A multimodal dynamic structure-function multi-layer brain network computing method and terminal

Abstract

The present invention discloses a multimodal dynamic structure-function multi-layer brain network calculation method and terminal, the method comprising: acquiring multimodal brain images, preprocessing the multimodal brain images to obtain target multimodal brain images; constructing dynamic functional brain networks, structural brain networks and gray matter morphologically similar brain networks according to the target multimodal brain images; calculating according to the dynamic functional brain networks, the structural brain networks and the gray matter morphologically similar brain networks according to the principle of network topological similarity to obtain structure-function coupling. After preprocessing the MRI data, the present invention uses sMRI to obtain a morphological brain network, fMRI to obtain a functional connection brain network, and DTI to obtain a structural connection brain network, and calculates a multi-layer brain network coupling index, taking into account the structural and functional connections while also taking into account the morphological information, thereby improving the model's exploration of brain structure and function, and providing more comprehensive and accurate brain structure and function information.

Description

Technical Field

[0001] The present invention relates to the field of medical image processing technology, and in particular to a multimodal dynamic structure-function multi-layer brain network calculation method, terminal and computer-readable storage medium. Background Art

[0002] At present, the detection of neuropsychiatric behaviors such as Alzheimer's disease, schizophrenia, depression, etc. mainly includes biomarker testing, neuropsychological assessment, brain imaging, etc. Among them, brain imaging has the advantages of being simple, safe, non-invasive, and objective data compared with biomarker testing and neuropsychological assessment. In recent years, it has been widely used in the early detection and research of neuropsychiatric diseases. Brain imaging of different modalities provides multidimensional information, which can be used by doctors for comprehensive analysis and diagnosis. For example, diffusion tensor imaging (DTI) and structural magnetic resonance imaging (sMRI) can provide structural information of the brain, that is, anatomical information such as the thickness and surface curvature of the cortex in different brain regions, gyri, sulci, or the direction and connection mode of white matter fiber bundles. Resting state functional magnetic resonance imaging (rs-fMRI) can use changes in blood oxygen levels in the brain to obtain information about the functional state of these regions.

[0003] The brain is a complex network system composed of connections between neurons, which are separated and integrated in a specific connection pattern. Analysis from the perspective of brain networks can better understand the structural or functional interactions between specific brain regions. For example, from a structural perspective, AD patients (Alzheimer's disease, AD, Alzheimer's disease patients) are often accompanied by thinning of the cerebral cortex, reduced gray matter volume, hippocampal atrophy, and abnormal deposition of beta-amyloid protein. From the perspective of brain function, common changes in AD patients include weakened functional connectivity within the default mode network. However, changes in brain function and brain structure in AD patients are not independent of each other. On the one hand, pathological changes in brain structure will destroy neurons and synaptic connections and thus affect the function of the entire brain network. On the other hand, changes in brain function may further promote neuronal degeneration. That is, changes in brain function and brain structure in AD patients are interrelated and mutually influential. Combining the structural and functional information of brain images can provide complementary combined advantages of multimodal imaging methods and improve diagnostic accuracy compared with a single modality.

[0004] Brain network structure-function coupling is the combination of brain structural networks and brain functional networks obtained from brain imaging to obtain information about the relationship between the two. Coupling describes the structural support for functional communication. Structure-function coupling occurs when inter-regional white matter connections in cortical regions can predict the strength of inter-regional functional connections. The functional connectivity of large-scale networks is limited by the underlying anatomical white matter pathways of the human brain, but the relationship between structure and function of the human brain has not yet been fully explored.

[0005] Therefore, the prior art still needs to be improved and developed. Summary of the invention

[0006] The main purpose of the present invention is to provide a multimodal dynamic structure-function multi-layer brain network calculation method, terminal and computer-readable storage medium, aiming to solve the problem that the existing technology currently has an insufficient understanding of the relationship between the structure and function of the human brain, and ignores the dynamic characteristics of brain function, morphological information and the relationship between indirectly connected structures and functions.

[0007] To achieve the above object, the present invention provides a multimodal dynamic structure-function multilayer brain network calculation method, which comprises the following steps:

[0008] Acquiring a multimodal brain image, and preprocessing the multimodal brain image to obtain a target multimodal brain image;

[0009] constructing a dynamic functional brain network, a structural brain network and a gray matter morphology-similar brain network according to the target multimodal brain images;

[0010] The structure-function coupling is obtained by calculating based on the dynamic functional brain network, the structural brain network and the gray matter morphology similar brain network according to the principle of network topology similarity.

[0011] Optionally, in the multimodal dynamic structure-function multilayer brain network calculation method, the preprocessing includes T1 anatomical image preprocessing, fMRI preprocessing and DTI preprocessing; the preprocessing is used to remove noise of no interest, correct data and physiological artifacts, and perform spatial and temporal standardization on the data;

[0012] The acquiring of the multimodal brain image and preprocessing the multimodal brain image to obtain the target multimodal brain image specifically includes:

[0013] Acquiring multimodal brain images by a magnetic resonance imaging acquisition machine, wherein the multimodal brain images include diffusion tensor imaging, T1-weighted imaging, and resting-state functional magnetic resonance imaging;

[0014] Performing T1 anatomical image preprocessing on the diffusion tensor imaging to obtain target diffusion tensor imaging;

[0015] performing fMRI preprocessing on the resting-state functional magnetic resonance imaging to obtain a target resting-state functional magnetic resonance imaging;

[0016] The T1-weighted imaging is subjected to DTI preprocessing to obtain a target T1-weighted imaging.

[0017] Optionally, in the multimodal dynamic structure-function multilayer brain network calculation method, the T1 anatomical image preprocessing is used to remove the skull portion and then segment the brain tissue into cerebrospinal fluid, white matter and gray matter to obtain a brain mask, and the T1W1 image is spatially normalized with a selected standard spatial template through nonlinear registration and resampled to a preset size, and a processed anatomical image is obtained after full width at half maximum Gaussian smoothing of a preset selected size;

[0018] The fMRI preprocessing is used to remove the first ten images, perform head motion correction to eliminate the influence of slight head motion displacement and slice timing correction to correct the time difference of different slice acquisition, align the fMRI image and T1-weighted anatomical image to a standard space of preset size by nonlinear change, then calculate frame displacement, DVARS and three regional-level whole-brain signals based on the preprocessed fMRI, perform spatial time series after removing the non-steady-state volume on the preprocessed functional image on MNI and perform spatial smoothing using an isotropic Gaussian kernel of preset size FWHM, and obtain a denoised and registered NIFTI format brain image of preset size to the standard space;

[0019] The DTI preprocessing is used for eddy current and motion artifact correction, diffusion tensor estimation and fractional anisotropy calculation, b0 image extraction, skull stripping, affine alignment of DTI and b0 image, correction of eddy current distortion and motion artifacts in diffusion tensor imaging dataset, and application of transformation to register DTI to standard space.

[0020] Optionally, the multimodal dynamic structure-function multilayer brain network calculation method, wherein the step of constructing a dynamic functional brain network, a structural brain network and a gray matter morphology similar brain network according to the target multimodal brain image, specifically comprises:

[0021] constructing a dynamic functional brain network according to the target resting-state functional magnetic resonance imaging;

[0022] constructing a structural brain network according to the target T1-weighted imaging;

[0023] Based on the target diffusion tensor imaging, a brain network with similar gray matter morphology is constructed.

[0024] Optionally, the multimodal dynamic structure-function multilayer brain network calculation method, wherein the step of constructing a dynamic functional brain network based on the target resting-state functional magnetic resonance imaging, specifically comprises:

[0025] Select a brain map to divide the gray matter into a preset number of brain regions.

[0026] Calculate the element-wise product of the z-scores of the blood oxygen level dependence time series between pairs of brain regions. A pair of regions with baseline ipsilateral activity will have a positive co-fluctuation value, while two regions fluctuating in opposite directions will have a negative co-fluctuation value.

[0027] For each participant, at each time point, a co-fluctuation matrix was obtained, in which each element in the co-fluctuation matrix represented the intensity of the co-fluctuation of two brain regions at the current time point. The matrix was stacked into a time series to form a dynamic functional connectivity matrix, in which nodes represented brain regions and edges represented the functional connections between brain regions.

[0028] Optionally, the multimodal dynamic structure-function multilayer brain network calculation method, wherein the step of constructing a structural brain network based on the target T1-weighted imaging specifically includes:

[0029] Based on FSL's PANDA, the diffusion tensor was calculated, diagonalized to obtain N eigenvalues ​​and corresponding eigenvectors, and the partial anisotropy value of the whole brain was calculated based on each voxel;

[0030] Using the BedpostX and ProbtrackX methods, each identified brain node was selected as a seed region, and a preset number of streamlined fibers were sampled at each voxel in each seed region for probabilistic fiber tracking imaging;

[0031] The probability of connection from the source region to the target region is defined as the number of streamlines crossing the target region divided by the total number of streamlines sampled in the source region;

[0032] After symmetry, the structural connection matrix was constructed by calculating the number of fibers connecting a preset number of brain regions. The consistency threshold method was used for analysis, and a weighted preset number × preset number of white matter structural brain network was constructed for each participant.

[0033] Optionally, the multimodal dynamic structure-function multilayer brain network calculation method, wherein the step of constructing a gray matter morphology-similar brain network based on the target diffusion tensor imaging, specifically comprises:

[0034] The morphological features of each brain region in a subject’s sMRI image were extracted based on the brain atlas;

[0035] The brain is divided into different brain regions according to the brain map, and the similarity of feature distribution between two brain regions is calculated to generate a structural brain network based on morphological similarity for each subject.

[0036] Optionally, the multimodal dynamic structure-function multilayer brain network calculation method, wherein the calculation is performed based on the principle of network topology similarity according to the dynamic functional brain network, the structural brain network and the gray matter morphology similar brain network to obtain structure-function coupling, specifically includes:

[0037] According to the dynamic functional brain network, the structural brain network and the gray matter morphology similar brain network, the structure-function coupling is obtained by the principle of network topological similarity:

[0038] ;

[0039] in, represents the SCN-FCN-GMN coupling of the i-th brain region, where i, j = 1, 2, 3…n, n is the number of brain regions, Corresponding to the mth sample of SCN, Represented by the mth sample of FCN, Represented by the mth sample of GMN, Represents the connectivity between node i and node j in the network;

[0040] The average topological similarity of all nodes was calculated, representing the structural and functional coupling between nodes.

[0041] In addition, to achieve the above-mentioned purpose, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a multimodal dynamic structure-function multi-layer brain network calculation program stored in the memory and executable on the processor, wherein the multimodal dynamic structure-function multi-layer brain network calculation program, when executed by the processor, implements the steps of the multimodal dynamic structure-function multi-layer brain network calculation method as described above.

[0042] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a multimodal dynamic structure-function multilayer brain network calculation program, and when the multimodal dynamic structure-function multilayer brain network calculation program is executed by a processor, the steps of the multimodal dynamic structure-function multilayer brain network calculation method as described above are implemented.

[0043] In the present invention, a multimodal brain image is obtained, and the multimodal brain image is preprocessed to obtain a target multimodal brain image; a dynamic functional brain network, a structural brain network, and a gray matter morphologically similar brain network are constructed according to the target multimodal brain image; and the dynamic functional brain network, the structural brain network, and the gray matter morphologically similar brain network are calculated according to the principle of network topological similarity to obtain structure-function coupling. After preprocessing the MRI data, the present invention uses sMRI to obtain a morphological brain network, fMRI to obtain a functional connection brain network, and DTI to obtain a structural connection brain network, and calculates a multi-layer brain network coupling index. While taking into account the structural and functional connections, morphological information is also taken into account, thereby improving the model's exploration of brain structure and function, and can provide more comprehensive and accurate brain structure and function information. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of a preferred embodiment of the multimodal dynamic structure-function multi-layer brain network calculation method of the present invention;

[0045] Figure 2 It is a schematic diagram of the data collection and data processing process in a preferred embodiment of the multimodal dynamic structure-function multi-layer brain network calculation method of the present invention;

[0046] Figure 3 It is a schematic diagram of a brain network obtained after preprocessing an input image in a preferred embodiment of the multimodal dynamic structure-function multi-layer brain network calculation method of the present invention;

[0047] Figure 4 It is a schematic diagram of the classification performance of the model in the external data set in a preferred embodiment of the multimodal dynamic structure-function multi-layer brain network calculation method of the present invention;

[0048] Figure 5 is a schematic diagram of the correlation between brain regions and scales in a preferred embodiment of the multimodal dynamic structure-function multi-layer brain network calculation method of the present invention;

[0049] Figure 6 It is a structural diagram of a preferred embodiment of the multimodal dynamic structure-function multi-layer brain network computing system of the present invention;

[0050] Figure 7 It is a structural diagram of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0052] The current analysis models for structure-function coupling include:

[0053] (1) Statistical models that calculate structure-function coupling based on the correlation between structural connections directly derived from anatomically connected regions and their corresponding functional connections; however, this model does not take into account the correlation between structural and functional connections in anatomically unconnected regions.

[0054] (2) Communication model, which uses the method of multiple linear regression to link the structural communicability of each brain region with the functional connection brain network. The independent variables of this regression are obtained by combining graph theory knowledge to obtain the structural connection network properties, such as Euclidean distance, path length and transitivity. This network communication model reveals the important role of anatomical networks and communication processes in shaping brain functional activities, and takes into account the relationship between the structure and function of indirect connections. However, there is a collinearity problem between these communication independent variables.

[0055] (3) The graph harmonic model uses the harmonic and Fourier decomposition methods in graph theory to decompose the structural network of the brain into a series of frequency components, thereby identifying coupling patterns at different scales. However, the physiological significance of this model is low.

[0056] (4) Biophysical model: Each brain region in the brain is regarded as a node, and differential equations are used to describe the dynamic behavior of each node, thereby capturing the regulatory effect of structural connections on functional activities. However, biophysical models are too complex to calculate. These models often involve a large number of differential equations and parameters, such as the membrane potential of neurons, synaptic strength, and neurotransmitter release rate. Therefore, the computational cost is high and it is difficult to apply them to large-scale brain networks.

[0057] Therefore, the present invention proposes a novel multimodal dynamic structure-function multi-layer brain network calculation method. After preprocessing MRI (Magnetic Resonance Imaging, magnetic resonance imaging), sMRI is used to obtain the morphological brain network, fMRI is used to obtain the functional connection brain network, and DTI is used to obtain the structural connection brain network. The multi-layer brain network coupling index is calculated. While taking into account the structural and functional connections, the morphological information is also taken into account, thereby improving the model's exploration of brain structure and function.

[0058] The multimodal dynamic structure-function multi-layer brain network calculation method described in the preferred embodiment of the present invention is as follows: Figure 1 and Figure 2 As shown, the multimodal dynamic structure-function multi-layer brain network calculation method includes the following steps:

[0059] Step S10: acquiring a multimodal brain image, and preprocessing the multimodal brain image to obtain a target multimodal brain image.

[0060] Specifically, multimodal brain images are acquired by a magnetic resonance imaging acquisition machine (for example, multimodal data of subjects are acquired by a 5.0T United Imaging magnetic resonance imaging device uMR Jupiter), and the multimodal brain images include diffusion tensor imaging (DTI), T1-weighted imaging (T1WI) and resting-state functional magnetic resonance imaging (fMRI); the preprocessing includes T1 anatomical image preprocessing, fMRI preprocessing and DTI preprocessing; the preprocessing is used to remove noise that is not of interest, correct data and physiological artifacts, and perform spatial and temporal standardization on the data; T1 anatomical image preprocessing is performed on the diffusion tensor imaging (DTI) to obtain target diffusion tensor imaging; fMRI preprocessing is performed on the resting-state functional magnetic resonance imaging (fMRI) to obtain target resting-state functional magnetic resonance imaging; DTI preprocessing is performed on the T1 weighted imaging (T1WI) to obtain target T1 weighted imaging.

[0061] The preprocessing process is as follows:

[0062] (1) T1 anatomical image preprocessing: The SPM software-based toolkit CAT12 was used as a preprocessing tool. The T1 anatomical image preprocessing was used to remove the skull and segment the brain tissue into cerebrospinal fluid, white matter, and gray matter to obtain a brain mask. The T1W1 image was spatially normalized with the selected standard spatial template (MNI152NLin6Asym) through nonlinear registration and resampled to a preset size (e.g., 2 mm × 2 mm × 2 mm). The processed anatomical image was then obtained after Gaussian smoothing with a preset full width at half maxima (FWHM) of a preset selected size (e.g., 4 mm).

[0063] (2) fMRI preprocessing: The fMRI images were preprocessed using preprocessing software such as fMRIprep, which mainly included the following steps: removing the first ten images, performing head motion correction to eliminate the influence of small head motion displacement and slice timing correction to correct the time difference between different slices, registering the fMRI images and T1-weighted anatomical images to a standard space of a preset size (e.g., 2 mm × 2 mm × 2 mm) by nonlinear transformation (e.g., the registration method used for brain imaging data analysis, Diffeomorphic Anatomical Registration through Exponentiated Lie algebra, DARTEL), and then calculating the frame displacement (FD), DVARS, and three regional-level whole-brain signals based on the preprocessed fMRI. The spatial-temporal series after removing the non-steady-state volume and spatially smoothing (full-width half-maximum) using an isotropic Gaussian kernel with a preset size (e.g., 6 mm) FWHM were performed on the preprocessed functional images on the MNI to obtain a denoised and registered NIFTI-formatted brain image of a preset size (e.g., 2 mm × 2 mm × 2 mm) in the standard space.

[0064] (3) DTI preprocessing: DTI images are preprocessed based on FSL’s PANDA software, which mainly includes eddy current and motion artifact correction, diffusion tensor estimation, and fractional anisotropy calculation. First, the b0 image is extracted (b0 image is the image corresponding to b=0 when the DTI image is generated during scanning), and then the skull is stripped. After that, the DTI is affinely aligned with the b0 image to correct the eddy current distortion and motion artifacts in the diffusion tensor imaging data set. Finally, a transformation is applied to align the DTI to the standard space.

[0065] Step S20: constructing a dynamic functional brain network, a structural brain network and a gray matter morphology similar brain network according to the target multimodal brain images.

[0066] Specifically, Figure 2 As shown, according to the target resting-state functional magnetic resonance imaging, a dynamic functional brain network (i.e. Figure 2 According to the target T1-weighted imaging, a structural brain network (i.e. Figure 2 According to the target diffusion tensor imaging, a gray matter morphologically similar brain network (i.e. Figure 2 morphologically similar brain networks in the context of

[0067] The construction process of each brain network is as follows:

[0068] (1) Constructing dynamic functional brain networks:

[0069] After preprocessing the fMRI images, a dynamic functional brain network (FCN) was constructed. The brain atlas selected was the 7 brain network version of Schaefer 2018. The brain atlas was used to segment the gray matter of the brain (excluding the cerebellum) into a preset number (e.g., 400) of brain regions. The dynamic functional brain network is the element product of the z-scores of the blood oxygen level dependence (BOLD) time series between pairs of brain regions. A pair of regions with baseline ipsilateral activity will have a positive co-fluctuation value, while two regions that fluctuate in opposite directions at the same time will have a negative co-fluctuation value. For each participant, at each time point, a co-fluctuation matrix is ​​obtained, in which each element in the co-fluctuation matrix represents the intensity of the co-fluctuation of the two brain regions at the current time point. The matrix is ​​stacked into a time series to form a dynamic functional connectivity matrix. The cross-time average of these co-fluctuation matrices can restore the Pearson correlation coefficient commonly used to define functional connectivity. The Pearson correlation coefficient of functional connectivity is to calculate the average time series of all voxels in the region to obtain the average time series, and then calculate the Pearson correlation coefficient between the average time series of each pair of brain regions. In order to improve the normality of the correlation coefficient, Fisher rz transformation is performed. For each participant, a symmetric functional connectivity (FCN) matrix of a preset number * preset number (for example, 400*400) can be obtained, with nodes representing brain regions and edges representing functional connections between brain regions.

[0070] (2) Construction of structural brain network: The construction of white matter structural brain network (SCN) from DTI is performed using PANDA based on FSL. The diffusion tensor is first calculated and diagonalized to obtain N (for example, 3) eigenvalues ​​and corresponding eigenvectors. Then, the fractional anisotropy (FA) of the whole brain is calculated based on each voxel. The BedpostX (Bayesian estimation of diffusion parameters using sampling technology) and ProbtrackX (probabilistic tracking using crossed fibers) methods are used with the following parameters: 5000 samples per voxel, curvature threshold 0.2, step size 0.5 mm, 2000 steps per sample, each identified brain node was selected as a seed region, and a preset number (e.g. 5000) of streamlined fibers were sampled at each voxel in each seed region for probabilistic fiber tracing imaging; the connection probability from the source area to the target area was defined as the number of streamlines passing through the target area divided by the total number of streamlines sampled from the source area; after symmetry, a structural connection matrix was constructed by calculating the number of fibers (probabilistic streamlines) connecting a preset number (e.g. 400) of brain regions. Similarly, in order to reduce the proportion of false connections in the probabilistic fiber tracking structural network, the consistency threshold method was used for analysis to improve the reliability of the connection; a weighted white matter structural brain network of a preset number × a preset number (e.g. 400×400) was constructed for each participant.

[0071] (3) Constructing gray matter morphologically similar brain networks: A gray matter morphologically similar brain network (GMN) is constructed from sMRI. First, the morphological features of each brain region in a subject’s sMRI image (such as gray matter volume, cortical thickness, or gyrification index) are extracted based on the brain atlas. Then, the brain is divided into different brain regions based on the brain atlas. The similarity of the feature distribution between each pair of brain regions (such as Jensen-Shannon divergence) is calculated, thereby generating a structural brain network based on morphological similarity for each subject.

[0072] like Figure 3 As shown, after the input image is preprocessed by fMRIprep (an open source neuroimaging preprocessing tool), PANDA (a data analysis software, a powerful data processing and analysis tool in Python language) and CAT12 (a statistical analysis tool based on MATLAB), the dynamic functional brain network is obtained by calculating the correlation of bold signals between two brain regions, the structural brain network is obtained by the number of fiber connections between two brain regions, and the gray matter morphological similarity brain network is obtained by the similarity of the characteristic distribution of two brain regions. Figure 3The PDF (Probability Distribution Function) in it represents the probability distribution function, which is used to describe continuous random variables. ROI (Region Of Interest) represents the region of interest, ROI1 represents brain region of interest 1, and ROI2 represents brain region of interest 2.

[0073] Individual gray matter morphologically similar brain networks were constructed on T1-weighted images using a method based on Jensen-Shannon (JS) divergence similarity estimation. JS is a variant of Kullback-Leibler divergence (KL) and is a measure of the distance between two probability distributions. Previous studies have found that JS divergence can produce more test-retest reliable estimates for single-subject cortical morphological brain networks compared with KL divergence. First, for each morphological index, all values ​​in each brain region were extracted based on the brain atlas. Then, the normal kernel function was used to obtain probability density estimates for each brain region and each morphological index. Each of the obtained probability density estimates was further converted into a probability distribution function (PDF). For the probability density functions P(x) and Q(x) of the two regions, the calculation formula of JS is:

[0074]

[0075] in, for P Brain area and Q Brain area KL Divergence, Refers to sampling points, is the number of sampling points, Indicates Sampling points P Brain area, Indicates Sampling points Q Brain area, is the JS divergence, which is the gray matter morphological similarity between the two brain regions. This method is based on the view that brain regions involved in similar cognitive or behavioral processes show developmental coordination or synchronous maturation. Therefore, gray matter that undergoes a common pathological process between brain regions will experience synchronous atrophy. According to the above method, an undirected adjacency matrix can be calculated from the T1-weighted image of each subject. In the matrix, the value range of morphological similarity between regions is 0 to 1, where 0 and 1 respectively represent that the PDFs of the two regions are completely different and completely the same. Therefore, almost the entire network is greater than 0, which is equivalent to that there are connections between brain regions, which is not conducive to the analysis of multi-layer brain networks, so thresholding is performed using a consistency threshold.

[0076] Step S30, calculating according to the dynamic functional brain network, the structural brain network and the gray matter morphology similar brain network based on the principle of network topology similarity to obtain structure-function coupling.

[0077] Specifically, after constructing the dynamic functional brain network, the structural brain network and the gray matter morphology similar brain network respectively, the three brain networks are combined into a multimodal structural-functional brain network. According to the dynamic functional brain network, the structural brain network and the gray matter morphology similar brain network, the structure-function coupling (S-Fcoupling, an indicator obtained by a computational model, which can be used for classification) is obtained through the principle of network topological similarity:

[0078] ;

[0079] in, represents the SCN-FCN-GMN coupling of the i-th brain region, where i, j = 1, 2, 3…n, n is the number of brain regions, Corresponding to the mth sample of SCN, Represented by the mth sample of FCN, Represented by the mth sample of GMN, Represents the connectivity between node i and node j in the network; if node j is and are all connected to node i, then , and are set to 1, otherwise, they are all 0; the average topological similarity of all nodes is also calculated, representing the structural and functional coupling between nodes.

[0080] The present invention first examines the multimodal MRI images to determine whether there are obvious quality problems in the images, such as motion blur, uneven intensity or other artifacts. Then, the brain MRI images are preprocessed using fmriprep and FSL software, and structural brain networks, functional brain networks, and morphological brain networks are constructed; the relationship between structure and function is explored using the network topology similarity between multimodal brain networks, and the pathological characteristics of the structure, function and morphology of gray matter and white matter in neurodegenerative diseases are comprehensively considered, and the calculation resources required are relatively small, thereby achieving a more in-depth exploration of the structural and functional changes of AD, which helps to assist in the early diagnosis of AD.

[0081] Compared with the prior art, the present invention not only preserves the exploration of the relationship between structure and function, but also takes into account morphological information, and does not require parameter adjustment like biophysical models. The calculation is simple and easy to implement, and is more suitable for MRI brain imaging studies of diseases such as AD that can cause disconnection between brain structure and function.

[0082] For example, the present invention analyzed the DTI and fMRI data of 214 ADNI patients, including 112 MCI patients, 41 AD patients and 102 healthy controls. The final results show (such as Figure 4 and Figure 5 As shown, Figure 4 The ns (nosignificance) in D indicates no significant difference. The SF coupling obtained by the method proposed in the present invention has different distributions in the whole brain. The SF coupling values ​​of the visual network and the somatomotor network are generally lower than those of other networks ( Figure 4 , Group average SF coupling of the HC group, A: Regional SF coupling coupling varies in the cortex, ranging from -0.0082 to 0.0277, the SF coupling index of the occipital and parietal cortices is low, while the coupling of the frontal and parietal regions is relatively high, B: Seven brain networks of the Yeo atlas C: Distribution of SF coupling in seven brain networks D: Pairwise t-tests were performed between networks, and the SF coupling values ​​of the visual network and the somatomotor network were generally lower than those of other networks). The SF coupling values ​​of each brain region in the MCI group and the AD group were correlated with the MOCA scale. And a brain region in the default mode network had the strongest correlation with the scale (R = 0.024, P < 0.001, R indicates correlation, P indicates significance, as shown in Figure 5 A and B in the figure, where MOCA (Montrealcognitive assessment) stands for Montreal cognitive assessment). The present invention extracted the S-Fcoupling value of this brain region and evaluated the differences between APOE subgroups. The present invention found that SF coupling in non-APOE4 carriers was significantly higher than that in carriers (such as Figure 5 C in, P ≤ 0.01, for example, P = 0.008, APOE4 represents a gene, namely apolipoprotein E, which is a soluble alkaline glycosylated secretory protein).

[0083] Furthermore, if Figure 6 As shown, based on the above-mentioned multimodal dynamic structure-function multi-layer brain network calculation method, the present invention also provides a multimodal dynamic structure-function multi-layer brain network calculation system, wherein the multimodal dynamic structure-function multi-layer brain network calculation system includes:

[0084] A data preprocessing module 51 is used to obtain multimodal brain images and preprocess the multimodal brain images to obtain target multimodal brain images;

[0085] A brain network construction module 52, used to construct a dynamic functional brain network, a structural brain network and a gray matter morphology similar brain network according to the target multimodal brain image;

[0086] The brain network calculation module 53 is used to calculate the dynamic functional brain network, the structural brain network and the gray matter morphology similar brain network according to the principle of network topology similarity to obtain structure-function coupling.

[0087] Furthermore, if Figure 7 As shown, based on the above-mentioned multimodal dynamic structure-function multi-layer brain network calculation method and system, the present invention also provides a terminal accordingly, and the terminal includes a processor 10, a memory 20 and a display 30. Figure 7 Only some components of the terminal are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0088] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the terminal. Further, the memory 20 may also include both an internal storage unit of the terminal and an external storage device. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code of the installation terminal. The memory 20 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a multimodal dynamic structure-function multilayer brain network calculation program 40 is stored on the memory 20, and the multimodal dynamic structure-function multilayer brain network calculation program 40 can be executed by the processor 10, thereby realizing the multimodal dynamic structure-function multilayer brain network calculation method in the present application.

[0089] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code or process data stored in the memory 20, such as executing the multimodal dynamic structure-function multi-layer brain network calculation method.

[0090] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 30 is used to display information on the terminal and to display a visual user interface. The components 10-30 of the terminal communicate with each other via a system bus.

[0091] In one embodiment, when the processor 10 executes the multimodal dynamic structure-function multilayer brain network calculation program 40 in the memory 20, the steps of the multimodal dynamic structure-function multilayer brain network calculation method described above are implemented.

[0092] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a multimodal dynamic structure-function multilayer brain network calculation program, and when the multimodal dynamic structure-function multilayer brain network calculation program is executed by a processor, the steps of the multimodal dynamic structure-function multilayer brain network calculation method as described above are implemented.

[0093] In summary, the present invention provides a multimodal dynamic structure-function multi-layer brain network calculation method, system, terminal and storage medium, the method comprising: obtaining multimodal brain images, preprocessing the multimodal brain images to obtain target multimodal brain images; constructing dynamic functional brain networks, structural brain networks and gray matter morphologically similar brain networks according to the target multimodal brain images; calculating according to the dynamic functional brain networks, the structural brain networks and the gray matter morphologically similar brain networks according to the principle of network topological similarity to obtain structure-function coupling. After preprocessing the MRI data, the present invention uses sMRI to obtain a morphological brain network, fMRI to obtain a functional connection brain network, and DTI to obtain a structural connection brain network, and calculates a multi-layer brain network coupling index. While taking into account the structural and functional connections, morphological information is also taken into account, thereby improving the model's exploration of brain structure and function, and can provide more comprehensive and accurate brain structure and function information.

[0094] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or terminal including the element.

[0095] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the program can be stored in a computer-readable storage medium that can be read by a computer, and the program can include the processes of the above-mentioned method embodiments when executed. The computer-readable storage medium can be a memory, a disk, an optical disk, etc.

[0096] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A multimodal dynamic structure-function multi-layer brain network calculation method, characterized in that: The multimodal dynamic structure-function multi-layer brain network calculation method comprises: Acquiring a multimodal brain image, and preprocessing the multimodal brain image to obtain a target multimodal brain image; constructing a dynamic functional brain network, a structural brain network and a gray matter morphology-similar brain network according to the target multimodal brain images; The structure-function coupling is obtained by calculating the dynamic functional brain network, the structural brain network and the gray matter morphology similar brain network according to the principle of network topology similarity; The preprocessing includes T1 anatomical image preprocessing, fMRI preprocessing and DTI preprocessing; the preprocessing is used to remove noise that is not of interest, correct data and physiological artifacts, and perform spatial and temporal standardization on the data; The acquiring of the multimodal brain image and preprocessing the multimodal brain image to obtain the target multimodal brain image specifically includes: Acquiring multimodal brain images by a magnetic resonance imaging acquisition machine, wherein the multimodal brain images include diffusion tensor imaging, T1-weighted imaging, and resting-state functional magnetic resonance imaging; Performing T1 anatomical image preprocessing on the diffusion tensor imaging to obtain target diffusion tensor imaging; performing fMRI preprocessing on the resting-state functional magnetic resonance imaging to obtain a target resting-state functional magnetic resonance imaging; Performing DTI preprocessing on the T1-weighted imaging to obtain target T1-weighted imaging; The step of constructing a dynamic functional brain network, a structural brain network and a gray matter morphology similar brain network according to the target multimodal brain images specifically includes: constructing a dynamic functional brain network according to the target resting-state functional magnetic resonance imaging; constructing a structural brain network according to the target T1-weighted imaging; constructing a brain network with similar gray matter morphology according to the target diffusion tensor imaging; The constructing of a dynamic functional brain network according to the target resting-state functional magnetic resonance imaging specifically includes: Select a brain map to divide the gray matter into a preset number of brain regions. Calculate the element-wise product of the z-scores of the blood oxygen level dependence time series between pairs of brain regions. A pair of regions with baseline ipsilateral activity will have a positive co-fluctuation value, while two regions fluctuating in opposite directions will have a negative co-fluctuation value. For each participant, at each time point, a co-fluctuation matrix was obtained, where each element in the co-fluctuation matrix represents the co-fluctuation intensity of two brain regions at the current time point. The matrix was stacked into a time series to form a dynamic functional connectivity matrix, where nodes represent brain regions and edges represent functional connections between brain regions. The constructing of a structural brain network according to the target T1-weighted imaging specifically includes: Based on FSL's PANDA, the diffusion tensor was calculated, diagonalized to obtain N eigenvalues ​​and corresponding eigenvectors, and the partial anisotropy value of the whole brain was calculated based on each voxel; Using the BedpostX and ProbtrackX methods, each identified brain node was selected as a seed region, and a preset number of streamlined fibers were sampled at each voxel in each seed region for probabilistic fiber tracking imaging; The probability of connection from the source region to the target region is defined as the number of streamlines crossing the target region divided by the total number of streamlines sampled in the source region; After symmetrization, the number of fibers connecting the preset number of brain regions was calculated to construct a structural connection matrix, which was analyzed using the consistency threshold method to construct a weighted preset number × preset number of white matter structural brain networks for each participant. The step of constructing a gray matter morphology-similar brain network according to the target diffusion tensor imaging specifically includes: The morphological features of each brain region in a subject’s sMRI image were extracted based on the brain atlas; The brain is divided into different brain regions according to the brain map, and the similarity of feature distribution between two brain regions is calculated to generate a structural brain network based on morphological similarity for each subject.

2. The multimodal dynamic structure-function multilayer brain network calculation method according to claim 1, characterized in that: The T1 anatomical image preprocessing is used to remove the skull part and segment the brain tissue into cerebrospinal fluid, white matter and gray matter to obtain a brain mask, and to spatially normalize the T1W1 image and the selected standard spatial template through nonlinear registration and resample to a preset size, and to obtain a processed anatomical image after Gaussian smoothing with a preset selected size at half maximum full width; The fMRI preprocessing is used to remove the first ten images, perform head motion correction to eliminate the influence of slight head motion displacement and slice timing correction to correct the time difference of different slice acquisition, align the fMRI image and T1-weighted anatomical image to a standard space of preset size by nonlinear change, then calculate frame displacement, DVARS and three regional-level whole-brain signals based on the preprocessed fMRI, perform spatial time series after removing the non-steady-state volume on the preprocessed functional image on MNI and perform spatial smoothing using an isotropic Gaussian kernel of preset size FWHM, and obtain a denoised and registered NIFTI format brain image of preset size to the standard space; The DTI preprocessing is used for eddy current and motion artifact correction, diffusion tensor estimation and fractional anisotropy calculation, b0 image extraction, skull stripping, affine alignment of DTI and b0 image, correction of eddy current distortion and motion artifacts in diffusion tensor imaging dataset, and application of transformation to register DTI to standard space.

3. The multimodal dynamic structure-function multilayer brain network calculation method according to claim 1, characterized in that: The calculation is performed based on the dynamic functional brain network, the structural brain network and the gray matter morphology similar brain network according to the principle of network topology similarity to obtain structure-function coupling, which specifically includes: According to the dynamic functional brain network, the structural brain network and the gray matter morphology similar brain network, the structure-function coupling is obtained by the principle of network topological similarity: ； in, represents the SCN-FCN-GMN coupling of the i-th brain region, SCN represents the structural brain network, FCN represents the dynamic functional brain network, and GMN represents the gray matter morphological similarity brain network, where i, j = 1, 2, 3…n, n is the number of brain regions, Corresponding to the mth sample of SCN, Represented by the mth sample of FCN, Represented by the mth sample of GMN, Represents the connectivity between node i and node j in the network; The average topological similarity of all nodes was calculated, representing the structural and functional coupling between nodes.

4. A terminal, characterized in that: The terminal includes: a memory, a processor, and a multimodal dynamic structure-function multi-layer brain network calculation program stored in the memory and executable on the processor. When the multimodal dynamic structure-function multi-layer brain network calculation program is executed by the processor, the steps of the multimodal dynamic structure-function multi-layer brain network calculation method as described in any one of claims 1 to 3 are implemented.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a multimodal dynamic structure-function multilayer brain network calculation program, and when the multimodal dynamic structure-function multilayer brain network calculation program is executed by a processor, the steps of the multimodal dynamic structure-function multilayer brain network calculation method as described in any one of claims 1 to 3 are implemented.

Images