelectrophysiological modeling method for stroke patients. This method integrates stroke lesion identification, image structure completion, fine brain tissue segmentation, and multi-layer head model construction to form an end-to-end personalized electrophysiological modeling process. This method fully accounts for tissue structural abnormalities and conductivity differences in stroke patients, improving the accuracy of source localization and neural control simulation, providing reliable support for clinical precision treatment. Example 2

[0090] A neurophysiological modeling system for stroke patients based on image completion, using the above method, for a detailed description of the method, please refer to the corresponding description in the above method embodiment, which will not be repeated here. Figure 9 As shown, the system 800 includes a lesion area automatic segmentation module 810, a lesion area structure completion module 820, a whole brain multi-tissue fine segmentation module 830, a multi-layer head model construction module 840, and an EEG positive problem modeling module 850; wherein,

[0091] The lesion area automatic segmentation module is used to obtain T1-weighted magnetic resonance imaging data of a stroke patient and perform lesion area segmentation on the T1-weighted magnetic resonance imaging data;

[0092] The lesion region structure completion module is used to complete the image of the lesion region and generate a brain tissue image with continuous structure;

[0093] The whole-brain multi-tissue fine segmentation module is used to perform high-precision structural segmentation of brain tissue in the brain tissue image and identify multiple types of brain tissue;

[0094] The multi-layer head model construction module is used to fuse the lesion labels obtained by lesion area segmentation with multiple types of brain tissue labels to construct a multi-type tissue label map including the lesion. It sets multiple conductivity gradients for the lesion area based on the tissue state and sets standard conductivity values ​​for other tissues to generate a multi-layer personalized head model.

[0095] The EEG forward problem modeling module is used to construct an individualized cortical gray matter surface model, perform affine transformation registration between the electrode coordinate system and the individual MRI space to achieve precise alignment, and solve the forward problem to calculate the conduction matrix between the electrodes and the cortical source under the individualized head model to provide a basic physical model.

[0096] Based on the technical solution of the above embodiment, optionally, the step of segmenting the T1-weighted magnetic resonance imaging data into lesion areas includes:

[0097] The T1-weighted magnetic resonance imaging data is segmented into three-dimensional lesion areas based on the nn-UNet framework, and a three-dimensional label map is output to mark the location and range of the lesion.

[0098] Based on the technical solution of the above embodiment, optionally, the step of completing the image of the lesion area includes:

[0099] Based on the pix2pix conditional generative adversarial network, the image of the lesion area is completed to generate a structurally continuous brain tissue image.

[0100] Based on the technical solutions of the above embodiments, optionally, the adversarial network adopts a U-Net structure with jump connections as a generator, and cooperates with the PatchGAN local discriminator to improve texture consistency, uses ATLAS public stroke image data to construct a training set of healthy-pseudo lesion image pairs, constructs a degraded image by simulating lesion occlusion masks, and combines the L1 reconstruction loss and the perceptual consistency loss function to guide the generator to learn the structural prior of healthy tissue, thereby achieving reasonable anatomical repair of the lesion area.

[0101] Based on the technical solution of the above embodiment, optionally, the step of performing high-precision structural segmentation on the brain tissue in the brain tissue image includes:

[0102] The brain tissue image after image completion is used as input, and the SPM12 and Freesurfer tool chains integrated in the Fieldtrip platform are used to perform high-precision structural segmentation of the brain tissue.

[0103] Based on the technical solutions of the above embodiments, optionally, the multiple types of brain tissues identified include seven types of tissues: scalp, skull, cerebrospinal fluid, gray matter, white matter, cerebellar gray matter, and cerebellar white matter.

[0104] Based on the technical solution of the above embodiment, optionally, after the step of performing high-precision structural segmentation on the brain tissue in the brain tissue image and identifying multiple types of brain tissue, the step further includes:

[0105] Based on the structural segmentation results, a seven-category nn-UNet segmentation model is constructed to achieve fast and automatic segmentation of new samples.

[0106] Based on the technical solution of the above embodiment, optionally, the step of fusing the lesion label obtained by lesion region segmentation with multiple types of brain tissue labels to construct a multi-type tissue label map including the lesion includes:

[0107] The lesion labels are fused with seven types of brain tissue labels to construct an eight-type tissue label map including lesions.

[0108] Based on the technical solution of the above embodiment, optionally, the step of setting multiple conductivity gradients for the lesion area according to the tissue state and setting standard conductivity values ​​for other tissues to generate a multi-layer personalized head model includes:

[0109] The mri2mesh module of the SimNIBS platform was called to generate a multi-layer finite element mesh, and the conductivity was assigned in combination with the Fieldtrip toolkit;

[0110] For the lesion area, three conductivity gradients were set according to the tissue state to simulate tissue necrosis, liquefaction or congestion;

[0111] For other tissues, the Fieldtrip recommended standard conductivity values ​​were used;

[0112] Finally, an eight-layer individualized head model with complete structure and reasonable conductivity distribution is generated.

[0113] Based on the technical solution of the above embodiment, optionally, the steps of constructing an individualized cortical gray matter surface model, performing affine transformation registration between the electrode coordinate system and the individual MRI space to achieve precise alignment, solving the forward problem, and calculating the conduction matrix between the electrodes and the cortical source under the individualized head model to provide a basic physical model include:

[0114] We constructed a personalized cortical gray matter surface model based on Freesurfer and the Human Connectome Project MRI data processing pipeline and simplified it to support source model construction.

[0115] The Fieldtrip template electrode coordinate system is aligned with the individual MRI space through affine transformation and fine-tuned to achieve precise alignment;

[0116] The Fieldtrip-SimBio module is used to solve the forward problem and calculate the conduction matrix between the electrodes and the cortical sources under the individual head model, providing a basic physical model for subsequent source localization or neural regulation simulation.

[0117] Based on the technical solution of the above embodiment, optionally, a modeling accuracy verification step is further included:

[0118] In the cortical gray matter surface model, dipoles are set as electrical activity sources, and their neighborhood expansion is simulated to construct spatially continuous activation areas, generating corresponding simulated EEG signals.

[0119] The inverse problem of the simulated EEG signals was solved using the reweighted generalized total variation fusion group sparsity constraint algorithm. The model accuracy was evaluated using the dipole localization error and the normalized root mean square error by comparing the source reconstruction results with the original activation areas. The advantages of the multi-layer individualized head model in the source localization task were quantified.

[0120] This embodiment provides a neuroelectrophysiological modeling system for stroke patients. This system integrates stroke lesion identification, image structure completion, fine brain tissue segmentation, and multi-layer head model construction, forming an end-to-end personalized electrophysiological modeling process. This method fully accounts for tissue structural abnormalities and conductivity differences in stroke patients, improving the accuracy of source localization and neural regulation simulation, and providing reliable support for clinical precision treatment. Example 3

[0121] A computer device 900, such as Figure 10 As shown, the present invention includes a memory 910, a processor 920, and a computer program 930 stored in the memory and executable on the processor. When the processor executes the computer program, the steps of a method for neuroelectrophysiological modeling of stroke patients based on image completion are implemented. For a detailed description of the method, please refer to the corresponding description in the above method embodiment and will not be repeated here. Example 4

[0122] A computer-readable storage medium such as Figure 11 As shown, a computer program is stored thereon, and when the computer program is executed by the processor, the steps of a method for neuroelectrophysiological modeling of stroke patients based on image completion are implemented. For a detailed description of the method, reference can be made to the corresponding description in the above method embodiment, and no further details will be given here.

[0123] The number of devices and processing scales described herein are intended to simplify the description of the present invention. Applications, modifications, and variations of the present invention will be readily apparent to those skilled in the art.

[0124] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

[0125] The apparatus, computer device, non-volatile computer storage medium, and method provided in the embodiments of this specification correspond to each other. Therefore, the apparatus, computer device, and non-volatile computer storage medium also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding apparatus, computer device, and non-volatile computer storage medium will not be repeated here.

[0126] Those skilled in the art will also appreciate that, in addition to implementing the controller in pure computer-readable program code, it is entirely possible to implement the same functionality by programming the method steps logically, such as through logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be considered structures within the hardware component. Alternatively, the devices for implementing various functions can be considered both software units implementing the method and structures within the hardware component.

[0127] The systems, devices, or units described in the above embodiments can be implemented by computer chips or physical devices, or by products with certain functions. For ease of description, the above devices are described separately by function, with each unit described separately. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware components.

[0128] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Thus, the embodiments of this specification may take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware. Furthermore, the embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. 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.

[0130] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 The function specified in one or more boxes.

[0131] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0132] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0133] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program units. Generally, program units include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communications network. In a distributed computing environment, program units may be located in local and remote computer storage media, including storage devices.

[0134] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

[0135] The foregoing is merely an example of the present invention and is not intended to limit the present invention to one or more embodiments. It will be apparent to those skilled in the art that various modifications and variations may be made to the present invention to one or more embodiments. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention to one or more embodiments shall be included within the scope of the claims of the present invention to one or more embodiments.

Claims

1. A neuroelectrophysiological modeling method for stroke patients based on image completion, characterized in that: The following steps are involved: Acquiring T1-weighted magnetic resonance imaging data of a stroke patient, and performing lesion area segmentation on the T1-weighted magnetic resonance imaging data; Performing image completion on the lesion area to generate a structurally continuous brain tissue image; performing high-precision structural segmentation on the brain tissue in the brain tissue image to identify multiple types of brain tissue; wherein the multiple types of brain tissue identified include seven types of tissue: scalp, skull, cerebrospinal fluid, gray matter, white matter, cerebellar gray matter, and cerebellar white matter; The lesion labels obtained from lesion region segmentation are integrated with multiple brain tissue labels to construct a multi-tissue label map including the lesion. Multiple conductivity gradients are set for the lesion region based on tissue status, while standard conductivity values ​​are set for other tissues to generate a multi-layered personalized head model. An individualized cortical gray matter surface model is constructed, and the electrode coordinate system is aligned with the individual MRI space through affine transformation to achieve precise alignment. The forward problem is solved to calculate the conduction matrix between the electrodes and the cortical source under the individualized head model to provide a basic physical model.

2. The method for neuroelectrophysiological modeling of stroke patients based on image completion according to claim 1, characterized in that: The step of segmenting the T1-weighted magnetic resonance imaging data into lesion areas comprises: The T1-weighted magnetic resonance imaging data is segmented into three-dimensional lesion areas based on the nn-UNet framework, and a three-dimensional label map is output to mark the location and range of the lesion.

3. The method for neuroelectrophysiological modeling of stroke patients based on image completion according to claim 1, characterized in that: The step of completing the image of the lesion area includes: Based on the pix2pix conditional generative adversarial network, the image of the lesion area is completed to generate a structurally continuous brain tissue image.

4. The method for neuroelectrophysiological modeling of stroke patients based on image completion according to claim 3, characterized in that: The adversarial network adopts a U-Net structure with skip connections as the generator, and cooperates with the PatchGAN local discriminator to improve texture consistency. The ATLAS public stroke image data is used to construct a training set of healthy-pseudo-lesion image pairs. The degraded image is constructed by simulating lesion occlusion masks. The L1 reconstruction loss is combined with the perceptual consistency loss function to guide the generator to learn the structural prior of healthy tissue and achieve reasonable anatomical repair of the lesion area.

5. The method for neuroelectrophysiological modeling of stroke patients based on image completion according to claim 1, characterized in that: The step of performing high-precision structural segmentation on the brain tissue in the brain tissue image comprises: The brain tissue image after image completion is used as input, and the SPM12 and Freesurfer tool chains integrated in the Fieldtrip platform are used to perform high-precision structural segmentation of the brain tissue.

6. The method for neuroelectrophysiological modeling of stroke patients based on image completion according to claim 5, characterized in that: After the step of performing high-precision structural segmentation of the brain tissue in the brain tissue image and identifying multiple types of brain tissue, the following step further comprises: Based on the structural segmentation results, a seven-category nn-UNet segmentation model is constructed to achieve fast and automatic segmentation of new samples.

7. The method for neuroelectrophysiological modeling of stroke patients based on image completion according to claim 5, characterized in that: The steps of fusing the lesion labels obtained by lesion region segmentation with the multi-class brain tissue labels to construct a multi-class tissue label map including the lesions include: The lesion labels are fused with seven types of brain tissue labels to construct an eight-type tissue label map including lesions.

8. The method for neuroelectrophysiological modeling of stroke patients based on image completion according to claim 7, characterized in that: The steps of setting multiple conductivity gradients for the lesion area according to the tissue state and setting standard conductivity values ​​for other tissues to generate a multi-layered individualized head model include: The mri2mesh module of the SimNIBS platform was called to generate a multi-layer finite element mesh, and the conductivity was assigned in combination with the Fieldtrip toolkit; For the lesion area, three conductivity gradients were set according to the tissue state to simulate tissue necrosis, liquefaction or congestion; For other tissues, the Fieldtrip recommended standard conductivity values ​​were used; Finally, an eight-layer individualized head model with complete structure and reasonable conductivity distribution is generated.

9. The method for neuroelectrophysiological modeling of stroke patients based on image completion according to claim 1, characterized in that: The steps of constructing an individualized cortical gray matter surface model, performing affine transformation registration between the electrode coordinate system and the individual MRI space to achieve precise alignment, solving a forward problem, and calculating the conduction matrix between the electrodes and the cortical source under the individualized head model to provide a basic physical model include: We constructed a personalized cortical gray matter surface model based on Freesurfer and the Human Connectome Project MRI data processing pipeline and simplified it to support source model construction. The Fieldtrip template electrode coordinate system is aligned with the individual MRI space through affine transformation and fine-tuned to achieve precise alignment; The Fieldtrip-SimBio module is used to solve the forward problem and calculate the conduction matrix between the electrodes and the cortical sources under the individual head model, providing a basic physical model for subsequent source localization or neural regulation simulation.

10. The method for neuroelectrophysiological modeling of stroke patients based on image completion according to claim 1, characterized in that: It also includes the modeling accuracy verification step: In the cortical gray matter surface model, dipoles are set as electrical activity sources, and their neighborhood expansion is simulated to construct spatially continuous activation areas, generating corresponding simulated EEG signals. The inverse problem of the simulated EEG signals was solved using the reweighted generalized total variation fusion group sparsity constraint algorithm. The model accuracy was evaluated using the dipole localization error and the normalized root mean square error by comparing the source reconstruction results with the original activation areas. The advantages of the multi-layer individualized head model in the source localization task were quantified.

11. A neuroelectrophysiological modeling system for stroke patients based on image completion, applying the method according to any one of claims 1 to 10, characterized in that: It includes the lesion area automatic segmentation module, the lesion area structure completion module, the whole brain multi-tissue fine segmentation module, the multi-layer head model construction module, and the EEG positive problem modeling module; among them, The lesion area automatic segmentation module is used to obtain T1-weighted magnetic resonance imaging data of a stroke patient and perform lesion area segmentation on the T1-weighted magnetic resonance imaging data; The lesion region structure completion module is used to complete the image of the lesion region and generate a brain tissue image with continuous structure; The whole-brain multi-tissue fine segmentation module is used to perform high-precision structural segmentation of brain tissue in the brain tissue image and identify multiple types of brain tissue; wherein the multiple types of brain tissue identified include seven types of tissue: scalp, skull, cerebrospinal fluid, gray matter, white matter, cerebellar gray matter, and cerebellar white matter; The multi-layer head model construction module is used to fuse the lesion labels obtained by lesion area segmentation with multiple types of brain tissue labels to construct a multi-type tissue label map including the lesion. It sets multiple conductivity gradients for the lesion area based on the tissue state and sets standard conductivity values ​​for other tissues to generate a multi-layer personalized head model. The EEG forward problem modeling module is used to construct an individualized cortical gray matter surface model, perform affine transformation registration between the electrode coordinate system and the individual MRI space to achieve precise alignment, and solve the forward problem to calculate the conduction matrix between the electrodes and the cortical source under the individualized head model to provide a basic physical model.

Images