Single sample network method and system for fMRI brain function connection analysis

Abstract

The invention provides a single-sample network method and system for fMRI brain function connection analysis, and belongs to the field of fMRI data community division algorithms, and the method comprises the steps: carrying out the preprocessing of an initial fMRI data set; performing brain region division on the preprocessed fMRI data set; obtaining a brain region average correlation matrix of the healthy control group, calculating a difference correlation matrix of each sample in the disease group relative to a sample of the healthy control group, and further obtaining a binary single sample correlation matrix of each sample in the disease group so as to construct a single sample network of each sample in the disease group; performing community division on the single-sample network to obtain a brain region co-community probability matrix of disease group samples; performing significance test on the brain region co-community probability matrix, and reserving brain region pairs of which the co-community probabilities are significantly greater than randomly allocated; and carrying out function enrichment on the reserved brain region pairs, and extracting remarkably-changed brain region connections related to diseases. According to the method, the influence of individual differences can be reduced, component differences can be reserved, and remarkably-changed brain region connections related to diseases can be accurately extracted.

Description

Technical Field

[0001] The present invention belongs to the field of fMRI data community partitioning algorithms, and in particular relates to a single sample network method and system for fMRI brain functional connection analysis. Background Art

[0002] The human brain is a complex structure composed of different regions. Each region, that is, each brain area, has its own mission and function. When studying the relationship between brain regions, there is a biological connection between brain regions, which is called structural connection. Functional magnetic resonance imaging (fMRI) refers to magnetic resonance brain function imaging based on blood oxygen level dependent (BOLD) signals. As early as the 1990s, researchers found that the activation of brain regions can be detected based on changes in BOLD signals. With the study of the correlation of BOLD signals between different brain regions, the degree of temporal correlation generated by the time series of BOLD signals in different brain regions is called "functional connectivity", which is functional connectivity that is different from structural connectivity. Functional connectivity is defined as the temporal dependence of neuronal activation patterns in anatomically separated brain regions, which attempts to establish a connection between two spatial regions of interest with the help of linear temporal correlation. fMRI is an effective method for locating various functional areas of the human brain in vivo, integrating function, imaging and anatomy. It has many advantages such as non-invasiveness, non-radioactivity, high temporal and spatial resolution, and multiple repeated operations. Therefore, in recent years, it has been widely used in the study of functional connections between brain regions and in revealing changes in functional connections related to the development of brain diseases and in the diagnosis of brain degenerative diseases.

[0003] In neuroscience, graph theory is used to establish mathematical models of complex network functions in the human brain. These networks are associated and connected between various regions and sub-regions of the brain, and the dynamic connections between networks constitute a larger single network. Graph theory methods mainly study the relationship between nodes and edges, and the calculation of correlation coefficients refers to calculating the correlation based on the BOLD signals between brain regions, converting resting-state fMRI into a set of functional networks of nodes and edges, so that graph theory methods can be applied to analyze brain functional networks. Calculating the correlation coefficient can analyze the interaction relationship between brain regions, and further express the changes in functional connectivity between brain regions through the numerical size and change of the correlation coefficient parameter, which is an important research object for fMRI data analysis. The functional connectivity between brain regions within the functional network of the brain is considered to be the functional basis of advanced cognitive functions and neurodegenerative diseases. Therefore, the calculation of the correlation coefficient can not only accurately reflect the functional association between brain regions, but also play an important role in the field of brain disease research.

[0004] Although the method of calculating the correlation coefficient provides an effective statistical tool for correlation analysis between two brain regions, it is not specifically designed for fMRI data. The fMRI image data directly collected has the characteristics of individual differences. In the data collection stage, although in the same experimental environment, the subjects will react differently to the stimulus of data collection, and there are differences between the patients' own physical constitutions, resulting in individual differences in the collected data. This makes the results obtained by directly applying the correlation coefficient method to fMRI data low in stability and low in recognition accuracy, and it is impossible to measure the changes in functional connectivity of brain regions. Therefore, a new method is needed that can reduce the impact of individual differences, retain inter-group differences, and accurately extract functional connectivity and related functional modules with significant changes related to the disease. Summary of the invention

[0005] The purpose of the present invention is to propose a single-sample network method and system for fMRI brain functional connectivity analysis, which can reduce the impact of individual differences, retain component differences, and accurately extract brain region connections with significant changes related to diseases.

[0006] The present invention is achieved through the following technical solutions:

[0007] A single sample network method for fMRI brain functional connectivity analysis, comprising the following steps:

[0008] Step S1, preprocessing the initial fMRI data set to obtain a noise-reduced and spatially standardized fMRI data set, wherein the preprocessed fMRI data set includes healthy control group samples and disease group samples;

[0009] Step S2, dividing the preprocessed fMRI data set into brain regions according to a standardized brain region atlas, and obtaining multiple brain regions and BOLD signals of each brain region;

[0010] Step S3, obtaining the average correlation matrix of the brain regions of the healthy control group according to the BOLD signals of each brain region corresponding to the samples of the healthy control group, and using the average correlation matrix of the brain regions of the healthy control group to calculate the difference correlation matrix of each sample in the disease group relative to the healthy control group samples, and then obtaining the binary single sample association matrix of each sample in the disease group, taking each brain region as a node and the binary single sample association matrix as an adjacency matrix, constructing a single sample network of each sample in the disease group;

[0011] Step S4, performing community division on the single sample network, and calculating the brain region common community probability matrix of the disease group samples;

[0012] Step S5, performing a significance test on the brain region common community probability matrix, and retaining brain region pairs with common community probabilities significantly greater than random assignment;

[0013] Step S6: Perform functional enrichment on the brain region pairs retained in step S5 to extract brain region connections with significant changes related to the disease.

[0014] Furthermore, in step S1, the preprocessing of the initial fMRI data set includes temporal layer correction, head motion correction, image registration, segmentation, standard space mapping and denoising.

[0015] Furthermore, in step S2, the fetch_atlas_aal function in the datasets package of the Python script is used to divide the preprocessed fMRI dataset into brain regions, and the number of brain regions divided for each sample in the preprocessed fMRI dataset is N, where N is an integer.

[0016] Furthermore, the step S3 specifically includes the following steps:

[0017] Step S31: According to the formula Get the average brain region correlation matrix of H samples in the healthy control group in, for Elements of represents the BOLD signal of any two brain regions in the hth sample of the healthy control group and The Pearson correlation coefficient between them, H is the number of samples in the healthy control group,

[0018] Step S32: According to the formula Obtain the difference correlation matrix D of the kth sample in the disease group relative to the healthy control group sample k ,in, Represents the BOLD signal of any two brain regions in the kth sample of the disease group and The Pearson correlation coefficient between D k Elements of

[0019] Step S33: According to the formula The difference correlation matrix D k Perform z-transformation to obtain the weight matrix of the kth sample in the disease group, and then use the formula Calculate the binary single sample association matrix of the kth sample in the disease group, where T is the given threshold;

[0020] Step S34: Take each brain region as a node and the binary single sample association matrix as an adjacency matrix to construct a single sample network G = (V, E k ), where V = {vi , i = 1, 2, ..., N} is the brain region node set of the kth sample in the disease group, is the brain region association edge set of the kth sample in the disease group.

[0021] Furthermore, the step S4 specifically includes the following steps:

[0022] Step S41, calculate the initial modularity of the single sample network G according to each independent community in the single sample network G of the kth sample of the disease group, and calculate the combined modularity of any community in the sample network G and each community connected to the community, wherein the initial modularity and the combined modularity are calculated according to the formula Calculation, ∑in represents the community C i The sum of the weights of the edges in community C, ∑tot represents i The sum of the weights of all edges connecting the nodes in G, m represents the number of edges in the single sample network G, and C represents the set of communities participating in the calculation;

[0023] Step S42: Calculate the changes of the initial modularity and each merged modularity respectively, and select the community C with the largest modularity change as j With Community C i Merge into a new community and update the single sample network G;

[0024] Step S43: Repeat steps S41 to S42 to continuously divide and merge communities until all brain region nodes v i The community to which the user belongs no longer changes;

[0025] Step S44: According to the formula Calculate any two brain region nodes v of the kth sample in the disease group i and v j The co-community probability matrix P k , where f is the number of community divisions and p is the number of total communities.

[0026] Furthermore, in step S1, the disease group samples are classified according to the severity of the disease, and for each severity, they are grouped according to age to obtain multiple sample groups.

[0027] Further, in step S5, according to the formula A significance test was performed on the common community probability matrix, and the brain region pairs with common community probabilities significantly greater than those randomly assigned were retained. t represents the probability of the observed sample statistic or more extreme values ​​appearing when no significant changes occur, and X1 and X2 represent the same pair of brain regions in the sample group with two different severity levels of the same age group. ij The set of common community probability values, and represent the means of X1 and X2 respectively, represents the variance of the joint estimate of X1 and X2, and They represent the variance of X1 and X2 respectively, and n1 and n2 represent the sample sizes of X1 and X2 respectively.

[0028] Furthermore, in step S6, Fisher's exact test is performed on the brain region pairs retained in step S5 to extract brain region connections with significant changes related to the disease.

[0029] The present invention is also achieved through the following technical solutions:

[0030] A single-sample network system for fMRI brain functional connectivity analysis, comprising:

[0031] Preprocessing module: used to preprocess the initial fMRI dataset to obtain a noise-reduced and spatially standardized fMRI dataset. The preprocessed fMRI dataset contains samples from the healthy control group and samples from the disease group.

[0032] BOLD signal analysis module: used to divide the preprocessed fMRI data set into brain regions according to the standardized brain region map, and obtain multiple brain regions and the BOLD signals of each brain region;

[0033] Single sample network acquisition module: used to obtain the average correlation matrix of the brain regions of the healthy control group according to the BOLD signals of each brain region corresponding to the samples of the healthy control group, and use the average correlation matrix of the brain regions of the healthy control group to calculate the difference correlation matrix of each sample in the disease group relative to the healthy control group, and then obtain the binary single sample association matrix of each sample in the disease group, and use each brain region as a node and the binary single sample association matrix as an adjacency matrix to construct a single sample network of each sample in the disease group;

[0034] Common community probability matrix acquisition module: used to divide the single sample network into communities and calculate the brain region common community probability matrix of the disease group samples;

[0035] Data verification module: used to perform significance test on the brain region common community probability matrix, and retain the brain region pairs with common community probability significantly greater than random assignment;

[0036] Functional connectivity analysis module: used to perform functional enrichment on the brain region pairs retained in step S5 and extract brain region connections with significant changes related to the disease.

[0037] The present invention has the following beneficial effects:

[0038] 1. The present invention first preprocesses the initial fMRI data set to obtain a preprocessed fMRI data set including healthy control group samples and disease group samples, and divides the preprocessed fMRI data set into brain regions according to the standardized brain region atlas to obtain multiple brain regions and BOLD signals of each brain region. Next, the average correlation matrix of the brain regions of the healthy control group is obtained according to the BOLD signals of each brain region corresponding to the healthy control group samples, and the average correlation matrix of the brain regions of the healthy control group is obtained to calculate the difference correlation matrix of each sample in the disease group relative to the healthy control group samples, thereby obtaining a binary single sample association matrix of each sample in the disease group, taking each brain region as a node and the binary single sample association matrix as an adjacency matrix, constructing a single sample network of each sample in the disease group, and then dividing the single sample network into communities, calculating the brain region common community probability matrix of the disease group samples, and performing a significance test on the brain region common community probability matrix, retaining brain region pairs with a common community probability significantly greater than that of random assignment, and finally performing functional enrichment on the retained brain region pairs to extract brain region connections with significant changes related to the disease. In the process, the difference between the disease sample and the healthy control group in the fMRI data is defined as a constraint term, and a weight matrix is ​​established through mathematical transformation. The healthy control group is used as a reference. This approach fully reduces the influence of individual differences and retains the differences between groups. This single-sample network calculation method improves the stability of the sample brain network, making the subsequent community division processing results more reliable. In the present invention, the parameter of common community probability is used instead of the traditional correlation coefficient to measure the functional connectivity of brain regions, which improves the accuracy of analyzing changes in functional connectivity of brain regions, so that the subsequent significance test and functional enrichment of the common community probability matrix can fully extract the significantly changed brain region connections related to the disease and related functional modules. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The present invention is further described in detail below with reference to the accompanying drawings.

[0040] Figure 1 It is a flow chart of the present invention.

[0041] Figure 2 It is the modularity change of the single sample network of the present invention compared with the correlation coefficient network.

[0042] Figure 3 It is the change of participation coefficient of the single sample network of the present invention compared with the correlation coefficient network.

[0043] Figure 4 This is the community probability matrix of each stage of the disease in different age groups obtained by the present invention.

[0044] Figure 5 This is the distribution of the connecting edges with significant decrease in the common community probability in different functional areas of the disease groups in different age groups as the symptoms worsen (EMCI-LMCI-AD) in the present invention. DETAILED DESCRIPTION

[0045] like Figure 1 As shown in FIG. 1 , the single-sample network method for fMRI brain functional connectivity analysis includes the following steps:

[0046] Step S1, preprocessing the initial fMRI data set to obtain a noise-reduced and spatially standardized fMRI data set, wherein the preprocessed fMRI data set includes healthy control group samples and disease group samples;

[0047] Specifically, this embodiment uses Alzheimer's disease (AD) as a research case, and the initial fMRI data set comes from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. In order to study the changes in brain functional connectivity between different age groups from early mild cognitive impairment (EMCI) to late mild cognitive impairment (LMCI) to Alzheimer's disease (AD), a total of 182 fMRI subject data samples were selected, including 20 samples of the healthy control group (CN) and 162 samples of the disease group. The disease group samples were classified according to the severity of the disease, namely 57 cases of EMCI, 55 cases of LMCI and 50 cases of AD. For the three severity levels, they were grouped according to age groups, namely, 60-year-old group (age between 60-69), 70-year-old group (age between 70-79), and 80-year-old group (age between 80-89), thereby obtaining multiple sample groups.

[0048] In this embodiment, the initial fMRI data set is preprocessed using Matlab's toolbox SPM12, specifically including:

[0049] Step S11, time layer correction, remove the files with distorted imaging, reset the origin of each fMRI image after retention, and use top-down slice time correction to correct the inconsistency of slice contrast time;

[0050] Step S12, head motion correction, correcting the motion artifacts caused by the head movement of the sample participants in the process of acquiring images in six directions: up and down, left and right, and front and back;

[0051] Step S13, image registration, aligning the weighted structural image and the functional image, wherein the alignment process is the prior art;

[0052] Step S14: segmentation, segmenting the structural image registered with the functional image into gray matter, white matter, cerebrospinal fluid, bone and others;

[0053] Step S15: standard space mapping, using affine transformation to map the weighted structural image and functional image to the MNI standard space;

[0054] Step S16: De-noising: spatial smoothing is performed using a Gaussian kernel to increase the signal-to-noise ratio.

[0055] In this embodiment, the final preprocessed fMRI sample data matrix X is obtained. 91×109×91×140 In this embodiment, the fMRI data is collected by collecting a whole-brain fMRI image every 3 seconds, so 140 represents the number of time points for data collection, and 91×109×91 represents the size and position of the scanned brain volume.

[0056] Step S2, dividing the preprocessed fMRI data set into brain regions according to a standardized brain region atlas, and obtaining multiple brain regions and BOLD signals of each brain region;

[0057] In this embodiment, the fetch_atlas_aal function in the Python script datasets package is used to fetch the preprocessed fMRI sample data matrix X 91×109×91×140 Perform brain region division and obtain the BOLD signal time series matrix T 90×140 The number of brain regions divided by each sample in the preprocessed fMRI data set is N = 90. The BOLD signal of each brain region is obtained by scanning at 140 time points. The time series matrix T 90×140 Each row represents the BOLD signal of a brain region.

[0058] Step S3, obtaining the average correlation matrix of the brain regions of the healthy control group according to the BOLD signals of each brain region corresponding to the samples of the healthy control group, and using the average correlation matrix of the brain regions of the healthy control group to calculate the difference correlation matrix of each sample in the disease group relative to the healthy control group samples, and then obtaining the binary single sample association matrix of each sample in the disease group, taking each brain region as a node and the binary single sample association matrix as an adjacency matrix, constructing a single sample network of each sample in the disease group;

[0059] The specific steps include:

[0060] Step S31: According to the formula Get the average brain region correlation matrix of H samples in the healthy control group in, for Elements of Indicates the BOLD signal of any two brain regions in the hth sample of the healthy control group and The Pearson correlation coefficient between them, H is the number of samples in the healthy control group,

[0061] Step S32: According to the formula Obtain the difference correlation matrix D between the kth sample of the disease group and the healthy control group sample k ,in, Indicates the BOLD signal of any two brain regions in the kth sample in the disease group and The Pearson correlation coefficient between D k Elements of

[0062] Step S33: According to the formula The difference correlation matrix D k Perform z-transformation to obtain the weight matrix of the kth sample in the disease group, and then use the formula Calculate the binary single sample association matrix E of the kth sample in the disease group k , where T is a given threshold;

[0063] Step S34: Take each brain region as a node and the binary single sample association matrix as an adjacency matrix to construct a single sample network G = (V, E k ), where V = {v i , i = 1, 2, ..., N} is the brain region node set of the kth sample in the disease group, and each element v i is called a node of the graph. is the brain region associated edge set of the kth sample in the disease group, each element is a line from v inside the sample association network G i to v j edge.

[0064] Step S4, performing community division on the single sample network, and calculating the brain region common community probability matrix of the disease group samples;

[0065] The specific steps include:

[0066] Step S41: Calculate the initial modularity of the single sample network G according to each independent community in the single sample network G of the kth sample of the disease group, and calculate the initial modularity of any community C in the sample network G respectively. i and with the community C i The combined modularity of the connected communities after merging, where the initial modularity and the combined modularity are calculated according to the formula Calculation, ∑in represents the community C i The sum of the weights of the edges in community C, ∑tot represents i The sum of the weights of all edges connecting the nodes in the network, m represents the number of edges in the single sample network G, that is, the binary single sample association matrix E k The number of non-zero elements in , C represents the set of communities participating in the calculation;

[0067] Step S42: Calculate the changes of the initial modularity and each merged modularity respectively, and select the community C with the largest modularity change as j With Community C i Merge into a new community and update the single sample network G;

[0068] Step S43: Repeat steps S41 to S42 to continuously divide and merge communities until all brain region nodes v i The community to which no longer changes, thus finding a more closely connected sub-network from the entire single sample network G;

[0069] More specifically, when calculating the initial modularity for the first time, the single sample network G of the kth sample in the disease group is divided into 90 communities C i , C i Corresponds to a node v in V i , at this time according to 90 communities C i Calculate the initial modularity. For any community C i , merge all communities connected to this community with this community, and then calculate the merged modularity corresponding to each merge; the single sample network G after one merge can be divided into 89 communities, and then calculate the initial modularity again based on the 89 communities, and so on, until all brain region nodes v i The community to which the community belongs will no longer change, that is, all communities will be merged into a new community;

[0070] When calculating the initial modularity, community C i Refers to the communities that have not been merged. When calculating the merged modularity, it is the same as community C i Connecting Community A and Community C i The community formed after the merger is also called community C i express;

[0071] Step S44: According to the formula Calculate any two brain region nodes v of the kth sample in the disease group i and v j The co-community probability matrix P k , where f is the number of community divisions, p is the number of co-communities,

[0072] Step S5, performing a significance test on the brain region common community probability matrix, and retaining brain region pairs with common community probabilities significantly greater than random assignment;

[0073] Specifically, according to the formula A significance test was performed on the common community probability matrix, and the brain region pairs with common community probabilities significantly greater than those randomly assigned (i.e., brain region pairs when t < 0.05) were retained. Among them, t represents the probability of the observed sample statistic or more extreme values ​​appearing when no significant change occurs (i.e., it is set to true). A lower t value means that the observed result is very different from the expected result under the null hypothesis. Usually, if the t value is lower than the predetermined significance level (such as 0.05), the null hypothesis (i.e., a significant change has occurred) will be rejected. X1 and X2 represent the same pair of brain region pairs in the sample group with two different severity levels of the same age group. ij The set of common community probability values, and represent the means of X1 and X2 respectively, represents the variance of the joint estimate of X1 and X2, and They represent the variance of X1 and X2 respectively, and n1 and n2 represent the sample sizes of X1 and X2 respectively.

[0074] In this embodiment, all samples were divided into 9 different sample groups, and the common community probability matrix calculated for the sample groups of the same age group with different diseases was subjected to inter-group significance test. The brain region matrix P with significantly reduced community probability in the three age groups was obtained. 60 , P 70 , P 80 .

[0075] Step S6, performing functional enrichment on the brain region pairs retained in step S5, and extracting brain region connections with significant changes related to the disease;

[0076] Fisher's exact test is performed on the brain region pairs retained in step S5 to extract the brain region connections with significant changes related to the disease. The probability of common communities in this embodiment significantly reduces the brain region pair matrix P 60 , P 70 , P 80 Functional enrichment was performed into seven functional connection networks of the brain, namely, visual network, motor sensory network, dorsal attention network, ventral attention network, limbic network, frontal-parietal network, and default mode network. Functional areas with p≤0.05 were retained as enrichment results.

[0077] Correspondingly, the single-sample network system for fMRI brain functional connectivity analysis includes:

[0078] Preprocessing module: used to preprocess the initial fMRI dataset to obtain a noise-reduced and spatially standardized fMRI dataset. The preprocessed fMRI dataset contains samples from the healthy control group and samples from the disease group.

[0079] BOLD signal analysis module: used to divide the preprocessed fMRI data set into brain regions according to the standardized brain region map, and obtain multiple brain regions and the BOLD signals of each brain region;

[0080] Single sample network acquisition module: used to obtain the average correlation matrix of the brain regions of the healthy control group according to the BOLD signals of each brain region corresponding to the samples of the healthy control group, and use the average correlation matrix of the brain regions of the healthy control group to calculate the difference correlation matrix of each sample in the disease group relative to the healthy control group, and then obtain the binary single sample association matrix of each sample in the disease group, and use each brain region as a node and the binary single sample association matrix as an adjacency matrix to construct a single sample network of each sample in the disease group;

[0081] Common community probability matrix acquisition module: used to divide the single sample network into communities and calculate the brain region common community probability matrix of the disease group samples;

[0082] Data verification module: used to perform significance test on the brain region common community probability matrix, and retain the brain region pairs with common community probability significantly greater than random assignment;

[0083] Functional connectivity analysis module: used to perform functional enrichment on the brain region pairs retained in step S5 and extract brain region connections with significant changes related to the disease.

[0084] Figure 2 and Figure 3 It shows that the average modularity (Modularity Q) of the single sample network (Individual Metabolic Brain Network) after community division is improved by about 70% and the average participation coefficient (Participation Coefficient P) is significantly reduced by 50% compared with the traditional calculation of correlation coefficient (Pearson Brain Network) method, which has obvious performance superiority. Figure 4 and Figure 5 The results of the single-sample network after the subsequent calculation of the co-community probability matrix, the significance test and the functional enrichment are shown. Figure 4 It can be seen that there are differences in the distribution of strong functional connections between different age groups and different disease severity. Figure 5It can be seen that through the t-test method, in the same age group, different disease groups have different connections (p≤0.05) with significantly reduced common community probability in the seven functional networks as the severity of the disease deepens (EMCI-LMCI-AD). After enrichment, the functional networks affected by diseases in different age groups can be quickly located. In the 60-year-old group, the significantly reduced functional connection edges are enriched in the motor sensory network (SMN) and the frontal-parietal network (FN); in the 70-year-old group, the significantly reduced functional connection edges are enriched in the dorsal attention network (DAN); in the 80-year-old group, the significantly reduced functional connection edges are enriched in the visual network (VIS) and the limbic network (LN); according to the analysis results after the test, in different age groups, by paying attention to the enrichment of different functional areas in the diagnosis, the patient's disease state can be accurately located and the diagnosis of Alzheimer's disease can be achieved.

[0085] In summary, the present invention improves the traditional method of calculating correlation coefficients. The results show that compared with the traditional method of calculating correlation coefficients, the embodiments of the present invention fully reduce the impact of individual differences on fMRI data, retain the differences between groups, and improve the stability of the sample brain network, making the subsequent community division processing results more reliable. The present invention also provides a new method for defining the functional correlation between brain regions through the common community probability, replacing the traditional correlation coefficient to measure the correlation, and improving the accuracy of analyzing changes in functional connectivity between brain regions, so that subsequent significance tests and differential function enrichment can fully extract brain region connections with significant changes related to the disease. Further promote the application of fMRI data in brain neurodegenerative diseases.

[0086] It is worth mentioning that the present invention is not only applicable to the analysis of Alzheimer's disease patient data, but also applicable to other disease patient data, such as Parkinson's disease, etc., which is not limited in this embodiment.

[0087] The above description is only a preferred embodiment of the present invention, and therefore cannot be used to limit the scope of implementation of the present invention. That is, equivalent changes and modifications made according to the scope of the patent application of the present invention and the contents of the specification should still fall within the scope covered by the patent of the present invention.

Claims

1. A single-sample network method for fMRI brain functional connectivity analysis, characterized by: The steps include: Step S1, preprocessing the initial fMRI data set to obtain a noise-reduced and spatially standardized fMRI data set, wherein the preprocessed fMRI data set includes healthy control group samples and disease group samples; Step S2, dividing the preprocessed fMRI data set into brain regions according to a standardized brain region atlas, and obtaining multiple brain regions and BOLD signals of each brain region; Step S3, obtaining the average correlation matrix of the brain regions of the healthy control group according to the BOLD signals of each brain region corresponding to the samples of the healthy control group, and using the average correlation matrix of the brain regions of the healthy control group to calculate the difference correlation matrix of each sample in the disease group relative to the healthy control group samples, and then obtaining the binary single sample association matrix of each sample in the disease group, taking each brain region as a node and the binary single sample association matrix as an adjacency matrix, constructing a single sample network of each sample in the disease group; Step S4, performing community division on the single sample network, and calculating the brain region common community probability matrix of the disease group samples; Step S5, performing a significance test on the brain region common community probability matrix, and retaining brain region pairs with common community probabilities significantly greater than random assignment; Step S6: Perform functional enrichment on the brain region pairs retained in step S5 to extract brain region connections with significant changes related to the disease.

2. A single sample network method for fMRI brain functional connectivity analysis according to claim 1, characterized in that: In step S1, the preprocessing of the initial fMRI data set includes temporal layer correction, head motion correction, image registration, segmentation, standard space mapping and denoising.

3. The single sample network method for fMRI brain functional connectivity analysis according to claim 1, characterized in that: In step S2, the fetch_atlas_aal function in the datasets package of the Python script is used to divide the preprocessed fMRI dataset into brain regions, and the number of brain regions divided for each sample in the preprocessed fMRI dataset is N, where N is an integer.

4. A single sample network method for fMRI brain functional connectivity analysis according to claim 3, characterized in that: The step S3 specifically includes the following steps: Step S31: According to the formula Get the average brain region correlation matrix of H samples in the healthy control group in, for Elements of represents the BOLD signal of any two brain regions in the hth sample of the healthy control group and The Pearson correlation coefficient between them, H is the number of samples in the healthy control group, Step S32: According to the formula Get the difference correlation matrix D of the kth sample in the disease group relative to the healthy control group sample k ,in, Represents the BOLD signal of any two brain regions in the kth sample of the disease group and The Pearson correlation coefficient between D k Elements of Step S33: According to the formula The difference correlation matrix D k Perform z-transformation to obtain the weight matrix of the kth sample in the disease group, and then use the formula Calculate the binary single sample association matrix of the kth sample in the disease group, where T is the given threshold; Step S34: Take each brain region as a node and the binary single sample association matrix as an adjacency matrix to construct a single sample network G = (V, E k ), where V = {v i , i = 1, 2, ..., N} is the brain region node set of the kth sample in the disease group, is the brain region association edge set of the kth sample in the disease group.

5. A single sample network method for fMRI brain functional connectivity analysis according to claim 1, 2 or 3, characterized in that: The step S4 specifically includes the following steps: Step S41, calculate the initial modularity of the single sample network G according to each independent community in the single sample network G of the kth sample of the disease group, and calculate the combined modularity of any community in the sample network G and each community connected to the community, wherein the initial modularity and the combined modularity are calculated according to the formula Calculation, ∑in represents the community C i The sum of the weights of the edges in community C, ∑tot represents i The sum of the weights of all edges connecting the nodes in G, m represents the number of edges in the single sample network G, and C represents the set of communities participating in the calculation; Step S42: Calculate the changes of the initial modularity and each merged modularity respectively, and select the community C with the largest modularity change as j With Community C i Merge into a new community and update the single sample network G; Step S43: Repeat steps S41 to S42 to continuously divide and merge communities until all brain region nodes v i The community to which the user belongs no longer changes; Step S44: According to the formula Calculate any two brain region nodes v of the kth sample in the disease group i and v j The co-community probability matrix P k , where f is the number of community divisions and p is the number of total communities.

6. A single sample network method for fMRI brain functional connectivity analysis according to claim 1, 2 or 3, characterized in that: In step S1, the disease group samples are classified according to the severity of the disease, and for each severity, they are grouped according to age to obtain multiple sample groups.

7. A single sample network method for fMRI brain functional connectivity analysis according to claim 1, 2 or 3, characterized in that: In step S5, according to the formula A significance test was performed on the common community probability matrix, and the brain region pairs with common community probabilities significantly greater than those randomly assigned were retained. t represents the probability of the observed sample statistic or more extreme values ​​appearing when no significant changes occur, and X1 and X2 represent the same pair of brain regions in the sample group with two different severity levels of the same age group. ij The set of common community probability values, and represent the means of X1 and X2 respectively, represents the variance of the joint estimate of X1 and X2, and They represent the variance of X1 and X2 respectively, and n1 and n2 represent the sample sizes of X1 and X2 respectively.

8. A single sample network method for fMRI brain functional connectivity analysis according to claim 1, 2 or 3, characterized in that: In step S6, Fisher's exact test is performed on the brain region pairs retained in step S5 to extract brain region connections with significant changes related to the disease.

9. A single-sample network system for fMRI brain functional connectivity analysis, characterized by: include: Preprocessing module: used to preprocess the initial fMRI dataset to obtain a noise-reduced and spatially standardized fMRI dataset. The preprocessed fMRI dataset contains samples from the healthy control group and samples from the disease group. BOLD signal analysis module: used to divide the preprocessed fMRI data set into brain regions according to the standardized brain region map, and obtain multiple brain regions and the BOLD signals of each brain region; Single sample network acquisition module: used to obtain the average correlation matrix of the brain regions of the healthy control group according to the BOLD signals of each brain region corresponding to the samples of the healthy control group, and use the average correlation matrix of the brain regions of the healthy control group to calculate the difference correlation matrix of each sample in the disease group relative to the healthy control group, and then obtain the binary single sample association matrix of each sample in the disease group, and use each brain region as a node and the binary single sample association matrix as an adjacency matrix to construct a single sample network of each sample in the disease group; Common community probability matrix acquisition module: used to divide the single sample network into communities and calculate the brain region common community probability matrix of the disease group samples; Data verification module: used to perform significance test on the brain region common community probability matrix, and retain the brain region pairs with common community probability significantly greater than random assignment; Functional connectivity analysis module: used to perform functional enrichment on the brain region pairs retained in step S5 and extract brain region connections with significant changes related to the disease.

Images