Graph theory analysis method and device for mild cognitive impairment dynamic functional connectivity

Abstract

The present application relates to the technical field of neuroimaging processing, in particular to a graph theory analysis method for mild cognitive impairment dynamic functional connectivity. The method comprises the following steps: extracting BOLD signals corresponding to the region of interest in the rs-fMRI data of the subject; performing sliding windowing on the BOLD signals and performing first correlation analysis to obtain dynamic functional connectivity information between each region of interest; performing second correlation analysis on the optimal feature subset selected from the dynamic functional connectivity information and extracting effective dynamic functional connectivity information; performing thresholding processing on the effective dynamic functional connectivity information and constructing a network; and extracting graph features of the network by using a graph theory method to analyze differences and classification. The purpose of the graph theory analysis method for mild cognitive impairment dynamic functional connectivity is to solve the problem of low data classification accuracy caused by the lack of time domain correlation analysis in the traditional graph theory analysis method.

Description

Technical Field

[0001] This invention relates to the field of neuroimaging processing technology, specifically to a graph theory analysis method and apparatus for dynamic functional connectivity in mild cognitive impairment. Background Technology

[0002] Alzheimer's disease (AD) is a degenerative disease of the central nervous system induced by various factors. The main pathological feature of AD is the loss of nerves in the hippocampus and cortex, clinically manifesting as cognitive and memory impairments, as well as personality and behavioral abnormalities. AD primarily affects the elderly, severely impacting their quality of life, and patients typically die within 10 years of diagnosis. The causes of AD are currently unclear, and there are no drugs that can stop or reverse its progression; only a few methods can temporarily alleviate symptoms. Mild cognitive impairment (MCI) refers to the transitional stage between normal aging-related cognitive decline and AD, not reaching the level of cognitive decline seen in AD, and is considered a prodromal period of AD. However, research indicates that MCI can potentially progress to AD, and the conversion rate is increasing annually, with more than half of MCI patients developing AD within 5 years. Therefore, early diagnosis of MCI is crucial for AD intervention, and this protocol is primarily used for the auxiliary diagnosis of MCI patients.

[0003] Current auxiliary diagnostic methods typically employ rest-state functional magnetic resonance imaging (rs-fMRI) to extract changes in blood oxygen level dependent (BOLD) signals in different brain regions, indirectly reflecting the function of local brain regions and neural networks. Functional connectivity (FC) calculated using BOLD signals can measure the synergistic relationships of functions in different brain regions, allowing for the study of differences between cognitive impairment (MCI) and normal cognition (NC). Recently, researchers have proposed dynamic functional connectivity (DFC) analysis. DFC can capture changes in intrinsic FC in the brain under various physiological states and is a more sensitive marker than static FC. In the field of neuroimaging, graph theory is often used to build mathematical models of complex network functions in the human brain. The nodes and edges in the model can reflect the connections between different brain regions. Existing studies have applied graph theory to FC analysis and shown that the topological organization of the whole-brain functional network in patients with cognitive impairment is abnormal, including the loss of some connectivity structures and the redistribution of core brain regions. Research on graph theory in the field of DFC (Digital Fractional Analytical) is not extensive. Existing methods mainly include dimensionality reduction and slicing. Dimensionality reduction reduces three-dimensional DFC data to two dimensions and then analyzes it based on the construction of brain networks. Slicing slices the time dimension of DFC data, constructs brain networks at each time point, and analyzes the differences between groups by extracting the dynamic changes of features. Finally, by distinguishing the abnormal features of the disease group from the normal group, it completes the auxiliary diagnosis of patients with the disease.

[0004] Traditional graph theory analysis methods neglect the dynamic nature of subjects' functional connectivity and the consistency of temporal functional connectivity. Different construction methods can lead to varying topological properties in the human brain network when defining network nodes and measuring network connectivity, resulting in a lack of unified standards for analysis. Existing methods, such as dimensionality reduction, obscure temporal information, have high requirements for the dimensionality reduction method itself, and the biological significance after dimensionality reduction remains unclear. Slicing methods are complex, difficult to classify, and prone to information asymmetry.

[0005] Therefore, the inventors provide a graph theory analysis method and apparatus for dynamic functional connectivity in mild cognitive impairment. Summary of the Invention

[0006] (1) Technical problems to be solved

[0007] This invention provides a graph theory analysis method and apparatus for dynamic functional connectivity in mild cognitive impairment, which solves the technical problem of low data classification accuracy caused by the lack of temporal correlation analysis in traditional graph theory analysis methods.

[0008] (2) Technical solution

[0009] The first aspect of this invention provides a graph theory analysis method for dynamic functional connectivity in mild cognitive impairment, comprising the following steps:

[0010] Extract the BOLD signal corresponding to the region of interest from the subject's rs-fMRI data;

[0011] The BOLD signal is windowed and a first correlation analysis is performed to obtain the dynamic functional connectivity information between each region of interest.

[0012] A second correlation analysis is performed on the optimal feature subset selected from the dynamic functional connection information to extract effective dynamic functional connection information;

[0013] The effective dynamic function connection information is thresholded and a network is constructed.

[0014] Graph theory methods are used to extract the graph features of the network for analysis and classification.

[0015] Furthermore, the step of sliding windowing the BOLD signal and performing the first correlation analysis to obtain the dynamic functional connectivity information between each region of interest specifically includes the following steps:

[0016] The BOLD signal of each extracted region of interest is divided into multiple windows by sliding segmentation using a rectangular window with a set step size.

[0017] The Pearson correlation coefficient of the BOLD signal between two adjacent regions of interest within each window is calculated to obtain the dynamic functional connection information.

[0018] Furthermore, the step of performing a second correlation analysis on the optimal feature subset selected from the dynamic functional connectivity information and extracting effective dynamic functional connectivity information specifically includes the following steps:

[0019] Extract the trace-free upper triangular elements from the correlation coefficient matrix of the w-th window, and determine the expansion vector of the upper triangular elements;

[0020] Based on the expansion vector, determine the dynamic functional connection information between two adjacent windows;

[0021] The optimal feature subset is selected based on the variance of the dynamic functional connection information between two adjacent windows.

[0022] The effective dynamic functional connection information between two adjacent windows is calculated using the optimal feature subset.

[0023] Further, the step of extracting the trace-free upper triangular elements from the correlation coefficient matrix of the w-th window and determining the expansion vector of the upper triangular elements specifically includes the following steps:

[0024] The correlation coefficient matrix of the w-th window is determined based on the Pearson correlation coefficient of the BOLD signal between two adjacent regions of interest within each window.

[0025] Extract the upper triangular elements of the correlation coefficient matrix of the w-th window, removing the trace.

[0026] Determine the expansion vector of the upper triangular elements.

[0027] Further, determining the dynamic functional connection information between two adjacent windows based on the expansion vector specifically involves: determining the dynamic functional connection information between the two adjacent windows based on the first expansion vector and the second expansion vector corresponding to the w1-th window and the w2-th window, as well as the mean values ​​corresponding to the first expansion vector and the second expansion vector, respectively.

[0028] Furthermore, the step of selecting the optimal feature subset based on the variance of the dynamic functional connection information between two adjacent windows specifically involves selecting the optimal feature subset from the dynamic functional connection information based on the minimum intra-class distance of the variance of the dynamic functional connection information between two adjacent windows.

[0029] Furthermore, the step of thresholding the effective dynamic functional connection information and constructing the network specifically involves: thresholding the matrix corresponding to the effective dynamic functional connection information to obtain an adjacency matrix in graph theory that represents the connection relationship of the network.

[0030] A second aspect of the present invention provides a graph theory analysis apparatus for dynamic functional connectivity in mild cognitive impairment, comprising:

[0031] The signal extraction module is used to extract the BOLD signal corresponding to the region of interest in the subject's rs-fMRI data;

[0032] The first correlation module is used to slide window the BOLD signal and perform the first correlation analysis to obtain the dynamic functional connection information between each region of interest.

[0033] The second correlation module is used to perform a second correlation analysis on the optimal feature subset selected from the dynamic functional connection information and extract effective dynamic functional connection information.

[0034] The network construction module is used to perform thresholding processing on the effective dynamic functional connection information and construct the network;

[0035] The graph theory analysis module is used to extract the graph features of the network using graph theory methods to analyze differences and classify them.

[0036] A third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, performs the steps of the method as described in any of the preceding claims.

[0037] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method as described in any of the preceding claims.

[0038] (3) Beneficial effects

[0039] In summary, this invention extracts dynamic functional connectivity information through a quadratic correlation method and reduces feature dimensionality and enhances inter-group differences by selecting the optimal feature subset based on the DFC stationarity criterion. It employs a thresholding method to construct a dynamic brain functional network and utilizes the relief algorithm to achieve adaptive threshold selection. Furthermore, it combines graph theory analysis to extract the network's graphical features and explores the significant differences between MCI and NC under different network construction methods and indices. This addresses the lack of temporal correlation analysis in traditional graph theory analysis methods, considering both the correlation between different ROIs and the correlation between different time windows. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart illustrating a graph theory analysis method for dynamic functional connectivity in mild cognitive impairment, provided by an embodiment of the present invention.

[0042] Figure 2 This is a structural block diagram of a graph theory analysis device for dynamic functional connectivity in mild cognitive impairment, provided in an embodiment of the present invention.

[0043] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0044] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. The following detailed description of the embodiments and the accompanying drawings are used to illustrate the principles of the present invention by way of example, but should not be used to limit the scope of the present invention. That is, the present invention is not limited to the described embodiments, and any modifications, substitutions and improvements to the parts, components and connection methods are covered without departing from the spirit of the present invention.

[0045] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0046] Figure 1 This is a flowchart illustrating a graph theory analysis method for dynamic functional connectivity in mild cognitive impairment, provided by an embodiment of the present invention. Figure 1 As shown, the method may include the following steps:

[0047] S101. Extract the BOLD signal corresponding to the region of interest in the subject's rs-fMRI data;

[0048] S102. Apply sliding windowing to the BOLD signal and perform the first correlation analysis to obtain the dynamic functional connectivity information between each region of interest;

[0049] S103. Perform a second correlation analysis on the optimal feature subset selected from the dynamic functional connectivity information and extract effective dynamic functional connectivity information;

[0050] S104. Threshold the effective dynamic function connection information and construct the network;

[0051] S105. Use graph theory methods to extract the graph features of the network to analyze differences and classify them.

[0052] In the above implementation, the application of graph theory in DFC requires constructing a network after obtaining the DFC. Since FC is a two-dimensional data model, traditional DFC analysis methods, after adding a time dimension, become three-dimensional data models. However, graph theory methods are based on two-dimensional network models of nodes and edges. While it is easy to construct graph theory networks in FC, it is very difficult to do so in DFC. This solution proposes a quadratic correlation method to extract dynamic functional connectivity information. After sliding windows for each ROI (region of interest), the first correlation of each ROI within each window is calculated to obtain functional connectivity information. Then, a second correlation is calculated for each window to obtain DFC information. This quadratic correlation result is used as the network construction object, i.e., a re-correlation of the correlation of the BOLD signal. This method compresses three-dimensional data into a two-dimensional temporal correlation matrix, making it easy to apply to graph theory analysis. Furthermore, this method can reflect the temporal consistency of human brain functional connectivity and has clear biological significance.

[0053] Specifically, the BOLD signals of N ROIs are extracted from the subjects' rs-fMRI data, and X is set to... n =[x tn ]∈R T×1 Let X be the BOLD signal of the nth Region of Interest (ROI) extracted from the rs-fMRI data of a subject (n = 1, 2, ..., N, t = 1, 2, ..., T), where T is the time point of the signal. The BOLD signals of the extracted N ROIs are segmented using a rectangular window of length L with a step size T0, dividing the BOLD time series of each ROI into W windows, where W = (T - T0 + L) / L. The BOLD signal of the nth ROI within any w windows is X. n =[x t,n ]∈R L×1 This represents the number of windows.

[0054] The present invention has the following advantages:

[0055] 1. The quadratic correlation method can effectively extract dynamic functional connectivity information, while transforming DFC into a two-dimensional time-domain functional connectivity correlation matrix. Based on this method, a network can be constructed. It has low complexity, can contain most of the effective dynamic functional connectivity information, has clear biological significance, and is easy to analyze.

[0056] 2. The variance of the functional connections (FC) between windows is used as a feature for feature subset selection. The variance between windows can reflect the stability of the functional connectivity between two ROIs over time. Selecting the optimal feature subset based on the variance contains DFC stationarity information and can better reflect the differences in the dynamic changes of functional connectivity between groups.

[0057] 3. By treating the graph theory feature matrix at a certain threshold as a feature in the Relief algorithm, the multidimensional graph theory features of all subjects at that threshold are considered as a whole. The Euclidean distance of the multidimensional features is used to find nearest neighbor samples within the same and different classes, and weights are iterated. Finally, each threshold yields its corresponding weight; a larger weight indicates more significant differences between groups in the network constructed at that threshold, and a stronger classification ability. Applying this to threshold selection in graph theory construction avoids subjective factors and enables automated decision-making, ensuring high efficiency and high performance.

[0058] As an optional implementation, step S102 involves sliding a window onto the BOLD signal and performing a first correlation analysis to obtain dynamic functional connectivity information between regions of interest. This specifically includes the following steps:

[0059] S1021. Use a rectangular window to slide segment the BOLD signal of each extracted region of interest with a set step size, and divide the BOLD time series of each region of interest into multiple windows.

[0060] S1022. Calculate the Pearson correlation coefficient of the BOLD signal between two adjacent regions of interest within each window to obtain dynamic function connection information.

[0061] In the above implementation, a first correlation processing step is performed to extract relevant information between ROIs within each window. The functional connectivity of each ROI within that window is obtained by calculating the Pearson correlation coefficient of the BOLD signals between all ROIs in each window. The correlation between the n1th ROI and the n2th ROI within the wth window can be expressed as:

[0062]

[0063] In the formula, and This represents the time series of the n1-th ROI and the n2-th ROI within the w-th window. and They represent and The mean of the ROIs. If the number of extracted ROIs is N, then the correlation coefficient matrix of the w-th window for any subject can be obtained as follows:

[0064]

[0065] As an optional implementation, in step S103, a second correlation analysis is performed on the optimal feature subset selected from the dynamic functional connectivity information to extract effective dynamic functional connectivity information, specifically including the following steps:

[0066] S1031. Extract the trace-free upper triangular elements from the correlation coefficient matrix of the w-th window and determine the expansion vector of the upper triangular elements.

[0067] S1032. Based on the expansion vector, determine the dynamic functional connection information between two adjacent windows;

[0068] S1033. Select the optimal feature subset based on the variance of the dynamic functional connection information between two adjacent windows;

[0069] S1034. Calculate the effective dynamic functional connection information between two adjacent windows using the optimal feature subset.

[0070] In the above implementation, a second correlation processing step is performed to extract relevant information between each time window. The Pearson correlation coefficients between each pair of time windows are calculated from the correlation coefficient matrices obtained in the above steps, yielding Dynamic Functional Connection Information (DFC) that contains both relevant information between ROIs and relevant information between time windows.

[0071] Since the first correlation result is a symmetric matrix, its trace is the result of Pearson correlation of a certain ROI with respect to itself, and it needs to be removed. The specific steps are as follows:

[0072] Extraction of the correlation coefficient matrix for the w-th window Extract the upper triangular elements after removing the trace, and let u (w) =[r i (w) ]∈R 1×I Let I be the vector formed by expanding the elements of the upper triangular shape, where I = N × (N-1) / 2, and r is the number of valid elements selected. i (w) Let i be the i-th element of the expanded vector, where i = 1, 2, ..., I. Then the correlation between the w1-th window and the w2-th window can be expressed as:

[0073]

[0074] In the formula, and Let represent the vector formed by expanding the upper triangular elements of the correlation coefficient matrix between the w1-th and w2-th windows of the subject. and They represent and The mean of the values. For W windows, the correlation coefficient matrix between any two adjacent windows can be obtained as follows:

[0075]

[0076] The elements in R2 are the Dynamic Functional Connections (DFCs) of the corresponding ROIs within the window, and these secondary correlation results will be used as the network construction objects in the future.

[0077] After dividing the ROI, the full-factor (FC) of the subjects usually contains thousands of features. After adding the time dimension, the number of features will increase exponentially. These features include some redundant and useless features. Removing redundant features will not cause information loss and can effectively improve the accuracy of the model. It is necessary to select the optimal feature subset for R1 for subsequent processing.

[0078] Studies have found that the dynamic changes in functional connectivity differ between patients with brain diseases and healthy groups, a finding confirmed by standard deviation analysis. Functional connectivity in disease groups typically exhibits less variability. Based on this, this invention uses the variance of the functional connectivity (FC) between windows as a feature subset selection criterion. The variance between windows reflects the temporal stability of functional connectivity between two regions of interest (ROIs). Selecting features that better reflect the differences in dynamic changes in functional connectivity between groups based on variance is chosen as the optimal feature subset. The minimum intra-class distance of the variance of FC between windows is used as the feature subset selection criterion. This selection reduces computational complexity and effectively addresses the problem of disordered sample distribution at the boundary between two classes, demonstrating good suitability for the binary classification problems of MCI and NC in this study.

[0079] u (w) =[r i (w) ]∈R 1×I Let r be the vector expanded from the upper triangular elements of the w-th window of the first correlation coefficient matrix, where r i (w) Let i be the i-th element of the expanded vector, i = 1, 2, ..., I, where I = N × (N-1) / 2. Since the experimental data only involves the MCI and NC classes, suppose there are M subjects in a certain class, m = 1, 2, ..., M. Then the variance of the features between the DFC windows of the m-th subject in that class can be expressed as:

[0080]

[0081] The obtained V m Let μ1 be the variance eigenvector, with I elements. Here, μ1 represents the variance eigenvector across all windows of the subject. w The group average is expressed as:

[0082]

[0083] The intra-class distance of all subject characteristics in this class can be expressed as:

[0084]

[0085] d is a vector of length I, where each element represents the intra-class distance of the corresponding feature, and μ2 is the group mean of all subjects in that class, expressed as:

[0086]

[0087] The positions of the J features selected by the minimum intra-class distance criterion can be calculated as follows:

[0088]

[0089] The optimal feature subset selected is:

[0090]

[0091] Finally, the correlation between windows is calculated using the optimal feature subset. The correlation between any subject's w1-th window and w2-th window is expressed as follows:

[0092]

[0093] In the formula, The feature vectors selected from the features of the w1 and w2 windows. and They represent and The mean of the features. For W windows, after selecting the optimal feature subset, the correlation coefficient matrix between each pair of windows can be obtained as follows:

[0094]

[0095] As an optional implementation, in step S1031, the upper triangular elements of the correlation coefficient matrix of the w-th window are extracted after removing the trace, and the expansion vector of the upper triangular elements is determined. Specifically, this includes the following steps:

[0096] S10311. Determine the correlation coefficient matrix of the w-th window based on the Pearson correlation coefficient of the BOLD signal between two adjacent regions of interest within each window.

[0097] S10312. Extract the upper triangular elements of the correlation coefficient matrix of the w-th window and remove the trace.

[0098] S10313. Determine the expansion vector of the upper triangular elements.

[0099] As an optional implementation method, the dynamic functional connection information between two adjacent windows is determined based on the expansion vector. Specifically, the dynamic functional connection information between two adjacent windows is determined based on the first expansion vector and the second expansion vector corresponding to the w1-th window and the w2-th window, as well as the mean values ​​corresponding to the first expansion vector and the second expansion vector.

[0100] As an optional implementation, the optimal feature subset is selected based on the variance of the dynamic functional connection information between two adjacent windows. Specifically, the optimal feature subset in the dynamic functional connection information is selected based on the minimum intra-class distance of the variance of the dynamic functional connection information between two adjacent windows.

[0101] As an optional implementation, in step S104, the effective dynamic functional connection information is thresholded and a network is constructed. Specifically, the matrix corresponding to the effective dynamic functional connection information is thresholded to obtain the adjacency matrix in graph theory that represents the connection relationship of the network.

[0102] In the above implementation, the resulting quadratic correlation matrix P, where each element represents the dynamic information of functional connections between corresponding time windows, needs to be mapped to a two-dimensional graph when using graph theory for analysis. The values ​​of the elements in P range from -1 to 1, and the selection of which elements to map determines the properties of the final constructed network. Thresholding transforms the matrix into a binary matrix corresponding to the adjacency matrix in graph theory, which can characterize the connection relationships of the network and is a commonly used method in graph theory analysis. However, the choice of threshold has a significant impact on the constructed network. If the threshold is too small, most connections will be considered valid connections, resulting in too many "false positive" connections; if the threshold is too large, too much information will be removed, blurring the differences between the two sets of data.

[0103] This invention employs a thresholding method to threshold secondary related information. A threshold δ is set, and elements in P greater than δ represent strong functional connectivity between two time points, which are considered valid connections. Conversely, elements less than δ are considered invalid connections. Valid connections are retained, and invalid connections are removed to threshold P. Due to the sensitivity of the threshold δ setting, this invention uses the relief algorithm to set a cost function ω for the graph theory features of the network constructed under different thresholds δ. δ Select the ω with optimal performance δ Graph theory analysis is performed on the network constructed under the corresponding threshold δ, which enables adaptive threshold selection while ensuring objectivity. The specific steps are as follows:

[0104] The adjacency matrix E is obtained by thresholding the quadratic relevant information P of all subjects at a threshold of δ. δ and weight matrix C δ The calculation formula is as follows:

[0105]

[0106]

[0107] E δ and C δ Each matrix, W×W, represents an undirected graph G with W nodes. δ ,like This indicates that nodes w1 and w2 are connected, and their weights correspond to... Then, by extracting the undirected graph G δGraph theory properties are used to quantitatively represent each graph. The igraph package in R offers many such metrics, such as non-linear distance (average path length) to measure the size of each network, and properties like density and clustering coefficients to reflect network connectivity and density. These low-dimensional features clearly reflect the fundamental properties of each network, including the number of strong connections and the overall efficiency of the connections. These interrelated features provide a quantitative measurement of network properties and can be used to identify network differences associated with different diseases.

[0108] Some graph theory properties used in this invention include:

[0109] Average path length: the average of the shortest distances between each pair of nodes;

[0110] Maximum degree: The maximum number of edges incident to a vertex, counted twice in a loop;

[0111] Number of sides: The number of sides of a figure;

[0112] Graph density: the ratio of the number of edges to the total number of possible edges;

[0113] Network clustering coefficient: A coefficient representing the degree of clustering of nodes in a graph, indicating the degree of interaction between nodes;

[0114] Edge connectivity (adhesion): The minimum number of edges required to obtain a non-strongly connected graph;

[0115] Graph entropy: Shannon entropy is calculated by weighted edge normalization of a graph;

[0116] Node entropy: Calculate the Shannon entropy by normalizing the weighted edges of the nodes facing the street, and then iteratively count and take the average value;

[0117] Small-world property: The measure of small-world property is:

[0118]

[0119] In the formula, η is the ratio of the clustering coefficient of the real network to the average clustering coefficient of its random network, where the random network has the same number of edges and weights as the real network, but with random wiring; λ is similar to η, but λ is the path length ratio. Networks with σ>1 are considered small-world networks, and the larger the σ, the stronger the small-world property. Finally, each participant receives a graph-theoretic feature vector F of length L, used to adaptively select the threshold.

[0120] The graph theory feature matrix F of a subject is randomly selected from any class of subject data. m m = 1, 2, ..., M. Based on the Euclidean distance and F... m Select q nearest neighbors from subjects of the same type. Then from F mSelect q nearest neighbors from samples of different classes. Then the total Euclidean distance between classes of the same type and classes of different types and for:

[0121]

[0122]

[0123] In the formula, l = 1, 2, ..., L represents the l-th feature. By traversing the graph theory feature matrix of all subjects in this class, the cost function ω under the threshold δ is obtained. δ It can be represented as:

[0124]

[0125] In the formula, k is the number of iterations. ω δ The larger the value of ω, the more significant the differences between feature groups at that threshold, and the stronger the classification ability; conversely, the smaller the value, the weaker the classification ability. δ The threshold δ corresponding to the maximum value is the optimal threshold. The undirected graph G under the optimal threshold δ is ultimately selected. δ We conducted a differential analysis and experiments on the graph theory characteristics.

[0126] Corresponding to the indicator weight optimization method in the above embodiment, Figure 2 This is a structural block diagram of an index weight optimization device provided in an embodiment of the present invention. For ease of explanation, only the parts relevant to the embodiment of the present invention are shown. Figure 2 As shown, the graph theory analysis device for dynamic functional connectivity in mild cognitive impairment includes:

[0127] Signal extraction module 21 is used to extract the BOLD signal corresponding to the region of interest in the subject's rs-fMRI data;

[0128] The first correlation module 22 is used to slide the window on the BOLD signal and perform the first correlation analysis to obtain the dynamic functional connection information between each region of interest.

[0129] The second correlation module 23 is used to perform a second correlation analysis on the optimal feature subset selected from the dynamic functional connection information and extract effective dynamic functional connection information.

[0130] Network construction module 24 is used to threshold the effective dynamic functional connection information and construct the network;

[0131] Graph theory analysis module 25 is used to extract the graph features of a network using graph theory methods to analyze differences and classify them.

[0132] As an optional implementation, the first related module 22 is specifically used for:

[0133] The BOLD signal of each extracted region of interest is divided into multiple windows by sliding segmentation using a rectangular window with a set step size.

[0134] The Pearson correlation coefficient of the BOLD signal between two adjacent regions of interest within each window is calculated to obtain the dynamic functional connectivity information.

[0135] As an optional implementation, the second related module 23 is specifically used for:

[0136] Extract the trace-free upper triangular elements from the correlation coefficient matrix of the w-th window, and determine the expansion vector of the upper triangular elements;

[0137] Based on the expansion vector, determine the dynamic functional connection information between two adjacent windows;

[0138] The optimal feature subset is selected based on the variance of the dynamic functional connectivity information between two adjacent windows.

[0139] Calculate the effective dynamic functional connectivity information between two adjacent windows using the optimal feature subset.

[0140] As an optional implementation, the network construction module 24 is specifically used to: threshold the matrix corresponding to the effective dynamic functional connection information to obtain the adjacency matrix in graph theory that represents the connection relationship of the network.

[0141] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. For example... Figure 3 As shown, the electronic device 3 in this embodiment includes: a processor 30, a memory 31, and a computer program 32 stored in the memory 31 and executable on the processor 30. When the processor 30 executes the computer program 32, it implements the steps in the above-described embodiments of the index weight optimization method, for example... Figure 2 Steps S101 to S106 are shown. Alternatively, when processor 30 executes computer program 32, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 2 The functions of modules 21 to 25 are shown.

[0142] For example, computer program 32 can be divided into one or more modules / units, one or more of which are stored in memory 31 and executed by processor 30 to complete this application. One or more modules / units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 32 in electronic device 3. For example, computer program 32 can be divided into... Figure 2 Modules 21 to 25 are shown.

[0143] Electronic device 3 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. Electronic device 3 may include, but is not limited to, processor 30 and memory 31. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 3 and does not constitute a limitation on electronic device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device may also include input / output devices, network access devices, buses, etc.

[0144] The processor 30 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0145] The memory 31 can be an internal storage unit of the electronic device 3, such as a hard disk or RAM. The memory 31 can also be an external storage device of the electronic device 3, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory 31 can include both internal and external storage units of the electronic device 3. The memory 31 is used to store computer programs and other programs and data required by the electronic device. The memory 31 can also be used to temporarily store data that has been output or will be output.

[0146] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0147] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0148] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0149] In the embodiments provided in this application, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0150] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0151] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0152] If an integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above-described methods for optimizing the weights of various indicators. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0153] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A graph theory analysis method for dynamic functional connectivity in mild cognitive impairment, characterized in that, The method includes the following steps: Extract the BOLD signal corresponding to the region of interest from the subject's rs-fMRI data; The BOLD signal is windowed and a first correlation analysis is performed to obtain the dynamic functional connectivity information between each region of interest. A second correlation analysis is performed on the optimal feature subset selected from the dynamic functional connection information to extract effective dynamic functional connection information; The effective dynamic function connection information is thresholded and a network is constructed. Graph theory methods are used to extract the graph features of the network for analysis and classification; The process of applying a sliding window to the BOLD signal and performing the first correlation analysis to obtain the dynamic functional connectivity information between regions of interest specifically includes the following steps: The BOLD signal of each extracted region of interest is divided into multiple windows by sliding segmentation using a rectangular window with a set step size. Calculate the Pearson correlation coefficient of the BOLD signal between two adjacent regions of interest within each window to obtain the dynamic function connection information; The step of performing a second correlation analysis on the optimal feature subset selected from the dynamic functional connectivity information and extracting effective dynamic functional connectivity information specifically includes the following steps: Extract the trace-free upper triangular elements from the correlation coefficient matrix of the w-th window, and determine the expansion vector of the upper triangular elements; Based on the expansion vector, determine the dynamic functional connection information between two adjacent windows; The optimal feature subset is selected based on the variance of the dynamic functional connection information between two adjacent windows. The effective dynamic functional connection information between two adjacent windows is calculated using the optimal feature subset. The determination of dynamic functional connection information between two adjacent windows based on the expansion vector specifically involves: Based on the first expansion vector and the second expansion vector corresponding to the w1-th window and the w2-th window, and the mean values ​​corresponding to the first expansion vector and the second expansion vector, the dynamic functional connection information between the two adjacent windows is determined. The selection of the optimal feature subset based on the variance of the dynamic functional connectivity information between two adjacent windows specifically involves: The optimal feature subset in the dynamic functional connection information is selected based on the minimum intra-class distance of the variance of the dynamic functional connection information between two adjacent windows.

2. The graph theory analysis method for dynamic functional connectivity in mild cognitive impairment according to claim 1, characterized in that, The step of extracting the trace-free upper triangular elements from the correlation coefficient matrix of the w-th window and determining the expanded vector of the upper triangular elements specifically includes the following steps: The correlation coefficient matrix of the w-th window is determined based on the Pearson correlation coefficient of the BOLD signal between two adjacent regions of interest within each window. Extract the upper triangular elements of the correlation coefficient matrix of the w-th window, removing the trace. Determine the expansion vector of the upper triangular elements.

3. The graph theory analysis method for dynamic functional connectivity in mild cognitive impairment according to claim 1, characterized in that, The process of thresholding the effective dynamic functional connection information and constructing the network specifically involves: Thresholding the matrix corresponding to the effective dynamic function connection information yields a binary matrix that represents the adjacency matrix in graph theory, representing the connection relationship of the network.

4. An apparatus employing the graph theory analysis method for dynamic functional connectivity in mild cognitive impairment as described in claim 1, characterized in that, include: The signal extraction module is used to extract the BOLD signal corresponding to the region of interest in the subject's rs-fMRI data; The first correlation module is used to slide window the BOLD signal and perform the first correlation analysis to obtain the dynamic functional connection information between each region of interest. The second correlation module is used to perform a second correlation analysis on the optimal feature subset selected from the dynamic functional connection information and extract effective dynamic functional connection information. The network construction module is used to perform thresholding processing on the effective dynamic functional connection information and construct the network; The graph theory analysis module is used to extract the graph features of the network using graph theory methods to analyze differences and classify them.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 3 above.

6. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 3.

Images