Brain connection pattern detection method, device and equipment based on cognitive function defects

Abstract

The present application relates to the technical field of cognitive deficit assessment, and in particular to a method, device and equipment for detecting brain connection patterns based on cognitive function defects. First, the present application constructs a prediction model, which uses the individual's whole-brain structural and functional connection group data to obtain the brain weight network of the cognitive process through a machine learning algorithm, and then uses a large-scale multivariate edge-by-edge analysis method to compare the abnormal brain connection group patterns between different diagnostic groups. By comparing the abnormal connection group characteristics of each diagnostic group obtained by the large-scale multivariate edge-by-edge analysis method with the cognition-related connection group characteristics, it can be identified whether these abnormalities involve cognition-related core brain areas, thereby determining whether the abnormalities are related to cognitive defects. By comparing the degree of sharing and differences in abnormal connection group patterns between different diagnostic groups, it can be determined whether the cross-diagnostic cognitive defects are derived from shared macroscopic brain network abnormalities or from the neuropathological basis specific to each diagnosis.

Description

Technical Field

[0001] The present application relates to the technical field of cognitive deficit assessment, and in particular to a method, device and equipment for detecting brain connection patterns based on cognitive function deficits. Background Art

[0002] In recent years, neuropsychological research has revealed that patients with different mental illnesses have widespread commonalities in impaired cognitive function. Cognitive deficits are common in many mental illnesses and are closely related to the patient's psychopathological symptoms.

[0003] Studies have found that patients with mental illnesses such as schizophrenia, bipolar disorder, and attention deficit hyperactivity disorder perform significantly worse than healthy controls in various cognitive tests, revealing the prevalence of impaired cognitive function in various mental illnesses. Although the above studies have revealed the commonality of cognitive impairment in patients with mental illness, it is currently unclear whether these cognitive deficits are due to shared neurobiological factors across diagnoses or to unique pathological processes in specific diseases.

[0004] Therefore, how to determine whether the cognitive deficits of patients with mental illness are due to shared neurobiological factors across diagnoses or to unique pathological processes of a specific disease has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The present application provides a method, device and equipment for detecting brain connection patterns based on cognitive function defects, which can determine the factors causing cognitive defects in patients with mental illness.

[0006] In the first aspect, an embodiment of the present application provides a method for detecting brain connection patterns based on cognitive function defects, the method comprising: acquiring neuroimaging data, the neuroimaging data comprising functional magnetic resonance imaging data and diffusion weighted imaging data of a healthy control group and different categories of mental illnesses; preprocessing the functional magnetic resonance imaging data and the diffusion weighted imaging data, respectively, to obtain preprocessed fMRI data and preprocessed DWI data; constructing a functional connection group and a structural connection group based on the preprocessed fMRI data and the preprocessed DWI data; determining a brain weight network of a cognitive process based on the functional connection group, the structural connection group and a prediction model based on a multimodal connection group that has been constructed, the brain weight network of the cognitive process being used to characterize the importance weights between different brain regions and cognitive functions; determining abnormal connection group characteristics based on the functional connection group, the structural connection group and a large-scale multivariate edge-by-edge analysis method; and realizing brain connection pattern detection of cognitive function defects across diagnostic mental disorders based on the brain weight network of the cognitive process and the abnormal connection group characteristics.

[0007] In a possible implementation, the functional magnetic resonance imaging data is preprocessed to obtain preprocessed fMRI data, including: format conversion of the functional magnetic resonance imaging data to determine NIfTI format data corresponding to the functional magnetic resonance imaging data; based on the NIfTI format data corresponding to the functional magnetic resonance imaging data, brain extraction processing is performed to determine a brain image to be processed; based on the brain image to be processed, head motion correction is performed to determine a brain spatial registration image; based on the brain spatial registration image and the time corresponding to the brain spatial registration image, the brain is determined through time layer correction. The invention relates to a method for determining a brain coordinate image by performing a co-registration process on the brain smoothed image, wherein the brain regions in the brain coordinate image correspond to a unified standard space; based on the brain coordinate image, by performing a percentage signal change conversion, converting the signal intensity of the brain coordinate image into a percentage signal change value, and determining a de-differenced image; based on the de-differenced image, by performing a Nuisance regression, determining a denoised image, wherein the denoised image includes the pre-processed fMRI data.

[0008] In a possible implementation, the diffusion weighted imaging data is preprocessed to obtain preprocessed DWI data, including: performing format conversion on the diffusion weighted imaging data to determine NIfTI format data corresponding to the diffusion weighted imaging data; determining an eddy current corrected image through eddy current distortion correction based on the NIfTI format data corresponding to the diffusion weighted imaging data; performing head motion correction based on the eddy current corrected image to determine a spatial registration image of the eddy current corrected image; registering the spatial registration image of the eddy current corrected image to a preset spatial template to determine a standardized image; performing brain extraction processing based on the standardized image to determine the diffusion weighted imaging data to be processed; determining anisotropic diffusion data based on the diffusion weighted imaging data to be processed through diffusion tensor calculation; and determining fiber bundle reconstruction data based on the anisotropic diffusion data through fiber bundle reconstruction, wherein the fiber bundle reconstruction data includes the preprocessed DWI data.

[0009] In a possible implementation, the functional connection group is determined based on the preprocessed fMRI data, including: extracting time series of different brain regions based on the preprocessed fMRI data and the Shen268 atlas; calculating the Pearson correlation coefficient between each of the time series; constructing a functional connection matrix based on the Pearson correlation coefficient between each of the time series; determining a tangent space projection matrix based on the functional connection matrix and the Euclidean average of the functional connection matrix; and determining the functional connection group based on the tangent space projection matrix.

[0010] In a possible implementation, a structural connection group is determined based on the preprocessed DWI data, including: obtaining a fiber orientation distribution image based on the preprocessed DWI data, AAL116 atlas and Fournier algorithm; based on the fiber orientation distribution image and iFOD2 algorithm, using segmented images of three tissue types as labels of the fiber tracking area to perform fiber tracking to obtain fiber tracking data, the tissue types including white matter, gray matter and cerebrospinal fluid; based on the fiber tracking data and a preset fiber length threshold, a target fiber bundle is determined, and the preset fiber length threshold is 25-250mm; through a preset tool, the average weight and linear combination coefficient of the target fiber bundle are calculated; based on the average weight and linear combination coefficient of the target fiber bundle, important fiber bundles are determined; based on the important fiber bundles, a streamline number matrix between different nodes of the important fiber bundle is constructed; based on the streamline number matrix, overfitting is offset by a preset algorithm to determine the structural connection group.

[0011] In one possible implementation, the method for constructing the prediction model based on the multimodal connection group includes: extracting feature information from the functional connection group and the structural connection group, the feature information including the connection strength of the positive network, the negative network and the whole brain network; training the prediction model based on the feature information using a ridge regression algorithm; cross-validating the prediction model using a cross-validation formula; and permutation validating the cross-validated training model to determine the prediction model based on the multimodal connection group.

[0012] In a possible implementation, the abnormal connection group characteristics are determined based on the functional connection group, the structural connection group and the large-scale multivariate edge-by-edge analysis method, including: performing data fusion on the functional connection group and the structural connection group to obtain a connection group matrix; performing large-scale multivariate edge-by-edge analysis on each connection in the connection group matrix to obtain an F statistic, and the F statistic is used to characterize the degree of difference of each connection between different diagnostic groups; based on the F statistic, determining an F value matrix; performing multiple comparison correction on the F value matrix to determine the abnormal connection group characteristics.

[0013] In a possible implementation, the brain weight network of the cognitive process and the abnormal connection group characteristics are used to detect brain connection patterns of cognitive function defects across diagnosis mental disorders, including: calculating the contribution of each node in the brain weight network of the cognitive process and the contribution of the network based on the node contribution degree calculation formula; calculating the contribution of each node in the abnormal connection group characteristics and the contribution of the network; based on the contribution of each node in the brain weight network of the cognitive process and the contribution of the network, and the contribution of each node in the abnormal connection group characteristics and the contribution of the network, the brain connection pattern detection of cognitive function defects across diagnosis mental disorders is realized; wherein, the node contribution degree calculation formula includes:

[0014]

[0015] Among them, W n represents the contribution of the nth node in the brain weight network of cognitive processes, B(k, m) represents whether the kth edge is selected in the mth connection group, std(Ek(:, m)) represents the standard deviation of the kth edge in the mth connection group, βm k Represents the weight of the kth connection group in the mth connection group.

[0016] In a second aspect, an embodiment of the present application provides a brain connection pattern detection device based on cognitive function defects, the device comprising: an acquisition module for acquiring neuroimaging data, the neuroimaging data comprising functional magnetic resonance imaging data and diffusion weighted imaging data of a healthy control group and different types of mental illnesses; a preprocessing module for preprocessing the functional magnetic resonance imaging data and the diffusion weighted imaging data, respectively, to obtain preprocessed fMRI data and preprocessed DWI data; a model building module for building a functional connection group and a structure based on the preprocessed fMRI data and the preprocessed DWI data. The model building module is further used to determine the brain weight network of the cognitive process based on the functional connection group, the structural connection group and the constructed prediction model based on the multimodal connection group, and the brain weight network of the cognitive process is used to characterize the importance weights between different brain regions and cognitive functions; the model building module is further used to determine the abnormal connection group characteristics based on the functional connection group, the structural connection group and the large-scale multivariate edge-by-edge analysis method; the detection module is used to realize the brain connection pattern detection of cognitive function defects in cross-diagnosis mental disorders based on the brain weight network of the cognitive process and the abnormal connection group characteristics.

[0017] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method described in the first aspect or any one of the implementation methods thereof is implemented.

[0018] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect or any one of the implementation methods thereof is implemented.

[0019] This application converts neuroimaging data into functional connectomes and structural connectomes, and determines the brain weight network of cognitive processes based on functional connectomes and structural connectomes and the constructed prediction model based on multimodal connectomes; based on functional connectomes and structural connectomes and large-scale multivariate edge-by-edge analysis, abnormal connectome features are determined; then based on the brain weight network of cognitive processes and the node and network contributions in the abnormal connectome features, the shared neural mechanisms and unique pathological processes of cognitive function defects in transdiagnostic mental disorders are identified. Specifically, by comparing the abnormal connectome features of each diagnostic group obtained by large-scale multivariate edge-by-edge analysis with the cognitive-related weighted brain network, it is possible to identify whether these abnormalities involve cognitive-related core brain areas, thereby determining whether the abnormalities are related to cognitive defects. By comparing the degree of sharing and differences in abnormal connectome patterns between different diagnostic groups, it can be determined whether the transdiagnostic cognitive defects are derived from shared macroscopic brain network abnormalities or from the neuropathological basis specific to each diagnosis. The method described in this application can further provide clues for discovering new diagnostic indicators for mental illness, and can also evaluate specific network abnormalities in individual patients and systematically explore the neural basis of cognitive function defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 A schematic diagram of a method for detecting brain connection patterns based on cognitive function defects provided in this application;

[0022] Figure 2 A schematic diagram of determining functional connectomes and structural connectomes based on preprocessed fMRI data and preprocessed DWI data provided in this application;

[0023] Figure 3A schematic diagram of abnormal patterns of multimodal connectomes between different diagnostic categories using large-scale multivariate edge-by-edge analysis based on functional connectomes and structural connectomes provided for this application;

[0024] Figure 4 A structural block diagram of a brain connection pattern detection device based on cognitive function defects provided in this application;

[0025] Figure 5 A schematic diagram of the structure of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0026] The following will be combined with the implementation methods of this application to clearly and completely describe the technical solutions in the implementation methods of this application. Obviously, the described implementation methods are only part of the implementation methods of this application, not all of the implementation methods. Based on the implementation methods in this application, all other implementation methods obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0027] In recent years, neuropsychological research has revealed that patients with different mental illnesses have widespread commonalities in terms of impaired cognitive function. Cognitive function refers to an individual's comprehensive ability to acquire, store, process and use information, including attention, memory, language, visual space and other fields. A large amount of evidence shows that cognitive function deficits are common in many mental illnesses and are closely related to the patient's psychopathological symptoms.

[0028] There are specific defects in the cognitive function performance of patients with different types of mental disorders. Previous studies have found that patients with schizophrenia, bipolar disorder, attention deficit hyperactivity disorder, etc. perform significantly lower than healthy controls in various cognitive tests, which reveals the prevalence of impaired cognitive function in various mental illnesses.

[0029] Although the above studies have revealed the commonalities of cognitive impairment in patients with mental illness, studies on a single disease are still limited in clarifying its neural mechanisms. Few studies have directly compared the differences in brain network abnormalities between different mental illnesses, making it difficult to understand the shared characteristics and individual differences in the neural basis of these diseases. Limited cross-sectional comparative studies have shown that cognitive deficits are associated with abnormal activity patterns of brain networks associated with specific mental illnesses. However, it is not yet certain whether these cognitive deficits are due to shared neurobiological factors across diagnoses or to unique pathological processes in specific diseases.

[0030] In recent years, the development of multimodal connectomics has provided new methods and perspectives for studying brain network dysfunction. Through brain network pattern analysis, we can better understand the shared neural representation of cognitive dysfunction and hopefully reveal the common pathophysiological basis of different mental illnesses. However, its application in transdiagnostic analysis of mental illnesses still needs further study.

[0031] To address the above issues, it is necessary to explore the differences in brain network abnormalities between different mental illnesses, identify the shared characteristics and individual differences in the neural basis of these diseases, and further clarify whether these cognitive deficits are caused by shared neurobiological factors across diagnoses or by unique pathological processes of specific diseases.

[0032] Therefore, the present application proposes a method, device and equipment for detecting brain connection patterns based on cognitive function defects to solve this problem.

[0033] Figure 1 A flowchart of a method for detecting a brain connection pattern based on cognitive function defects is provided in one embodiment of the present application. The method comprises:

[0034] S110, acquiring neuroimaging data, the neuroimaging data including functional magnetic resonance imaging data and diffusion weighted imaging data of a healthy control group and different types of mental illnesses.

[0035] Specifically, the neuroimaging data obtained in the present application are functional magnetic resonance imaging (fMRI) data and diffusion weighted imaging (DWI) data of a healthy control group and different types of mental illnesses with typical characteristics selected from a large number of data.

[0036] S120, preprocessing the functional magnetic resonance imaging data and the diffusion weighted imaging data respectively to obtain preprocessed fMRI data and preprocessed DWI data.

[0037] In a possible implementation, preprocessing the functional magnetic resonance imaging data to obtain preprocessed fMRI data includes: performing format conversion on the functional magnetic resonance imaging data to determine NIfTI format data corresponding to the functional magnetic resonance imaging data; performing brain extraction processing based on the NIfTI format data corresponding to the functional magnetic resonance imaging data to determine the brain image to be processed; performing head motion correction based on the brain image to be processed to determine the brain space registration image; based on the brain space registration image and the time corresponding to the brain space registration image, determining the brain time image through time layer correction, and the brain time image corresponds to the same reference time point; based on the brain time image, determining the brain smoothed image through spatial smoothing processing; determining the brain coordinate image through co-registration processing of the brain smoothed image, and the brain area in the brain coordinate image corresponds to a unified standard space; based on the brain coordinate image, converting the signal intensity of the brain coordinate image into a percentage signal change value through percentage signal change conversion to determine a de-differenced image; based on the de-differenced image, determining a denoised image through Nuisance regression, and the denoised image includes the preprocessed fMRI data.

[0038] Specifically, the fMRI data is preprocessed to obtain preprocessed fMRI data, including:

[0039] 1. Convert fMRI data in Digital Imaging and Communications in Medicine (DICOM) format to the unified format of Neuroimaging Informatics Technology Initiative (NIfTI) to facilitate subsequent analysis and data processing;

[0040] 2. Perform brain extraction on the converted NIfTI format fMRI data: remove the skull and other non-brain tissues from the MRI image to more accurately analyze the brain structure and function, that is, determine the brain image to be processed;

[0041] 3. Perform head motion correction based on the brain image to be processed. Use 6 rigid body motion parameters (3 translations and 3 rotations) to correct the displacement and rotation of the time series images caused by the subject's head movement, so that the image sequence can achieve the best spatial registration, that is, determine the brain spatial registration image;

[0042] 4. For the brain spatial registration image and the time corresponding to the brain spatial registration image, the brain time image is determined through time layer correction. The brain time image corresponds to the same reference time point. Since the acquisition of fMRI images is completed by continuously collecting multiple brain slices, there are differences in the collection time of different slices. This time difference needs to be corrected so that all slices correspond to the same reference time point;

[0043] 5. Based on the brain time image, the brain smoothed image is determined through spatial smoothing. Specifically, the image is spatially smoothed using a Gaussian kernel (FWHM = 6 mm), which can increase the signal-to-noise ratio and compensate for the differences between individual brain anatomical structures, which is used to improve the sensitivity of subsequent statistical analysis;

[0044] 6. Align the fMRI data of all individuals' brains through co-registration of the smoothed brain images, and register all the brain images of individuals to a unified standard space. Specifically, register each individual's smoothed brain image to the standard brain space, and use linear transformation to ensure that the brain regions of different subjects can correspond to the same standard space, which is convenient for subsequent group analysis. In this implementation, the standard brain space is the MNI152 brain template;

[0045] 7. Based on the brain coordinate image, the signal intensity of the brain coordinate image is converted into a percentage signal change value through percentage signal change conversion to reduce the influence of anatomical differences between individuals and determine the de-differentiation image.

[0046] 8. Based on the de-differenced image, Nuisance regression is used to identify and regress components related to irrelevant factors such as head movement and physiological noise using independent component analysis and other methods to improve the purity of the signal and determine the denoised image, which includes the preprocessed fMRI data.

[0047] In a possible implementation, diffusion weighted imaging data is preprocessed to obtain preprocessed DWI data, including: performing format conversion on the diffusion weighted imaging data to determine NIfTI format data corresponding to the diffusion weighted imaging data; determining an eddy current corrected image through eddy current distortion correction based on the NIfTI format data corresponding to the diffusion weighted imaging data; performing head motion correction based on the eddy current corrected image to determine a spatial registration image of the eddy current corrected image; registering the spatial registration image of the eddy current corrected image to a preset spatial template to determine a standardized image; performing brain extraction processing based on the standardized image to determine the diffusion weighted imaging data to be processed; determining anisotropic diffusion data through diffusion tensor calculation based on the diffusion weighted imaging data to be processed; and determining fiber bundle reconstruction data through fiber bundle reconstruction based on the anisotropic diffusion data, wherein the fiber bundle reconstruction data includes the preprocessed DWI data.

[0048] Specifically, the DWI data is preprocessed to obtain preprocessed DWI data, including:

[0049] 1. Convert DWI data in Digital Imaging and Communications in Medicine (DICOM) format to NIfTI format to facilitate subsequent analysis and data processing;

[0050] 2. Based on the NIfTI format data corresponding to the diffusion-weighted imaging data, the eddy current distortion correction is used to determine the eddy current corrected image. During the diffusion-weighted imaging process, the rapid switching of the diffusion gradient field will cause geometric distortion of the image. The inverse matrix correction method is usually used to estimate the distortion field in each diffusion gradient direction and correct the original data to reduce or eliminate this geometric distortion and finally determine the eddy current corrected image;

[0051] 3. Based on the eddy current correction image, head motion correction is performed to determine the spatial registration image of the eddy current correction image. Specifically, each voxel of the eddy current correction image is registered to the reference image (b=0) through rigid matrix transformation (translation and rotation) so that the entire data set achieves optimal spatial alignment;

[0052] 4. Register the spatial registration image of the eddy current corrected image to the preset spatial template to determine the standardized image. Specifically, register the spatial registration image of the individual eddy current corrected image to the standard brain space template (MNI template) using nonlinear transformation. Achieve spatial standardization between different subjects and prepare for subsequent group analysis;

[0053] 5. Based on the standardized image, perform brain extraction processing to determine the diffusion-weighted imaging data to be processed. Specifically, extract the complete brain tissue from the standardized image and remove non-brain tissues such as the skull, eyeball, and scalp;

[0054] 6. Based on the diffusion weighted imaging data to be processed, the anisotropic diffusion data is determined by diffusion tensor calculation. Specifically, for each voxel in the diffusion weighted imaging data, the diffusion tensor model of the voxel is fitted by minimizing the error using weighted data in different diffusion gradient directions; the diffusion tensor can describe the anisotropic diffusion of water molecules in the voxel;

[0055] 7. Based on the anisotropic diffusion data, fiber bundle reconstruction data is determined through fiber bundle reconstruction. The fiber bundle reconstruction data includes the preprocessed DWI data. Specifically, based on the anisotropic diffusion data, the main diffusion direction is determined and integrated to reconstruct the three-dimensional trajectory and structural connection pattern of the brain white matter fiber bundle. The three-dimensional trajectory and structural connection pattern of the brain white matter fiber bundle includes the preprocessed DWI data.

[0056] S130, determining a functional connectome and a structural connectome based on the preprocessed fMRI data and the preprocessed DWI data.

[0057] In a possible implementation, functional connectivity groups are determined based on preprocessed fMRI data, including: extracting time series of different brain regions based on the preprocessed fMRI data and Shen268 atlas; calculating the Pearson correlation coefficient between each time series; constructing a functional connectivity matrix based on the Pearson correlation coefficient between each time series; determining a tangent space projection matrix based on the functional connectivity matrix and the Euclidean average of the functional connectivity matrix; and determining a functional connectivity group based on the tangent space projection matrix.

[0058] Specifically, Figure 2 A schematic diagram of determining a functional connection group and a structural connection group based on preprocessed fMRI data and preprocessed DWI data provided in an embodiment of the present application, such as Figure 2 As shown, functional connectomes were determined using preprocessed fMRI data.

[0059] Exemplarily, the Shen-268 atlas is used to construct the functional connectome, which divides the brain into 268 nodes, covering the entire cortex, subcutaneous area and cerebellum, and further organized into 10 typical brain networks. For each individual, the BOLD time series of different brain regions are first extracted from the preprocessed fMRI data, and the Pearson correlation coefficient between these time series is calculated to construct the functional connectivity matrix of the individual, i.e., the FC matrix, which is a symmetrical correlation matrix.

[0060] Furthermore, the Euclidean mean of all functional connectivity matrices is calculated as the reference matrix C ref For each individual FC matrix, the main diagonal regularization is first performed to improve the subsequent recognition ability, and then it is projected to C through the preset projection formula. ref The tangent space with the origin as the tangent space is obtained, and the tangent space projection matrix tangent-FC is obtained. The elements of the tangent space projection matrix tangent-FC have escaped the symmetric semi-positive definite constraint and can be regarded as independent features. In the tangent space, appropriate distance metrics such as Euclidean distance can be used for analysis to better capture the functional connectivity differences between individuals. Therefore, the tangent space projection matrix tangent-FC is a functional connectivity group.

[0061] The default projection formula is:

[0062]

[0063] in, Represents the projection of the functional connectome onto the reference matrix C refThe result after the tangent space is the origin is the functional connection group, which is used as the functional connection group feature of subsequent predictive modeling. Q is a symmetric positive definite matrix on the manifold, and C ref is the reference matrix on the manifold, log m Represents the logarithmic transformation of the matrix.

[0064] In a possible implementation, the structural connection group is determined based on the preprocessed DWI data, including: obtaining a fiber orientation distribution image based on the preprocessed DWI data, AAL116 atlas and Fournier algorithm; based on the fiber orientation distribution image and iFOD2 algorithm, using segmented images of three tissue types as labels of the fiber tracking area to perform fiber tracking to obtain fiber tracking data, and the tissue types include white matter, gray matter and cerebrospinal fluid; based on the fiber tracking data and preset fiber length, angle and step threshold, the target fiber bundle is determined, the preset fiber length threshold is 25-250mm, the preset fiber bending angle is 30 degrees, and the preset distance of each tracking iteration is 1mm; based on the target fiber bundle, the geometric characteristics of the target fiber bundle are described by calculating the main direction and spatial distribution of the target fiber bundle, and the average weight and linear combination coefficient of the target fiber bundle are calculated; based on the average weight and linear combination coefficient of the target fiber bundle, the important fiber bundle is determined; based on the important fiber bundle, the streamline number matrix between different nodes of the important fiber bundle is constructed; based on the streamline number matrix, the step map filtering technology based on spherical deconvolution is used to filter the streamlines. Filtering of Tractograms, SIFT) was used as an optimization algorithm to counteract overfitting and determine the structural connectome.

[0065] Specifically, Figure 2 As shown, the structural connectome was determined by preprocessing the DWI data.

[0066] This application uses AAL-116 to define brain structure connection group nodes, adopts constrained spherical deconvolution algorithm to estimate fiber orientation distribution, and performs fiber bundle tracking.

[0067] Exemplarily, the Fournier algorithm is used to estimate the response function for the constrained spherical deconvolution (CSD) in the preprocessed DWI data, and the diffusion direction function of each voxel is calculated to obtain a fiber orientation distribution image.

[0068] Furthermore, in the fiber tract tracking process, the iFOD2 algorithm was used for probabilistic fiber tracking, and three tissue type segmentation images were used as labels for the fiber tracking area, including white matter, gray matter, and cerebrospinal fluid. In order to enhance the connectivity of the fiber tracts, a backtracking strategy was adopted and fiber tracking was terminated at the junction of gray matter and white matter to avoid crossing between different tissue types.

[0069] Furthermore, the sensitivity and selectivity of fiber tracking were controlled by setting the relative threshold and the maximum turning angle threshold (30 degrees) and the step size (1 mm), while the fiber length was limited to between 25 mm and 250 mm to filter out fiber bundles that were too short or too long.

[0070] The junction of gray matter and white matter was selected as the starting point for tracking, and one million fiber bundles were tracked for each subject. To avoid over-representation of certain areas due to the large number of fiber bundles, the iFOD2 algorithm was used to optimize the probabilistic fiber tracking process by presetting threshold constraints such as length and angle, and calculating the average weight and linear combination coefficient of each fiber bundle to determine the important fiber bundles.

[0071] When constructing the streamline number matrix between different nodes based on important fiber bundles, the streamline counts were normalized by node volume to solve the problem of node size differences, thereby determining the streamline number matrix between different nodes of important fiber bundles. At the same time, in order to avoid over-representation caused by too many fiber bundles, the SIFT algorithm was used to offset overfitting, and finally a weighted optimized structural connection group was obtained for each subject.

[0072] Furthermore, a unique structural connection group and functional connection group are obtained for each subject. According to their correlation with behavior, they are divided into positive, negative and overall brain networks. If a connection in the connection group is positively correlated with the behavioral variable, it will be attributed to the positive network, while the overall brain network will treat all related connections as a brain network for the construction of subsequent prediction models. By combining the functional connection group and the structural connection group, a more comprehensive brain network model can be constructed, which helps to understand the relationship between brain function and structure.

[0073] S140, based on the functional connectome, structural connectome and the constructed prediction model based on the multimodal connectome, determine the brain weight network of the cognitive process. The brain weight network of the cognitive process is used to characterize the importance weights between different brain regions and cognitive functions.

[0074] In one possible implementation, a method for constructing a prediction model based on a multimodal connectome includes: extracting feature information from a functional connectome and a structural connectome, wherein the feature information includes the connection strength of a positive network, a negative network, and a whole-brain network; training a prediction model based on the feature information using a ridge regression algorithm; cross-validating the prediction model using a cross-validation formula; and permutation validating the cross-validated training model to determine a prediction model based on a multimodal connectome.

[0075] Specifically, features are extracted from the functional connectome and the structural connectome, including the connection strength of the positive network, the negative network, and the whole-brain network, that is, correlation is used as the predictor variable of the prediction model; features are used as input and individual cognitive measurement scores are used as output to train the prediction model, and ridge regression is used as the machine learning algorithm to prevent the collinearity risk of high-dimensional data. The penalty term of ridge regression is further used as a parameter for optimization, and the prediction performance of different networks (positive network, negative network, whole-brain network) is compared. The combination of network and parameters with the best prediction effect is obtained as the final model for subsequent predictive modeling to obtain better prediction accuracy.

[0076] After determining the prediction model, the model needs to be validated. This application uses cross-validation and permutation validation methods for validation. Specifically, the robustness and generalization ability of the model are evaluated through leave-one-out cross-validation and leave-one-out cross-validation methods; the R between the actual behavior variables and the predicted behavior variables is calculated. 2 To evaluate the model performance, the cross-validation model is:

[0077]

[0078] The coefficient of determination that characterizes cross-validation measures the predictive performance of the model during the cross-validation process. The closer the value is to 1, the better the model's predictive performance is, and the closer the value is to 0, the worse the model's predictive performance is. i The actual value that represents the individual's cognitive performance; Characterize the predictive value of individual cognitive function performance; The average value of all actual observations of the subjects.

[0079] Furthermore, after the cross-validation is passed, a permutation test is performed to verify the statistical significance of the prediction effect and ensure the reliability of the model prediction results.

[0080] S150, determine abnormal connectome characteristics based on functional connectome, structural connectome and large-scale multivariate edge-by-edge analysis.

[0081] Based on the functional connectome and structural connectome, large-scale multivariate edge-by-edge analysis is used to quantify the abnormal patterns of multimodal connectomes between different diagnostic categories. The specific flow chart is as follows: Figure 3 shown.

[0082] In one possible implementation, abnormal connectome characteristics are determined based on the functional connectome, the structural connectome, and a large-scale multivariate edge-by-edge analysis method, including: performing data fusion on the functional connectome and the structural connectome to obtain a connectome matrix; performing a large-scale multivariate edge-by-edge analysis on each connection in the connectome matrix to obtain an F statistic, which is used to characterize the degree of difference between each connection in different diagnostic groups; determining an F value matrix based on the F statistic; and performing multiple comparison correction on the F value matrix to determine abnormal connectome characteristics.

[0083] Specifically, all the data of functional connectome and structural connectome were integrated into a multivariate analysis of variance framework to obtain the connectome matrix; for each connection in the connectome matrix, the corresponding connection strength value was extracted from all acquired modes (structural connectome, functional connectome), and it was input into the two-tailed multivariate analysis of variance model as the dependent variable, and the diagnostic category was used as the independent variable to obtain the F statistic value of the degree of difference between different diagnostic groups for each connection; based on the F statistic value, all edge connections were cyclically calculated to determine the F value matrix; the F value matrix was corrected for multiple comparisons using the method of False Discovery Rate (FDR) < 0.05 to determine the abnormal connectome characteristics.

[0084] S160, based on the brain weight network and abnormal connection group characteristics of cognitive processes, realizes the detection of brain connection patterns of cognitive function deficits in cross-diagnostic mental disorders.

[0085] In a possible implementation, based on the brain weight network of cognitive processes and abnormal connection group characteristics, brain connection pattern detection of cognitive function defects in trans-diagnostic mental disorders is realized, including: based on the node contribution degree calculation formula, calculating the contribution degree of each node in the brain weight network of cognitive processes and the contribution degree of the network; calculating the contribution degree of each node in the abnormal connection group characteristics and the contribution degree of the network; based on the contribution degree of each node in the brain weight network of cognitive processes and the contribution degree of the network, and the contribution degree of each node in the abnormal connection group characteristics and the contribution degree of the network, brain connection pattern detection of cognitive function defects in trans-diagnostic mental disorders is realized; wherein, the node contribution degree calculation formula includes:

[0086]

[0087] Among them, W nThe contribution of the edges between the nth node in the brain weight network of the cognitive process and all other nodes (268 nodes) in the brain network and the contributions of the different connection groups constructed are added together. B(k, m) indicates whether the kth edge is selected in the mth connection group. If it is selected, the function output is 1, otherwise it is 0; std(Ek(:, m)) represents the standard deviation of the kth edge in the mth connection group, βm k Represents the ridge regression algorithm as The weight of the kth connection in the mth connection group.

[0088] The contribution of each node in the abnormal connection group feature is obtained by accumulating the F value between each node and other nodes in the diagnostic group.

[0089] Finally, in terms of network-level contribution, the cumulative average of the contributions of the nodes in a certain network is used as the overall contribution of the network. Then, through correlation analysis, the relationship between the node-level and network-level contribution between the prediction model and the large-scale multivariate edge-by-edge analysis results is quantified. For example, if the Wn 268 data representing each weighted node behind a certain cognitive ability is correlated with the 268 difference weights accumulated by the F value, this indicates that the individual's cognitive deficits have diagnostic specificity; if they are not correlated, this indicates that the specific defects of the brain network between all categories of diseases do not affect the network of individual cognitive deficits, and the brain network of cognitive deficits between them is shared.

[0090] The method described in this application can identify the shared neural mechanisms and unique pathological processes of cognitive deficits in transdiagnostic mental disorders by comparing the contribution of each node and the network in the brain weight network of cognitive processes, as well as the contribution of each node and the network in the abnormal connection group characteristics. Specifically, evaluating the nodes and networks with high correlation in the degree of contribution in the prediction model and large-scale multivariate edge-by-edge analysis can reveal the shared neural basis of cognitive deficits and reflect the common biological pathological pathways of mental illnesses, while nodes and networks with significant differences in contribution may be associated with the unique clinical symptoms and cognitive deficits of each disease, representing specific pathological network abnormalities. Based on these shared and unique network patterns, this application can further explore related potential biomarkers to provide clues for the discovery of new diagnostic indicators for mental illness. It can also evaluate specific network abnormalities in individual patients and systematically explore the neural basis of cognitive deficits.

[0091] Figure 4 This is a structural block diagram of a brain connection pattern detection device based on cognitive function defects provided by an embodiment of the present application. For the sake of convenience, only the part related to the embodiment of the present application is shown. Figure 4The seismic signal denoising device 400 includes an acquisition module 401, a preprocessing module 402, a model building module 403, and a detection module 404.

[0092] In one implementation, the apparatus 400 may be used to implement the above Figure 1 For example, the acquisition module 401 is used to implement S110, the preprocessing module 402 is used to implement S120, the model building module 403 is used to implement S130 to S150, and the detection module 404 is used to implement S160.

[0093] This application can identify shared neural mechanisms and unique pathological processes of cognitive deficits in transdiagnostic mental disorders by comparing the contribution of each node and the network in the brain weight network of cognitive processes, as well as the contribution of each node and the network in the abnormal connection group characteristics. Specifically, evaluating nodes and networks with high correlation in contribution in predictive models and large-scale multivariate edge-by-edge analysis can reveal the shared neural basis of cognitive deficits and reflect the common biological pathological pathways of mental illnesses, while nodes and networks with significant differences in contribution may be associated with the unique clinical symptoms and cognitive deficits of each disease, representing specific pathological network abnormalities. Based on these shared and unique network patterns, this application can further explore related potential biomarkers to provide clues for the discovery of new diagnostic indicators for mental illness. It can also evaluate specific network abnormalities in individual patients and systematically explore the neural basis of cognitive deficits.

[0094] Figure 5 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 5 As shown, the electronic device 5 of this embodiment includes: at least one processor 50 ( Figure 5 Only one is shown in the figure) a processor, a memory 51, and a computer program 52 stored in the memory 51 and executable on the at least one processor 50, and when the processor 50 executes the computer program 52, the steps in any of the above method embodiments are implemented.

[0095] The electronic device 5 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The electronic device may include, but is not limited to, a processor 50 and a memory 51. Those skilled in the art will appreciate that Figure 5 It is only an example of the electronic device 5 and does not constitute a limitation on the electronic device 5. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, it may also include input and output devices, network access devices, etc.

[0096] The processor 50 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0097] In some embodiments, the memory 51 may be an internal storage unit of the electronic device 5, such as a hard disk or memory of the electronic device 5. In other embodiments, the memory 51 may also be an external storage device of the electronic device 5, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 5. Further, the memory 51 may also include both an internal storage unit and an external storage device of the electronic device 5. The memory 51 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program. The memory 51 may also be used to temporarily store data that has been output or is to be output.

[0098] The above description is only a preferred implementation mode of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for detecting brain connection patterns based on cognitive function defects, characterized in that: The method comprises: acquiring neuroimaging data, the neuroimaging data comprising functional magnetic resonance imaging data and diffusion weighted imaging data of healthy controls and different categories of psychiatric disorders; Preprocessing the functional magnetic resonance imaging data and the diffusion weighted imaging data respectively to obtain preprocessed fMRI data and preprocessed DWI data; constructing a functional connectome and a structural connectome based on the preprocessed fMRI data and the preprocessed DWI data; Based on the functional connectome, the structural connectome and the constructed prediction model based on the multimodal connectome, determining a brain weight network of the cognitive process, wherein the brain weight network of the cognitive process is used to characterize the importance weights between different brain regions and cognitive functions; Determining abnormal connectome characteristics based on the functional connectome, the structural connectome, and a large-scale multivariate edge-by-edge analysis method; Based on the brain weight network of the cognitive process and the abnormal connection group characteristics, the brain connection pattern detection of cognitive function deficits in trans-diagnostic mental disorders is realized; The brain weight network based on the cognitive process and the abnormal connection group characteristics are used to detect brain connection patterns of cognitive function defects in trans-diagnostic mental disorders, including: Based on the node contribution degree calculation formula, the contribution degree of each node in the brain weight network of the cognitive process and the contribution degree of the network are calculated; Calculating the contribution of each node and the network in the abnormal connection group feature; Based on the contribution of each node in the brain weight network of the cognitive process and the contribution of the network, and the contribution of each node in the abnormal connection group characteristics and the contribution of the network, the brain connection pattern detection of cognitive function defects in cross-diagnosis mental disorders is realized; The node contribution calculation formula includes: Among them, W n represents the contribution of the nth node in the brain weight network of cognitive processes, B(k,m) indicates whether the kth edge is selected in the mth connection group, and std(E k (:,m)) represents the standard deviation of the kth edge in the mth connected group, βm k Represents the weight of the kth connection group in the mth connection group.

2. The method according to claim 1, characterized in that Preprocessing the functional magnetic resonance imaging data to obtain preprocessed fMRI data includes: Performing format conversion on the functional magnetic resonance imaging data to determine NIfTI format data corresponding to the functional magnetic resonance imaging data; Based on the NIfTI format data corresponding to the functional magnetic resonance imaging data, brain extraction processing is performed to determine a brain image to be processed; Based on the brain image to be processed, head motion correction is performed to determine a brain spatial registration image; Based on the brain spatial registration image and the time corresponding to the brain spatial registration image, determine a brain time image through time layer correction, wherein the brain time image corresponds to the same reference time point; Based on the brain time image, determining a brain smoothed image through spatial smoothing processing; The brain smoothed image is subjected to co-registration processing to determine a brain coordinate image, wherein the brain region in the brain coordinate image corresponds to a unified standard space; Based on the brain coordinate image, converting the signal intensity of the brain coordinate image into a percentage signal change value through percentage signal change conversion to determine a de-differenced image; Based on the de-differenced image, a denoised image is determined by Nuisance regression, wherein the denoised image includes the pre-processed fMRI data.

3. The method according to claim 1, characterized in that Preprocessing the diffusion weighted imaging data to obtain preprocessed DWI data includes: Performing format conversion on the diffusion weighted imaging data to determine NIfTI format data corresponding to the diffusion weighted imaging data; Based on the NIfTI format data corresponding to the diffusion weighted imaging data, an eddy current corrected image is determined by eddy current distortion correction; Based on the eddy current corrected image, head motion correction is performed to determine a spatial registration image of the eddy current corrected image; Registering the spatial registration image of the eddy current correction image to a preset spatial template to determine a standardized image; Based on the standardized image, performing brain extraction processing to determine diffusion weighted imaging data to be processed; Based on the diffusion weighted imaging data to be processed, determining anisotropic diffusion data by diffusion tensor calculation; Based on the anisotropic diffusion data, fiber bundle reconstruction data is determined through fiber bundle reconstruction, and the fiber bundle reconstruction data includes the pre-processed DWI data.

4. The method according to claim 1, characterized in that: Based on the pre-processed fMRI data, a functional connection group is determined, including: Extracting time series of different brain regions based on the preprocessed fMRI data and Shen268 atlas; Calculate the Pearson correlation coefficient between each of the time series; constructing a functional connectivity matrix based on the Pearson correlation coefficients between the time series; determining a tangent space projection matrix based on the functional connectivity matrix and the Euclidean average of the functional connectivity matrix; Based on the tangent space projection matrix, the functional connectome is determined.

5. The method according to claim 1, characterized in that Based on the pre-processed DWI data, a structural connection group is determined, including: Based on the pre-processed DWI data, AAL116 atlas and Fournier algorithm, a fiber orientation distribution image is obtained; Based on the fiber orientation distribution image and the iFOD2 algorithm, segmented images of three tissue types are used as labels of the fiber tracking area to perform fiber tracking and obtain fiber tracking data, wherein the tissue types include white matter, gray matter and cerebrospinal fluid; Determining a target fiber bundle based on the fiber tracking data and a preset fiber length threshold, wherein the preset fiber length threshold is 25-250 mm; Based on the target fiber bundle, calculating the average weight and linear combination coefficient of the target fiber bundle; determining important fiber bundles based on the average weights and linear combination coefficients of the target fiber bundles; Based on the important fiber bundle, constructing a streamline quantity matrix between different nodes of the important fiber bundle; Based on the streamline number matrix, the structural connection group is determined by using a step graph filtering technique to counteract overfitting.

6. The method according to claim 1, characterized in that The method for constructing the prediction model based on the multimodal connectome comprises: Extracting feature information from the functional connectome and the structural connectome, wherein the feature information includes connection strengths of a positive network, a negative network, and a whole-brain network; Based on the feature information, a ridge regression algorithm is used to train a prediction model; Cross-validating the prediction model using a cross-validation formula; The cross-validated training model is subjected to permutation validation to determine the prediction model based on the multimodal connectome.

7. The method according to claim 1, characterized in that Determining abnormal connectome features based on the functional connectome, the structural connectome, and a large-scale multivariate edge-by-edge analysis method includes: Performing data fusion on the functional connectome and the structural connectome to obtain a connectome matrix; Performing large-scale multivariate edge-by-edge analysis on each connection in the connection group matrix to obtain an F statistic value, wherein the F statistic value is used to characterize the degree of difference of each connection between different diagnostic groups; Based on the F-statistic, determining an F-value matrix; The F-value matrix was corrected for multiple comparisons to determine the abnormal connectome characteristics.

8. A device for detecting brain connection patterns based on cognitive function defects, using the method described in any one of claims 1 to 7 for detection, characterized in that: The device comprises: An acquisition module, used for acquiring neuroimaging data, wherein the neuroimaging data includes functional magnetic resonance imaging data and diffusion weighted imaging data of a healthy control group and different types of mental illnesses; A preprocessing module, used for preprocessing the functional magnetic resonance imaging data and the diffusion weighted imaging data respectively to obtain preprocessed fMRI data and preprocessed DWI data; A model building module, used for building a functional connection group and a structural connection group based on the preprocessed fMRI data and the preprocessed DWI data; The model building module is further used to determine a brain weight network of a cognitive process based on the functional connectome, the structural connectome, and a prediction model based on a multimodal connectome that has been constructed, wherein the brain weight network of a cognitive process is used to characterize the importance weights between different brain regions and cognitive functions; The model building module is further used to determine abnormal connectome characteristics based on the functional connectome, the structural connectome and a large-scale multivariate edge-by-edge analysis method; A detection module is used to detect brain connection patterns of cognitive function defects in trans-diagnostic mental disorders based on the brain weight network of the cognitive process and the abnormal connection group characteristics.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

Images