Longitudinal analysis method and system for magnetic resonance imaging data of mild brain injury

Abstract

The present invention belongs to the field of rehabilitation therapy technology and discloses a longitudinal analysis method and system for magnetic resonance imaging data of mild brain injury. The method extracts BOLD signals from the magnetic resonance imaging data of subjects; constructs a symmetric positive definite sparse brain functional connectivity network set of subjects based on sparse inverse covariance matrix estimation; determines the brain functional connectivity network dictionary and sparse coefficient matrix in kernel space based on Riemannian manifold sparse coding; performs spatial distribution analysis of brain functional connectivity atomic networks; and performs longitudinal analysis of magnetic resonance imaging data of mild brain injury. By analyzing the differences in the spatial distribution of these highly present brain functional connectivity atomic networks in the brain, the present invention digs out brain functional connectivity imaging markers for distinguishing the three mild brain injury rehabilitation treatment stages: acute phase, subacute phase, and complete recovery, thereby realizing longitudinal analysis of the mild brain injury rehabilitation process.

Description

Technical Field

[0001] The present invention belongs to the technical field of rehabilitation medicine data analysis, and in particular relates to a longitudinal analysis method and system for magnetic resonance imaging data of mild brain injury. Background Art

[0002] The human brain is an organic, unified whole, with various brain regions collaborating to accomplish higher-level cognitive functions. The brain functional connectivity network intuitively describes the interactions between spatially distant brain regions during cognitive function execution. The resting-state brain functional connectivity network reflects the functional patterns of spontaneous neural activity in the brain and is an important technical tool for studying the neural mechanisms of cognitive disorders, with excellent clinical applicability.

[0003] Clinically, mild brain injury manifests primarily as varying degrees of cognitive, behavioral, and emotional impairments. Common medical imaging techniques, such as CT and conventional MRI, struggle to visualize lesions. Currently, CT images are ineffective in providing a reliable basis for clinical diagnosis of mild brain injury, while the rich resources of magnetic resonance imaging data remain underutilized. Clinically, physicians primarily rely on their own medical experience to make qualitative judgments based on patients' subjective experiences and self-described symptoms, lacking objective and reliable indicators.

[0004] At present, most brain function analysis methods for mild brain injury based on resting-state functional magnetic resonance imaging use the vectorization method in Euclidean space for analysis. The vectorization method is a linear analysis method that cannot accurately characterize the nonlinear essential characteristics of brain functional connections. Summary of the Invention

[0005] In response to the problems existing in the prior art, the present invention provides a longitudinal analysis method and system for magnetic resonance imaging data of mild brain injury.

[0006] The present invention is implemented as follows: a longitudinal analysis method for magnetic resonance imaging data of mild brain injury, the longitudinal analysis method for magnetic resonance imaging data of mild brain injury comprising the following steps:

[0007] Step 1: Using the BOLD signal extracted from magnetic resonance imaging data, we obtain measurement data reflecting the brain functional activity status of each brain region of interest;

[0008] Step 2: Based on the sparse inverse covariance matrix estimation, a symmetric positive definite sparse brain functional connection network set of each subject is obtained. The positive definite sparse brain functional connection network reflects the interaction relationship between the various brain regions of the subject;

[0009] Step 3: Based on Riemannian manifold sparse coding, the dictionary of the positive definite sparse brain functional connectivity network and its sparse coefficient matrix in the kernel space are determined, and the component characteristics of the positive definite sparse brain functional connectivity network are analyzed in the manifold space;

[0010] Step 4: Perform spatial distribution analysis of the brain functional connectivity atomic network to obtain the spatial distribution of the brain functional connectivity subnetwork characteristics of each subject;

[0011] Step five: longitudinal analysis of magnetic resonance imaging data of mild brain injury is conducted to identify diseases by analyzing the spatial distribution pattern differences of brain functional connection subnetwork characteristics between healthy subjects and diseased subjects.

[0012] Furthermore, in step 2, the construction of the sparse brain functional connectivity network includes:

[0013] Let {x1, x2, ..., x M} represents the BOLD signal time series of the brain region of interest with a length of M, where x i is a d-dimensional vector, corresponding to d brain regions of interest. Assume x i Obey an unknown multidimensional Gaussian distribution Among them, μ is the mean vector of the multidimensional Gaussian distribution, Σ is the covariance matrix of the multidimensional Gaussian distribution, both of which are unknown parameters. The brain functional connectivity network can be obtained by the inverse covariance matrix Σ of the BOLD signal time series. -1 Estimation is performed to construct,Σ -1 The non-main diagonal elements of represent a partial correlation between pairs of brain regions of interest after eliminating the influence of other brain regions of interest, that is, represents the connection strength between the brain regions of interest i and j. If the brain regions of interest i and j are independent of each other, then is 0. Estimate the partial correlation between each pair of brain regions of interest:

[0014]

[0015] in, is the BOLD signal time series {x1, x2, ...x M}; det(·) represents the matrix determinant operator, tr(·) represents the matrix trace operator, ||·||1 represents the sum of the absolute values ​​of all elements of the matrix, λ>0 is the preset sparse regularization parameter, Representing sparse brain functional connectivity networks Due to the sparse characteristics of limited connections between a brain region of interest and other brain regions in brain neural activity, by -1 Increase Sparsity constraint, i.e. A more realistic brain functional connection network is estimated. When the sparse regularization parameter λ is relatively small, the Graphical Lasso method is used to solve equation (1) to obtain Represented as a sparse brain functional connectivity network.

[0016] Furthermore, in step 3, determining the brain functional connectivity network dictionary and the sparse coefficient matrix in the kernel space according to Riemannian manifold sparse coding includes:

[0017] Sparse brain functional connectivity network training set for N subjects where s n represents the representative sparse brain functional connectivity network of the nth subject, estimated by the sparse inverse covariance matrix of the fMRI BOLD signal of the subject get, represents the d×d-dimensional symmetric positive definite matrix space.

[0018] Hypothetical dictionary Feature Map represents the mapping function from symmetric positive definite matrix space to Hilbert space. For any given sparse inverse covariance matrix s n , and obtain its sparse coding by solving the following formula:

[0019]

[0020] in, is the sparse coefficient vector, a j It is about dictionary atom D j The coefficient of .

[0021] When the sparse brain functional connection network training set s={s1,s2,…s N}By patients with mild brain damage G1 ={s1, ...s n} and healthy control group G2 ={s n+1 ,…,s N When two groups of types are formed, group sparse coding is used for dictionary learning. Its basic form is as follows:

[0022]

[0023] Among them, the sparse coefficient matrix A G1 ={A1, ..., A n} and A G2 ={A n+1 ,…,A N} represent the sparse coefficient matrices corresponding to the mild brain injury patient group and the healthy control group, respectively, indicating the spatial distribution of the brain functional connection atomic network of the mild brain injury patient group and the healthy control group.

[0024] By selecting a suitable symmetric positive definite kernel with reproducing kernel properties, the kernel technique is used to simplify Formula (3) into a kernel sparse coding problem on a Riemannian manifold with a convex optimization form.

[0025] Solve this problem using the stochastic gradient algorithm or ADMM algorithm, and obtain the dictionary That is, it is a set of brain functional connection atomic network patterns, representing a set of brain functional connection atomic networks shared by all subjects, including patients with mild brain injury and healthy controls, which can map the spatial distribution of brain functional connection networks with characteristic descriptions in the brain space of all subjects.

[0026] Furthermore, in step 4, the spatial distribution analysis of the brain functional connectivity atomic network includes:

[0027] For the longitudinal analysis of mild brain injury rehabilitation diagnosis, a sparse brain functional connectivity network training set s={s G1 , s G2 , s G3}, using group sparse coding for dictionary learning:

[0028]

[0029] Among them, the sparse coefficient matrix A G1 、A G2 、A G3 They represent the sparse coefficient matrices corresponding to the acute patient group, subacute rehabilitation patient group, and healthy control group, respectively, and reflect the distribution statistical information of the brain functional connection atomic network of the three groups in the brain network space.

[0030] Further, for A G1 、A G2 、A G3 Non-zero T-tests were performed separately to count the number of non-zero citations of each brain functional connection atomic network in each group, and to find the brain functional connection atomic network with a significantly high presentation rate in each group.

[0031] By analyzing the differences in the spatial distribution of these highly present brain functional connection atomic networks in the brain, we can use them to identify the brain functional connection characteristics in the three stages of rehabilitation treatment for mild brain injury: acute, subacute, and complete recovery, and achieve a longitudinal analysis of the rehabilitation process of mild brain injury.

[0032] Another object of the present invention is to provide a longitudinal analysis system for mild brain injury magnetic resonance imaging data using the longitudinal analysis method for mild brain injury magnetic resonance imaging data, the longitudinal analysis system for mild brain injury magnetic resonance imaging data comprising:

[0033] A signal extraction module, for extracting BOLD signals from magnetic resonance imaging data of a subject;

[0034] A network set determination module is used to construct a symmetric positive definite sparse brain functional connection network set of the subject based on the sparse inverse covariance matrix estimation;

[0035] The network feature pattern mining module is used to determine the brain functional connection network dictionary and sparse coefficient matrix in the kernel space based on Riemannian manifold sparse coding;

[0036] Network spatial distribution analysis module, used to analyze the spatial distribution of brain functional connectivity atomic networks;

[0037] Disease discrimination module for longitudinal analysis of magnetic resonance imaging data of mild brain injury.

[0038] Another object of the present invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the following steps:

[0039] Methods: BOLD signals were extracted from the subjects' magnetic resonance imaging data; the subjects' symmetric positive definite sparse brain functional connectivity network set was determined based on sparse inverse covariance matrix estimation; the brain functional connectivity network dictionary and sparse coefficient matrix in kernel space were determined based on Riemannian manifold sparse coding; the spatial distribution of brain functional connectivity atomic networks was analyzed; and longitudinal analysis of magnetic resonance imaging data of mild brain injury was performed.

[0040] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor performs the following steps:

[0041] Methods: BOLD signals were extracted from the subjects' magnetic resonance imaging data; the subjects' symmetric positive definite sparse brain functional connectivity network set was determined based on sparse inverse covariance matrix estimation; the brain functional connectivity network dictionary and sparse coefficient matrix in kernel space were determined based on Riemannian manifold sparse coding; the spatial distribution of brain functional connectivity atomic networks was analyzed; and longitudinal analysis of magnetic resonance imaging data of mild brain injury was performed.

[0042] Another object of the present invention is to provide an information data processing terminal, which is used to implement the longitudinal analysis system of mild brain injury magnetic resonance imaging data of mild brain injury.

[0043] Combining all the above technical solutions, the advantages and positive effects of the present invention are as follows:

[0044] The longitudinal analysis method of magnetic resonance imaging data of mild brain injury provided by the present invention selects a suitable symmetric positive definite kernel with reproducing kernel properties (such as the Stein kernel, etc.), and uses kernel techniques to simplify the formula into a kernel sparse coding problem on a Riemannian manifold with a convex optimization form. By solving this problem, the resulting dictionary is used to represent the atomic network pattern of brain functional connectivity.

[0045] The present invention analyzes the differences in the spatial distribution of these highly present brain functional connection atomic networks in the brain, and uses them to discriminate the brain functional connection imaging characteristics in the three stages of rehabilitation treatment for mild brain injury: acute stage, subacute stage, and complete recovery, thereby realizing a longitudinal analysis of the rehabilitation process of mild brain injury.

[0046] The present invention conducts brain network analysis from the perspective of Riemannian manifolds, which can obtain the nonlinear structural information contained in brain functional connections. It breaks through the common Euclidean space vectorization method that cannot deeply reflect the essential nonlinear characteristics of the brain functional connection network, and is more conducive to the accurate identification of brain functional diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0048] Figure 1 This is a flow chart of a longitudinal analysis method for magnetic resonance imaging data of mild brain injury provided by an embodiment of the present invention.

[0049] Figure 2 This is a schematic diagram of the principle of the longitudinal analysis method of magnetic resonance imaging data of mild brain injury provided by an embodiment of the present invention.

[0050] Figure 3 This is a structural block diagram of a longitudinal analysis system for magnetic resonance imaging data of mild brain injury provided by an embodiment of the present invention;

[0051] In the figure: 1. Signal extraction module; 2. Network set determination module; 3. Network feature pattern mining module; 4. Network space distribution analysis module; 5. Disease status discrimination module. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0053] In response to the problems existing in the prior art, the present invention provides a longitudinal analysis method and system for magnetic resonance imaging data of mild brain injury. The present invention is described in detail below with reference to the accompanying drawings.

[0054] like Figure 1 As shown, the longitudinal analysis method of magnetic resonance imaging data of mild brain injury provided by the embodiment of the present invention includes the following steps:

[0055] S101, using BOLD signals extracted from magnetic resonance imaging data to obtain measurement data reflecting brain functional activity status of each brain region of interest;

[0056] S102, constructing a symmetric positive definite sparse brain functional connectivity network set for each subject based on sparse inverse covariance matrix estimation, wherein the positive definite sparse brain functional connectivity network reflects the interaction relationship between the brain regions of the subject;

[0057] S103, based on Riemannian manifold sparse coding, determining a positive definite sparse brain functional connectivity network dictionary and its sparse coefficient matrix in the kernel space, and analyzing the component features of the positive definite sparse brain functional connectivity network in the manifold space;

[0058] S104, performing a spatial distribution analysis of the brain functional connectivity atomic network to obtain the spatial distribution pattern of the brain functional connectivity subnetwork characteristics of each subject;

[0059] S105: Longitudinal analysis of magnetic resonance imaging data of mild brain injury is performed. Disease identification is performed by analyzing the spatial distribution pattern differences of brain functional connectivity subnetwork characteristics between healthy subjects and diseased subjects.

[0060] The principle diagram of the longitudinal analysis method of magnetic resonance imaging data of mild brain injury provided by the embodiment of the present invention is as follows Figure 2 shown.

[0061] like Figure 3 As shown, the longitudinal analysis system for magnetic resonance imaging data of mild brain injury provided by the embodiment of the present invention includes:

[0062] A signal extraction module 1 is used to extract BOLD signals from the subject's magnetic resonance imaging data;

[0063] A network set determination module 2 is used to construct a symmetric positive definite sparse brain functional connection network set of the subject based on the sparse inverse covariance matrix estimation;

[0064] Network feature pattern mining module 3, used to determine the brain functional connection network dictionary and its sparse coefficient matrix in the kernel space based on Riemannian manifold sparse coding;

[0065] Network spatial distribution analysis module 4, used to analyze the spatial distribution pattern of brain functional connection atomic networks;

[0066] The disease condition identification module 5 is used for longitudinal analysis of magnetic resonance imaging data of mild brain injury.

[0067] The technical solution of the present invention is further described below in conjunction with embodiments.

[0068] Example: Spatial characteristics of brain functional connectivity network in mild brain injury based on Riemannian manifold sparse representation:

[0069] The technical route adopted by the present invention for the spatial characteristics of brain functional connection network of mild brain injury is as follows Figure 2 shown.

[0070] The longitudinal analysis method of magnetic resonance imaging data of mild brain injury provided by an embodiment of the present invention includes:

[0071] 1. Construction of sparse brain functional connectivity network

[0072] {x1, x2, ..., x M} represents the BOLD signal time series of the brain region of interest with a length of M, where x i is a d-dimensional vector, corresponding to d brain regions of interest. Assume x i Obey an unknown multidimensional Gaussian distribution Among them, μ is the mean vector of the multidimensional Gaussian distribution, Σ is the covariance matrix of the multidimensional Gaussian distribution, both of which are unknown parameters. The brain functional connectivity network can be obtained by the inverse covariance matrix Σ of the BOLD signal time series. -1 Estimation is performed to construct,Σ -1 The non-main diagonal elements of represent a partial correlation between pairs of brain regions of interest after eliminating the influence of other brain regions of interest, that is, represents the connection strength between the brain regions of interest i and j. If the brain regions of interest i and j are independent of each other, then is 0. Estimate the partial correlation between each pair of brain regions of interest:

[0073]

[0074] in, is the BOLD signal time series {x1, x2, ..., x M}; det(·), tr(·) and ||·||1 represent the matrix determinant operator, the matrix trace operator and the sum of the absolute values ​​of all matrix elements, respectively. λ>0 is the preset sparse regularization parameter. Representing sparse brain functional connectivity networks Considering the sparse characteristics of limited connections between a brain region of interest and other brain regions in brain neural activity, by -1 Increase Sparsity constraint, i.e. ||Σ -1 ||1, a more realistic brain functional connection network can be estimated. When the sparse regularization parameter λ is relatively small, the Graphical Lasso method is used to solve Equation (1) to obtain To represent sparse brain functional connectivity networks.

[0075] 2. Mining characteristic patterns of atomic networks of brain functional connectivity

[0076] Sparse brain functional connectivity network training set for N subjects where s n represents the representative sparse brain functional connectivity network of the nth subject, estimated by the sparse inverse covariance matrix of the fMRI BOLD signal of the subject get, represents the d×d-dimensional symmetric positive definite matrix space.

[0077] Hypothetical dictionary Feature Map represents the mapping function from symmetric positive definite matrix space to Hilbert space. For any given sparse inverse covariance matrix s n , we can solve the following formula to obtain its sparse coding:

[0078]

[0079] in, is the sparse coefficient vector, a j It is about dictionary atom D j The coefficient of .

[0080] Since the present invention is aimed at the auxiliary diagnosis and analysis of mild brain injury, the sparse brain functional connection network training set s={s1,s2,…s N ] Mainly by patients with mild brain damage G1 ={s1,…,s n} and healthy control group G2 ={s n+1 ,…,s N}It consists of two groups of types.

[0081] In order to improve the discrimination ability of the sparse representation classifier, the present invention proposes to use group sparse coding for dictionary learning, the basic form of which is as follows:

[0082]

[0083] Among them, the sparse coefficient matrix , A G1 ={A1, ..., A n} and A G2 ={A n+1 ,…,A N} represent the sparse coefficient matrices corresponding to the mild brain injury patient group and the healthy control group, respectively.

[0084] The present invention selects a suitable symmetric positive definite kernel (such as the Stein kernel) with reproducing kernel properties and uses kernel techniques to simplify formula (3) into a kernel sparse coding problem on a Riemannian manifold with a convex optimization form.

[0085] Solve this problem using the stochastic gradient algorithm or ADMM algorithm, and obtain the dictionary It is a set of brain functional connection atomic network patterns, representing a set of brain functional connection atomic networks shared by all subjects, including patients with mild brain injury and healthy controls, and mapping out the spatial distribution of brain functional connection networks with characteristic descriptions in the brain space of all subjects. n Combined with the sparse dictionary, a sparse brain functional connection network s can be reconstructed for each sample. n .

[0086] 3. Longitudinal analysis of MRI data of mild brain injury for the rehabilitation process of mild brain injury

[0087] Construct a sparse brain functional connectivity network training set s={s G1 , s G2 , s G3}. We plan to use group sparse coding for dictionary learning:

[0088]

[0089] Among them, the sparse coefficient matrix A G1 、A G2 、A G3 The sparse coefficient matrices respectively represent the acute patient group, subacute rehabilitation patient group, and healthy control group, and retain the statistical information of the spatial distribution patterns of the sparse brain functional connection networks of the three groups.

[0090] To A G1 、A G2 、A G3 Non-zero T-tests were performed separately to count the number of non-zero citations of each brain functional connectivity atomic network in each group, identifying the brain functional connectivity atomic networks with significantly high prevalence in each group. By analyzing the differences in the spatial distribution of these highly prevalent brain functional connectivity atomic networks in the brain, the researchers used them to identify brain functional connectivity characteristics across three stages of mild brain injury rehabilitation treatment: acute, subacute, and complete recovery, enabling a longitudinal analysis of the mild brain injury rehabilitation process.

[0091] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented in whole or in part in the form of a computer program product, the computer program product includes one or more computer instructions. When the computer program instructions are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more available media integrated therein.

[0092] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with this technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered by the scope of protection of the present invention.

Claims

1. A longitudinal analysis method for magnetic resonance imaging data of mild brain injury, characterized in that: The longitudinal analysis method of mild brain injury magnetic resonance imaging data comprises the following steps: Step 1: Using the BOLD signal extracted from magnetic resonance imaging data, we obtain measurement data reflecting the brain functional activity status of each brain region of interest; Step 2: Based on the sparse inverse covariance matrix estimation, a symmetric positive definite sparse brain functional connection network set of each subject is obtained. The positive definite sparse brain functional connection network reflects the interaction relationship between the various brain regions of the subject; Step 3: Based on Riemannian manifold sparse coding, the dictionary of the positive definite sparse brain functional connectivity network and its sparse coefficient matrix in the kernel space are determined, and the component characteristics of the positive definite sparse brain functional connectivity network are analyzed in the manifold space; Step 4: Perform spatial distribution analysis of the brain functional connectivity atomic network to obtain the spatial distribution pattern of the brain functional connectivity subnetwork characteristics of each subject; Step five: longitudinal analysis of magnetic resonance imaging data of mild brain injury is conducted to identify disease status by analyzing the spatial distribution pattern differences of brain functional connection subnetwork characteristics between healthy subjects and diseased subjects.

2. The longitudinal analysis method of magnetic resonance imaging data of mild brain injury according to claim 1, characterized in that: The method for constructing the sparse brain functional connection network includes: Let {x1, x2, ..., x M } represents the BOLD signal time series of the brain region of interest with a length of M, where x i is a d-dimensional vector, corresponding to d brain regions of interest; x i Obey an unknown multidimensional Gaussian distribution Where μ is the mean vector of the multidimensional Gaussian distribution, ∑ is the covariance matrix of the multidimensional Gaussian distribution, both of which are unknown parameters; the brain functional connection network is obtained by the inverse covariance matrix ∑ of the BOLD signal time series. -1 Conduct estimation construction;∑ -1 The non-main diagonal elements represent a partial correlation between pairs of brain regions of interest after eliminating the influence of other brain regions of interest. represents the connection strength between the brain regions of interest i and j; if the brain regions of interest i and j are independent of each other, then is 0; the partial correlation between each pair of brain regions of interest is estimated: in, is the BOLD signal time series {x1, x2, ..., x M }; det(·) represents the matrix determinant operator, tr(·) represents the matrix trace operator, ||·||1 represents the sum of the absolute values ​​of all elements of the matrix, λ>0 is the preset sparse regularization parameter, Representing sparse brain functional connectivity networks The estimated value of Σ -1 Add l1 sparsity constraint, i.e. ||Σ -1 ||1, a more realistic brain functional connection network is estimated; when the sparse regularization parameter λ is relatively small, the Graphical Lasso method is used to solve equation (1) to obtain Used to represent sparse brain functional connectivity networks.

3. The longitudinal analysis method of magnetic resonance imaging data of mild brain injury according to claim 1, characterized in that: According to the sparse inverse covariance matrix estimation, a symmetric positive definite sparse brain functional connection network set of each subject is obtained, including: Sparse brain functional connectivity network training set for N subjects Among them S n represents the sparse brain functional connectivity network of the nth subject, estimated by the sparse inverse covariance matrix of the subject's functional magnetic resonance BOLD signal get, represents the d×d dimensional symmetric positive definite matrix space; Hypothetical dictionary Feature Map represents the mapping function from symmetric positive definite matrix space to Hilbert space; for any given sparse inverse covariance matrix S n , and obtain its sparse coding by solving the following formula: in, is the sparse coefficient vector, a i It is about dictionary atom D i The coefficient of Sparse brain functional connectivity network training set S={s1,s2,…,S n ,S n+1 ,…S N }By patients with mild brain damage G1 ={S1, ..., S n } and healthy control group S G2 ={S n+1 ,…,S N }Two groups of types; The group sparse coding method is used for dictionary learning, and its basic form is as follows: Among them, the sparse coefficient matrix A G1 ={A1, ..., A n } and A G2 ={A n+1 ,…,A N } represent the sparse coefficient matrices corresponding to the mild brain injury patient group and the healthy control group, respectively.

4. The longitudinal analysis method of magnetic resonance imaging data of mild brain injury according to claim 3, characterized in that: By selecting a suitable symmetric positive definite kernel with reproducing kernel properties, the kernel technique is used to simplify formula (3) into a kernel sparse coding problem on a Riemannian manifold with a convex optimization form. Solve this problem using the stochastic gradient algorithm or ADMM algorithm, and obtain the dictionary That is, it is a set of brain functional connection atomic network patterns, which represents a common set of brain functional connection atomic networks in the brain functional connection networks of all subjects including patients with mild brain injury and healthy control groups. It can map the spatial distribution of brain functional connection networks with characteristic descriptions in the brain space of all subjects.

5. The longitudinal analysis method of magnetic resonance imaging data of mild brain injury according to claim 1, characterized in that: The brain functional connectivity atomic network spatial distribution analysis includes: Construct a sparse brain functional connectivity network training set S = {S G1 , S G2 , S G3 }, using group sparse coding for dictionary learning: Among them, the sparse coefficient matrix A G1 、A G2 、A G3 They represent the sparse coefficient matrices corresponding to the acute patient group, subacute rehabilitation patient group, and healthy control group, respectively, reflecting the spatial distribution statistical information of the sparse brain functional connection networks of the three groups.

6. The longitudinal analysis method of magnetic resonance imaging data of mild brain injury according to claim 5, characterized in that: To A G1 、A G2 、A G3 Non-zero T-tests were performed separately to count the number of non-zero citations of each brain functional connection atomic network in each group, and to find the brain functional connection atomic network with a significantly high presentation rate in each group; by analyzing the differences in the spatial distribution of these brain functional connection atomic networks with high presentation rates, they were used to identify the brain functional connection characteristics in the three stages of rehabilitation treatment for mild brain injury: acute stage, subacute stage, and complete recovery, thereby realizing a longitudinal analysis of the rehabilitation process of mild brain injury.

7. A longitudinal analysis system for magnetic resonance imaging data of mild brain injury implementing the longitudinal analysis method for magnetic resonance imaging data of mild brain injury according to any one of claims 1 to 6, characterized in that: The longitudinal analysis system for mild brain injury magnetic resonance imaging data of mild brain injury comprises: A signal extraction module, for extracting BOLD signals from magnetic resonance imaging data of a subject; A network set determination module is used to construct a symmetric positive definite sparse brain functional connection network set of the subject based on the sparse inverse covariance matrix estimation; The network feature pattern mining module is used to determine the brain functional connection network dictionary and sparse coefficient matrix in the kernel space based on Riemannian manifold sparse coding; Network spatial distribution analysis module, used to analyze the spatial distribution of brain functional connectivity atomic networks; A disease status discrimination module for longitudinal analysis of magnetic resonance imaging data of mild brain injury.

8. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the longitudinal analysis method of magnetic resonance imaging data of mild brain injury as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the longitudinal analysis method of magnetic resonance imaging data of mild brain injury according to any one of claims 1 to 6.

10. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the longitudinal analysis system for mild brain injury magnetic resonance imaging data of mild brain injury as described in claim 7.

Images