A brain metabolic network epicenter detection method and device

Abstract

The application provides a brain metabolism network epicenter detection method and device. The method comprises the following steps: establishing a healthy brain network based on MRI brain imaging data of healthy subjects, including a structural network and a functional network with the same brain network nodes; calculating a brain metabolism impairment degree value L of each brain network node based on FDG-PET brain imaging data of disease patients and healthy subjects; calculating a probability P1 and P2 that each brain network node is a brain metabolism network epicenter based on the healthy brain structural network and the functional network respectively; and comprehensively ranking the brain network nodes according to the sizes of L, P1 and P2, wherein the first few brain network nodes are the brain metabolism network epicenter. The application can accurately detect the brain metabolism network epicenter and has good robustness.

Description

Technical Field

[0001] This invention belongs to the field of brain network technology, specifically relating to a method and device for detecting the epicenter of a brain metabolic network. Background Technology

[0002] The human brain is composed of hundreds of billions of neurons, whose interconnected networks form the foundation for complex and diverse brain functions. Research indicates that the onset and progression of major brain diseases are related to the disruption of these brain networks. Connectome structure plays a crucial role in disease transmission, particularly in the transneuronal propagation of misfolded proteins. Studying brain network epicenters (the core starting points of functional network damage) is essential for understanding the pathogenesis of neurodegenerative diseases. Neurodegenerative diseases are not confined to specific brain regions but extend throughout a broader range of functional neural networks. Understanding the structural and functional characteristics of these networks can help identify disease transmission pathways and potential intervention points.

[0003] The epicenter of brain networks, the initial brain regions initiating major brain diseases, plays a crucial role in determining pathogenesis and developing new treatment strategies. Currently, brain network research generally employs inter-group statistical comparative analysis to identify which network connections show significant changes compared to healthy individuals. However, this method struggles to effectively pinpoint the epicenter of the brain network.

[0004] In view of this, the present invention proposes a method and device for detecting the epicenter of a brain metabolic network. Summary of the Invention

[0005] To address the aforementioned problems in the prior art, this invention provides a method and apparatus for detecting the epicenter of the brain metabolic network, which can obtain the brain functional damage area most relevant to brain diseases and the accurate location of the epicenter of the brain metabolic network.

[0006] To achieve the above objectives, the present invention adopts the following technical solution.

[0007] In a first aspect, the present invention provides a method for detecting epicenters of brain metabolic networks, comprising the following steps:

[0008] A healthy brain network was established based on MRI brain imaging data of healthy subjects, including structural and functional networks with identical brain network nodes.

[0009] Based on FDG-PET brain imaging data of disease patients and healthy subjects, the degree of brain metabolic damage L of each brain network node was calculated;

[0010] Based on healthy brain structural networks and functional networks, respectively, calculate the probability P1 and P2 that each brain network node may be the epicenter of the brain metabolic network.

[0011] The brain network nodes are ranked comprehensively based on the values ​​of L, P1, and P2, and the top-ranked brain network nodes are the epicenters of the brain metabolic network.

[0012] Furthermore, methods for establishing healthy brain structure networks include:

[0013] S1. Acquire MRI brain imaging data from healthy subjects, including diffusion tensor imaging (DTI) and functional brain imaging (fMRI).

[0014] S2. Convert MRI brain imaging data into NIFTI format data;

[0015] S3. Use the diffusion tensor model to fit the DTI data to the tensor, and use the least squares method to obtain the optimal solution λ1, λ2, λ3 of the eigenvalues ​​of the diffusion tensor matrix and their corresponding eigenvectors e1, e2, e3, where λ1≥λ2≥λ3.

[0016] S4. Calculate the diffusion anisotropy fraction FA for each pixel, using the following formula:

[0017]

[0018] In the formula, MT=(λ1+λ2+λ3) / 3;

[0019] S5. Use the FACT algorithm to perform fiber tracing on the DTI images to obtain the whole brain white matter fiber map for each healthy subject: use each pixel in the DTI image as the starting seed point, use the direction of the principal feature vector e1 as the tracing direction, and use the directions of the second and third feature vectors e2 and e3 to correct the tracing direction; when the FA of a pixel is less than 0.2, stop tracing the fiber bundle; then use the next pixel as the starting seed point to perform fiber tracing until all pixels have been traversed;

[0020] S6. Based on the brain region division of the human brain functional atlas, the whole brain is divided into multiple brain network nodes. The fiber bundle connection density between two brain network nodes is used as the value SC of the edge connecting the two brain network nodes, thus obtaining the healthy brain structure network; the formula for calculating SC is:

[0021]

[0022] In the formula, N represents the number of healthy subjects. fib The number of healthy subjects with a fiber tract root count greater than 0 between two brain network nodes, fib i Let represent the number of fiber bundles between the two nodes for the i-th healthy subject, where i = 1, 2, ..., N.

[0023] Furthermore, methods for establishing healthy brain function networks include:

[0024] Convert brain functional imaging fMRI data into 4D NIFTI format data including time information;

[0025] Based on fMRI signals from multiple consecutive time points of brain network nodes, the Pearson correlation coefficient between two brain network nodes is calculated, and the correlation coefficient is used as the value FC of the edge connecting the two brain network nodes to obtain a healthy brain functional network.

[0026] Furthermore, the method for establishing a healthy brain functional network also includes: preprocessing fMRI images of healthy subjects, including time correction, head motion correction, spatial normalization, Gaussian smoothing, white matter signal regression, cerebrospinal fluid signal regression, Friston24 parameter head motion regression, and low-frequency filtering.

[0027] Furthermore, the method for establishing a healthy brain function network also includes sparsifying the brain function network: if SC ij =0, then FC is set. ij =0, where i and j are the indices of the two brain network nodes.

[0028] Furthermore, the methods for calculating the degree of brain metabolic impairment L include:

[0029] FDG-PET brain imaging data of patients and healthy subjects were acquired and converted into NIFTI format data; the DARTEL algorithm was used to standardize the FDG-PET brain images to MNI space, and Gaussian kernel was used to smooth the spatially standardized data.

[0030] Divide the pixel value of each pixel by the baseline value to obtain the standardized uptake score SUVR, and save the SUVR image with SUVR as the pixel value. The baseline value is equal to the average pixel value of brain regions not affected by brain diseases.

[0031] Calculate the brain metabolic damage value l of the i-th pixel. i The formula is:

[0032]

[0033]

[0034] In the formula, t i Let be the t-value of the two-sample t-test for the i-th pixel. Let be the mean and variance of the SUVR value of the i-th pixel for each of the n1 members of the patient group. These are the mean and variance of the SUVR value of the i-th pixel for each of the n2 members of the healthy group;

[0035] The average value of the degree of brain metabolic damage of each brain network node in the corresponding region is calculated to obtain the degree of brain metabolic damage L of each brain network node.

[0036] Furthermore, the calculation method for P1 includes:

[0037] Calculate the intermediate probability value A of P1 for the i-th brain network node. i The formula is:

[0038]

[0039] In the formula, L ij SC represents the degree of brain metabolic impairment of the j-th neighbor node of the i-th brain network node. ij Let the edge value be the value of the edge connecting the i-th brain network node to its j-th neighbor node in a healthy brain structure network, where j = 1, 2, ..., N. i N i The number of neighboring nodes;

[0040] The formula for calculating P1 is:

[0041]

[0042] In the formula, P 1i Let P1 be the value of the i-th brain network node, and A be the value of P1. ij P1 is the intermediate probability of the j-th neighbor node of the i-th brain network node.

[0043] Furthermore, the calculation methods for P2 include:

[0044] Calculate the intermediate probability value B of P2 for the i-th brain network node. i The formula is:

[0045]

[0046] In the formula, FC ij The value of the edge connecting the i-th brain network node to its j-th neighbor node in the healthy brain function network;

[0047] Calculate P2 using the following formula:

[0048]

[0049] In the formula, P 2i Let P2 be the value of the i-th brain network node, and B be the value of P2. ij P2 is the intermediate probability of the j-th neighbor node of the i-th brain network node.

[0050] Furthermore, the method also includes assessing the extent of influence of the epicenter of the brain metabolic network, including:

[0051] The SUVR images of the patients and healthy subjects were merged into a 4D sequence including subject information to obtain the PET image subject sequence for each group.

[0052] Using brain network nodes as nodes of the PET network, the Pearson correlation coefficient between two nodes is calculated based on the standardized uptake score (SUVR) of all subjects at the nodes, and the correlation coefficient is used as the value of the edge connecting the two nodes of the PET network (Pet).

[0053] Sparsification of PET networks using healthy brain structure networks: If SC ij =0, then set Pet ij =0;

[0054] Calculate Pet for each member of the patient group ij Pets with each member of the health group ij The difference was analyzed using a permutation test to determine the significance of the difference.

[0055] The edges that connect to the epicenter nodes of the brain metabolism network are identified from the edges corresponding to the differences with significant differences. The region formed by the nodes contained in the edges is the influence range of the epicenter of the brain metabolism network.

[0056] In a second aspect, the present invention provides a brain metabolic network epicenter detection device, comprising:

[0057] The network building module is used to build healthy brain networks based on MRI brain imaging data of healthy subjects, including structural and functional networks with identical brain network nodes;

[0058] The L-value calculation module is used to calculate the degree of brain metabolic damage L for each brain network node based on FDG-PET brain imaging data of disease patients and healthy subjects.

[0059] The probability calculation module is used to calculate the probability P1 and P2 that each brain network node may be the epicenter of the brain metabolic network, based on the healthy brain structural network and the functional network, respectively.

[0060] The epicenter determination module is used to comprehensively rank brain network nodes based on the magnitudes of L, P1, and P2, with the top few brain network nodes being the epicenters of the brain metabolic network.

[0061] Compared with the prior art, the present invention has the following beneficial effects.

[0062] This invention establishes a healthy brain network based on MRI brain imaging data from healthy subjects. Using FDG-PET brain imaging data from both diseased and healthy subjects, it calculates the degree of brain metabolic impairment (L) at each network node. Based on both the healthy brain structural and functional networks, it calculates the probabilities (P1 and P2) that each network node might be an epicenter of the brain metabolic network. The network nodes are then ranked based on the magnitudes of L, P1, and P2, with the top-ranked nodes identified as epicenters. This enables automatic detection of epicenters in the brain metabolic network. This invention accurately detects epicenters in the brain metabolic network and exhibits excellent robustness. Attached Figure Description

[0063] Figure 1 This is a flowchart of a metabolic network epicenter detection method according to an embodiment of the present invention.

[0064] Figure 2 The distribution of the probabilities P1 and P2 that a node in a healthy brain structural network (A) and a functional network (B) may be an epicenter of the brain metabolic network.

[0065] Figure 3 This is a block diagram of a metabolic network epicenter detection device according to an embodiment of the present invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of this invention clearer and more understandable, the invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0067] Figure 1 This is a flowchart of a metabolic network epicenter detection method according to an embodiment of the present invention, including the following steps:

[0068] Step 101: Establish a healthy brain network based on MRI brain imaging data of healthy subjects, including structural and functional networks with identical brain network nodes;

[0069] Step 102: Based on the FDG-PET brain imaging data of the patient group and the healthy group, calculate the brain metabolic damage value L for each brain network node;

[0070] Step 103: Based on the healthy brain structural network and functional network, calculate the probability P1 and P2 that each brain network node may be the epicenter of the brain metabolic network.

[0071] Step 104: Rank the brain network nodes based on the magnitudes of L, P1, and P2. The top-ranked brain network nodes are the epicenters of the brain metabolic network.

[0072] In this embodiment, step 101 is mainly used to establish a healthy brain network. This embodiment establishes a healthy brain network based on MRI brain imaging data of healthy subjects. The healthy brain network includes a structural network and a functional network. Both networks consist of nodes and edges connecting the nodes, with the edge value representing the relationship between the two nodes connected by the edge. Both networks use the same brain network nodes, with the AAL90 brain atlas as the nodes for the brain structural network. Subsequent embodiments will provide specific methods for establishing these two networks.

[0073] In this embodiment, step 102 is mainly used to calculate the degree of brain metabolic impairment L for each brain network node. The larger the degree of brain metabolic impairment L for a brain network node, the greater the likelihood that the node is the epicenter of the metabolic network. Therefore, brain network nodes belonging to the epicenter of the metabolic network can be determined based on the magnitude of L. This embodiment calculates the degree of brain metabolic impairment for each brain network node based on FDG-PET brain imaging data from patients and healthy subjects. Members of the patient group are generally Alzheimer's disease (AD) patients. Compared with traditional structural imaging techniques, PET (positron emission tomography) imaging has unique advantages, providing information on brain function and metabolism. By using specific radiotracers, PET can detect early metabolic abnormalities in neurodegenerative diseases, even before any obvious structural changes appear. For example, in Alzheimer's disease, PET can identify reduced FDG (glucose metabolism) uptake, amyloid plaques, and the presence of pathological tau protein in the preclinical stage of the disease, which is beneficial for early diagnosis and disease monitoring.

[0074] In this embodiment, step 103 is mainly used to calculate the probability that each brain network node may be the epicenter of the brain metabolic network. Based on the healthy brain structural network and functional network, the probabilities P1 and P2 of each brain network node being the epicenter of the brain metabolic network are calculated respectively. Obviously, the larger the values ​​of P1 and P2, the greater the probability that the corresponding brain network node may be the epicenter of the brain metabolic network. Therefore, the nodes belonging to the epicenter of the brain metabolic network can be determined according to the magnitude of P1 and P2.

[0075] In this embodiment, step 104 is mainly used to determine the location of the epicenter of the brain metabolic network. As mentioned above, the values ​​of L, P1, and P2 of a brain network node are closely related to whether the node is the epicenter of the brain metabolic network. In principle, the approximate location of the epicenter can be estimated based on any one of these parameters. To improve the positioning accuracy, this embodiment ranks the brain network nodes comprehensively based on the values ​​of L, P1, and P2, and selects the top few (e.g., the first 3 to 5) brain network nodes as the epicenters of the brain metabolic network. There are many methods for comprehensive ranking. For example, L, P1, and P2 can be weighted and summed, and ranked according to the sum; or the three parameters can be ranked and filtered according to their importance; or they can be ranked separately, and the top few can be selected, and then the intersection can be found. This embodiment does not limit the specific method of comprehensive ranking.

[0076] As an optional embodiment, the method for establishing a healthy brain structure network includes:

[0077] S1. Acquire MRI brain imaging data from healthy subjects, including diffusion tensor imaging (DTI) and functional brain imaging (fMRI).

[0078] S2. Convert MRI brain imaging data into NIFTI format data;

[0079] S3. Use the diffusion tensor model to fit the DTI data to the tensor, and use the least squares method to obtain the optimal solution λ1, λ2, λ3 of the eigenvalues ​​of the diffusion tensor matrix and their corresponding eigenvectors e1, e2, e3, where λ1≥λ2≥λ3.

[0080] S4. Calculate the diffusion anisotropy fraction FA for each pixel, using the following formula:

[0081]

[0082] In the formula, MT=(λ1+λ2+λ3) / 3;

[0083] S5. Use the FACT algorithm to perform fiber tracing on the DTI images to obtain the whole brain white matter fiber map for each healthy subject: use each pixel in the DTI image as the starting seed point, use the direction of the principal feature vector e1 as the tracing direction, and use the directions of the second and third feature vectors e2 and e3 to correct the tracing direction; when the FA of a pixel is less than 0.2, stop tracing the fiber bundle; then use the next pixel as the starting seed point to perform fiber tracing until all pixels have been traversed;

[0084] S6. Based on the brain region division of the human brain functional atlas, the whole brain is divided into multiple brain network nodes. The fiber bundle connection density between two brain network nodes is used as the value SC of the edge connecting the two brain network nodes, thus obtaining the healthy brain structure network; the formula for calculating SC is:

[0085]

[0086] In the formula, N represents the number of healthy subjects. fib The number of healthy subjects with a fiber tract root count greater than 0 between two brain network nodes, fib i Let represent the number of fiber bundles between the two nodes for the i-th healthy subject, where i = 1, 2, ..., N.

[0087] This embodiment presents a technical solution for establishing a healthy brain structure network. Establishing a healthy brain structure network mainly includes brain network node partitioning and edge value calculation, with edge value calculation being of paramount importance. This embodiment establishes a whole-brain white matter fiber map of a healthy subject and calculates the edge value based on the number or density of fibers between the two nodes connected by each edge.

[0088] Steps S1 and S2 are used for data acquisition and data format conversion. First, MRI brain imaging data from healthy subjects is acquired, including diffusion tensor imaging (DTI) and functional brain imaging (fMRI). Then, the MRI images (DICOM data) are converted into processable NIFTI format data. Since the goal is to establish a healthy brain structural network, only MRI images from healthy subjects are needed. Diffusion tensor imaging (DTI) is currently the only imaging technique capable of non-invasively detecting white matter fiber structures in the living human brain. It can display living brain white matter fiber bundles in three dimensions, providing a three-dimensional and intuitive view of the fiber bundle's course and changes, thus achieving detailed imaging of human nerve fibers.

[0089] Steps S3 to S5 are used to generate a whole-brain white matter fiber map for each healthy subject. Step S3 calculates the three eigenvalues ​​of the diffusion tensor matrix and their corresponding eigenvectors. Step S4 calculates the diffusion anisotropy score FA for each pixel based on the three eigenvalues, using the formula shown in equation (1). Step S5 performs fiber tracing on the DTI image based on the three eigenvectors and the diffusion anisotropy score to obtain a whole-brain white matter fiber map for each healthy subject, specifically, determining all pixels that form each fiber bundle. In actual calculations, fiber bundles with higher resolution than the DTI image are obtained through image interpolation.

[0090] Step S6 is used for brain network node division and boundary value calculation. In this embodiment, based on the brain region division of the human brain functional atlas (such as the AAL90 brain atlas), the whole brain is divided into multiple brain network nodes. The fiber bundle connection density between two nodes is used as the edge value SC connecting the two nodes. The calculation formula of SC is as shown in equation (2). Equation (2) is a piecewise function with SC as the dependent variable and the number of fiber bundle connections between two nodes fib as the independent variable. The number of healthy subjects N when the number of fiber connections between two nodes is greater than 0 is calculated. fibLess than half the number of healthy subjects N (i.e., N) fib When N < 0.5N), the value of SC is 0, which is equivalent to no edge connecting the two nodes; when N fib When ≥0.5N, the value of SC is equal to the mean of the number of fiber bundles (fib) between the two nodes in all healthy subjects.

[0091] As an optional embodiment, the method for establishing a healthy brain function network includes:

[0092] Convert brain functional imaging fMRI data into 4D NIFTI format data including time information;

[0093] Based on fMRI signals from multiple consecutive time points of brain network nodes, the Pearson correlation coefficient between two brain network nodes is calculated, and the correlation coefficient is used as the value FC of the edge connecting the two brain network nodes to obtain a healthy brain functional network.

[0094] This embodiment presents a technical solution for establishing a healthy brain functional network. The healthy brain functional network uses the same brain network nodes as the structural network, but differs in that it calculates the boundary value FC based on brain functional imaging fMRI data. First, the fMRI data is converted into 4-dimensional (3-dimensional + time dimension) NIFTI format data that includes time information; then, the Pearson correlation coefficient between two brain network nodes is calculated, which is the Pearson correlation coefficient of the fMRI signals of two nodes at multiple consecutive time points. This correlation coefficient is the boundary value FC connecting the two nodes.

[0095] As an optional embodiment, the method for establishing a healthy brain functional network further includes: preprocessing the fMRI images of healthy subjects, including time correction, head motion correction, spatial normalization, Gaussian smoothing, white matter signal regression, cerebrospinal fluid signal regression, Friston24 parameter head motion regression, and low-frequency filtering.

[0096] This embodiment presents a technical solution for preprocessing fMRI images. To obtain high-quality fMRI image data and thus improve the accuracy of healthy brain functional networks, preprocessing of the fMRI images is necessary before boundary value calculation. Preprocessing includes time correction, head motion correction, Gaussian smoothing, and low-frequency filtering.

[0097] As an optional embodiment, the method for establishing a healthy brain function network further includes sparsifying the healthy brain function network: if SC ij =0, then FC is set. ij =0, where i and j are the indices of the two brain network nodes.

[0098] This embodiment presents a technical solution for sparsification of a healthy brain functional network. Sparsification refers to reducing the number of edges in the healthy brain functional network, which facilitates reduced computation and increased running speed without affecting the accuracy of epicenter detection. This embodiment utilizes a healthy brain structural network to perform sparsification on the functional network. The technical principle is: if two nodes are not connected by edges in the structural network (i.e., the edge value is 0), then these two nodes are also not connected by edges in the functional network. In other words, if there was originally an edge connecting them, that edge is deleted; if there was originally no edge connecting them, the edge connection is maintained.

[0099] As an optional embodiment, the method for calculating the degree of brain metabolic impairment L includes:

[0100] FDG-PET brain imaging data of patients and healthy subjects were acquired and converted into NIFTI format data; the DARTEL algorithm was used to standardize the FDG-PET brain images to MNI space, and Gaussian kernel was used to smooth the spatially standardized data.

[0101] Divide the pixel value of each pixel by the baseline value to obtain the standardized uptake score SUVR, and save the SUVR image with SUVR as the pixel value. The baseline value is equal to the average pixel value of brain regions not affected by brain diseases.

[0102] Calculate the brain metabolic damage value l of the i-th pixel. i The formula is:

[0103]

[0104] In the formula, t i Let be the t-value of the two-sample t-test for the i-th pixel. Let be the mean and variance of the SUVR value of the i-th pixel for each of the n1 members of the patient group. These are the mean and variance of the SUVR value of the i-th pixel for each of the n2 members of the healthy group;

[0105] The average value of the degree of brain metabolic damage of each brain network node in the corresponding region is calculated to obtain the degree of brain metabolic damage L of each brain network node.

[0106] This embodiment provides a technical solution for calculating the degree of brain metabolic impairment L. As mentioned above, PET imaging can provide information on brain function and metabolism. By using specific radioactive tracers, PET can detect early metabolic abnormalities in neurodegenerative diseases. Therefore, this embodiment calculates the degree of brain metabolic impairment based on glucose-positron emission tomography (FDG-PET).

[0107] First, data acquisition and processing are performed. In this embodiment, the degree of brain metabolic damage is calculated by comparing patients and healthy individuals. Therefore, it is necessary to acquire FDG-PET brain imaging data of the patient group and the healthy group, convert it into NIFTI format data, and then use the DARTEL algorithm to standardize the FDG-PET brain images to the MNI space. Finally, Gaussian smoothing is performed on the spatially standardized data using a Gaussian kernel.

[0108] Then, the standardized uptake score (SUVR) of each pixel is calculated. A brain region unaffected by brain disease is selected as a reference region, and the average pixel value (also known as PET metabolic value) of this reference region is calculated to obtain the baseline value. Then, the pixel value of each pixel is divided by the baseline value to obtain the standardized uptake score (SUVR) of each pixel.

[0109] Finally, the degree of brain metabolic impairment, L, is calculated. First, the degree of brain metabolic impairment, l, is calculated for each pixel. Then, the degree of brain metabolic impairment, L, for each node is obtained by calculating the mean of the degree of brain metabolic impairment, l, for all pixels within each node region. In this embodiment, the l value for each pixel is obtained by calculating the t-value of a two-sample t-test for each pixel. The two samples include the SUVR values ​​x of n1 members of the patient group for each pixel. i1 The SUVR value x of the n2 members in the healthy group i2 The t value of each pixel can be obtained by formula (4). Then, the sign of the t value is used to determine whether the patient group has experienced a decrease in brain metabolism. If t < 0, then there is a decrease in brain metabolism, and the absolute value of t (i.e. -t) is the value of the degree of brain metabolic damage l; otherwise, there is no decrease in brain metabolism, and l = 0, as shown in the piecewise function in formula (3).

[0110] As an optional embodiment, the calculation method for P1 includes:

[0111] Calculate the intermediate probability value A of P1 for the i-th brain network node. i The formula is:

[0112]

[0113] In the formula, L ij SC represents the degree of brain metabolic impairment of the j-th neighbor node of the i-th brain network node. ij Let the edge value be the value of the edge connecting the i-th brain network node to its j-th neighbor node in a healthy brain structure network, where j = 1, 2, ..., N. i N i The number of neighboring nodes;

[0114] The formula for calculating P1 is:

[0115]

[0116] In the formula, P1i Let P1 be the value of the i-th brain network node, and A be the value of P1. ij P1 is the intermediate probability of the j-th neighbor node of the i-th brain network node.

[0117] This embodiment provides a technical solution for calculating P1. P1 is the probability that each node may be the epicenter, calculated based on a healthy brain structure network. In this embodiment, the intermediate probability A for calculating P1 is obtained by weighted summing of the boundary values ​​SC connecting the node and its neighbors using the brain metabolic damage value L of each node's neighboring nodes, as shown in equation (5); then, the mean of A of each node's neighboring nodes is calculated to obtain P1, as shown in equation (6).

[0118] As an optional embodiment, the calculation method for P2 includes:

[0119] Calculate the intermediate probability value B of P2 for the i-th brain network node. i The formula is:

[0120]

[0121] In the formula, FC ij The value of the edge connecting the i-th brain network node to its j-th neighbor node in the healthy brain function network;

[0122] Calculate P2 using the following formula:

[0123]

[0124] In the formula, P 2i Let P2 be the value of the i-th brain network node, and B be the value of P2. ij P2 is the intermediate probability of the j-th neighbor node of the i-th brain network node.

[0125] This embodiment provides a technical solution for calculating P2. P2 is the probability that each node may be the epicenter, calculated based on the healthy brain functional network. The calculation method of P2 is almost exactly the same as that of P1, except that SC in the P1 calculation formula is replaced with FC. See equations (7) and (8) for details.

[0126] Figure 2 Distribution maps of P1 and P2 obtained through actual calculations are provided. The upper part A is the distribution map of P1, and the lower part B is the distribution map of P2. This example ultimately determined that the epicenter region of the AD patient was the right angular gyrus, the left posterior cingulate cortex, the left angular gyrus, the right anterior temporal gyrus, and the right middle temporal gyrus.

[0127] As an optional embodiment, the method further includes assessing the extent of influence of the epicenter of the brain metabolic network, including:

[0128] The SUVR images of the patients and healthy subjects were merged into a 4D sequence including subject information to obtain the PET image subject sequence for each group.

[0129] Using brain network nodes as nodes of the PET network, the Pearson correlation coefficient between two nodes is calculated based on the standardized uptake score (SUVR) of all subjects at the nodes, and the correlation coefficient is used as the value of the edge connecting the two nodes of the PET network (Pet).

[0130] Sparsification of PET networks using healthy brain structure networks: If SC ij =0, then set Pet ij =0;

[0131] Calculate Pet for each member of the patient group ij Pets with each member of the health group ij The difference was analyzed using a permutation test to determine the significance of the difference.

[0132] The edges that connect to the epicenter nodes of the brain metabolism network are identified from the edges corresponding to the differences with significant differences. The region formed by the nodes contained in the edges is the influence range of the epicenter of the brain metabolism network.

[0133] This embodiment presents a technical solution for assessing the impact range of the epicenter of the brain metabolic network. This embodiment establishes a PET network based on SUVR images of subjects in both the patient and healthy groups, and then calculates the PET network boundary value Pet for each member of the patient group. ij Pets with each member of the health group ij The difference is calculated, and the significance of the difference is statistically analyzed by performing a permutation test. Finally, the influence range of the epicenter is determined based on the node regions corresponding to the differences with significant differences.

[0134] First, a PET image subject sequence is established. In this embodiment, the SUVR images of the patient group and the healthy group are merged in four dimensions, including subject information (dimensions), to obtain a PET image subject sequence for each group consisting of the SUVR images of each subject.

[0135] Then, a PET network is constructed. The nodes of the PET network still use the previously obtained brain network nodes, and the boundary values ​​are still obtained by calculating the Pearson correlation coefficient between two nodes. The data used to calculate the correlation coefficient is a sequence composed of the standardized uptake scores (SUVR) of all subjects at each node. Next, the PET network will be sparsified using a healthy brain structure network to simplify the PET network.

[0136] Finally, the area of ​​influence of the epicenter is determined. First, a permutation test is performed on the aforementioned differences to obtain the edges and nodes corresponding to the differences with significant differences. The area formed by the nodes connected to the epicenter node and the epicenter node is the area of ​​influence of the epicenter.

[0137] Figure 3 This is a schematic diagram of the composition of a brain metabolic network epicenter detection device according to an embodiment of the present invention. The device includes:

[0138] Network building module 11 is used to build a healthy brain network based on MRI brain imaging data of healthy subjects, including structural and functional networks with the same brain network nodes;

[0139] L-value calculation module 12 is used to calculate the degree of brain metabolic damage L for each brain network node based on FDG-PET brain imaging data of disease patients and healthy subjects.

[0140] The probability calculation module 13 is used to calculate the probability P1 and P2 that each brain network node may be the epicenter of the brain metabolic network, based on the healthy brain structural network and the functional network, respectively.

[0141] The epicenter determination module 14 is used to comprehensively rank brain network nodes based on the magnitudes of L, P1, and P2, with the top few brain network nodes being the epicenters of the brain metabolic network.

[0142] The apparatus of this embodiment can be used to perform Figure 1 The technical solutions of the method embodiments shown are similar in principle and in effect, and will not be described again here.

[0143] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for detecting the epicenter of a brain metabolic network, characterized in that, Includes the following steps: A healthy brain network was established based on MRI brain imaging data of healthy subjects, including structural and functional networks with identical brain network nodes. Based on FDG-PET brain imaging data of disease patients and healthy subjects, the degree of brain metabolic damage L of each brain network node was calculated; Based on healthy brain structural networks and functional networks, respectively, calculate the probability P1 and P2 that each brain network node may be the epicenter of the brain metabolic network. The brain network nodes are ranked comprehensively based on the values ​​of L, P1, and P2, and the top few brain network nodes are the epicenters of the brain metabolic network. The establishment of the structured network includes: MRI brain imaging data, including diffusion tensor imaging (DTI) and functional brain imaging (fMRI), were acquired from healthy subjects. The MRI brain imaging data were converted to NIFTI format. A diffusion tensor model was used to fit the DTI data, and the optimal solutions λ1, λ2, λ3 of the diffusion tensor matrix eigenvalues ​​and their corresponding eigenvectors e1, e2, e3 were obtained using the least squares method, where λ1 ≥ λ2 ≥ λ3. The diffusion anisotropy fraction (FA) for each pixel was calculated. The FACT algorithm was used to trace fibers in the DTI images to obtain the whole-brain white matter fibers of each healthy subject. The Vitagram uses each pixel in the DTI image as a starting seed point and the direction of the principal feature vector e1 as the tracking direction. Simultaneously, it uses the directions of the second and third feature vectors e2 and e3 to correct the tracking direction. When the FA of a pixel is less than 0.2, tracking of that fiber bundle stops, and the next pixel is used as the starting seed point for fiber tracking, until all pixels have been traversed. Based on the brain region division of the human brain functional atlas, the whole brain is divided into multiple brain network nodes. The fiber bundle connection density between two brain network nodes is used as the value SC of the edge connecting the two brain network nodes, thus obtaining a healthy brain structure network. The calculation of the degree of brain metabolic impairment L includes: FDG-PET brain imaging data of patients and healthy subjects were acquired and converted into NIFTI format. The FDG-PET brain images were normalized to the MNI space using the DARTEL algorithm, and Gaussian smoothing was performed on the spatially normalized data using a Gaussian kernel. The normalized uptake fraction (SUVR) was obtained by dividing the pixel value of each pixel by a baseline value, and saved as an SUVR image with SUVR as the pixel value. The baseline value was equal to the average pixel value of brain regions unaffected by brain disease. The degree of brain metabolic impairment (l) at the i-th pixel was calculated. i ; Calculate the degree of brain metabolic damage value of pixels within the region corresponding to each brain network node l i The average value is used to obtain the brain metabolic damage value L for each brain network node.

2. The method for detecting epicenters of brain metabolic networks according to claim 1, characterized in that, The formula for calculating the diffusion anisotropy fraction FA for each pixel is: In the formula, MT=(λ1+λ2+λ3) / 3; The formula for calculating the value SC is: In the formula, N represents the number of healthy subjects. fib The number of healthy subjects with a fiber tract root count greater than 0 between two brain network nodes, fib i Let represent the number of fiber bundles between the two nodes for the i-th healthy subject, where i = 1, 2, ..., N.

3. The method for detecting epicenters of brain metabolic networks according to claim 2, characterized in that, Methods for establishing healthy brain functional networks include: Convert brain functional imaging fMRI data into 4D NIFTI format data including time information; Based on fMRI signals from multiple consecutive time points of brain network nodes, the Pearson correlation coefficient between two brain network nodes is calculated, and the correlation coefficient is used as the value FC of the edge connecting the two brain network nodes to obtain a healthy brain functional network.

4. The method for detecting epicenters of brain metabolic networks according to claim 3, characterized in that, The method for establishing a healthy brain functional network also includes: preprocessing fMRI images of healthy subjects, including time correction, head motion correction, spatial normalization, Gaussian smoothing, white matter signal regression, cerebrospinal fluid signal regression, Friston 24-parameter head motion regression, and low-frequency filtering.

5. The method for detecting the epicenter of a brain metabolic network according to claim 3, characterized in that, The method for establishing a healthy brain function network also includes sparsifying the brain function network: if SC ij =0, then FC is set. ij =0, where i and j are the indices of the two brain network nodes.

6. The method for detecting epicenters of brain metabolic networks according to claim 3, characterized in that, Calculate the brain metabolic damage value l of the i-th pixel. i The formula is: In the formula, t i Let be the t-value of the two-sample t-test for the i-th pixel. Let be the mean and variance of the SUVR value of the i-th pixel for each of the n1 members of the patient group. These are the mean and variance of the SUVR value of the i-th pixel for each of the n2 members of the healthy group.

7. The method for detecting epicenters of brain metabolic networks according to claim 6, characterized in that, The methods for calculating P1 include: Calculate the intermediate probability value A of P1 for the i-th brain network node. i The formula is: In the formula, L ij SC represents the degree of brain metabolic impairment of the j-th neighbor node of the i-th brain network node. ij Let the edge value be the value of the edge connecting the i-th brain network node to its j-th neighbor node in a healthy brain structure network, where j = 1, 2, ..., N. i N i The number of neighboring nodes; The formula for calculating P1 is: In the formula, P 1i Let P1 be the value of the i-th brain network node, and A be the value of P1. ij P1 is the intermediate probability of the j-th neighbor node of the i-th brain network node.

8. The method for detecting epicenters of brain metabolic networks according to claim 7, characterized in that, The methods for calculating P2 include: Calculate the intermediate probability value B of P2 for the i-th brain network node. i The formula is: In the formula, FC ij The value of the edge connecting the i-th brain network node to its j-th neighbor node in the healthy brain function network; Calculate P2 using the following formula: In the formula, P 2i Let P2 be the value of the i-th brain network node, and B be the value of P2. ij P2 is the intermediate probability of the j-th neighbor node of the i-th brain network node.

9. The method for detecting epicenters of brain metabolic networks according to claim 6, characterized in that, The method also includes assessing the extent of impact of the epicenter of the brain metabolic network, including: The SUVR images of the patients and healthy subjects were merged into a 4D sequence including subject information to obtain the PET image subject sequence for each group. Using brain network nodes as nodes of the PET network, the Pearson correlation coefficient between two nodes is calculated based on the standardized uptake score (SUVR) of all subjects at the nodes, and the correlation coefficient is used as the value of the edge connecting the two nodes of the PET network (Pet). Sparsification of PET networks using healthy brain structure networks: If SC ij =0, then set Pet ij =0; Calculate Pet for each member of the patient group ij Pets with each member of the health group ij The difference was analyzed using a permutation test to determine the significance of the difference. The edges that connect to the epicenter nodes of the brain metabolism network are identified from the edges corresponding to the differences with significant differences. The region formed by the nodes contained in the edges is the influence range of the epicenter of the brain metabolism network.

10. A brain metabolic network epicenter detection device, characterized in that, include: The network building module is used to build healthy brain networks based on MRI brain imaging data of healthy subjects, including structural and functional networks with identical brain network nodes; The L-value calculation module is used to calculate the degree of brain metabolic damage L for each brain network node based on FDG-PET brain imaging data of disease patients and healthy subjects. The probability calculation module is used to calculate the probability P1 and P2 that each brain network node may be the epicenter of the brain metabolic network, based on the healthy brain structural network and the functional network, respectively. The epicenter determination module is used to comprehensively rank brain network nodes based on the magnitudes of L, P1, and P2, with the top few brain network nodes being the epicenters of the brain metabolic network. The establishment of the structured network includes: MRI brain imaging data, including diffusion tensor imaging (DTI) and functional brain imaging (fMRI), were acquired from healthy subjects. The MRI brain imaging data were converted to NIFTI format. A diffusion tensor model was used to fit the DTI data, and the optimal solutions λ1, λ2, λ3 of the diffusion tensor matrix eigenvalues ​​and their corresponding eigenvectors e1, e2, e3 were obtained using the least squares method, where λ1 ≥ λ2 ≥ λ3. The diffusion anisotropy fraction (FA) for each pixel was calculated. The FACT algorithm was used to trace fibers in the DTI images to obtain the whole-brain white matter fibers of each healthy subject. The Vitagram uses each pixel in the DTI image as a starting seed point and the direction of the principal feature vector e1 as the tracking direction. Simultaneously, it uses the directions of the second and third feature vectors e2 and e3 to correct the tracking direction. When the FA of a pixel is less than 0.2, tracking of that fiber bundle stops, and the next pixel is used as the starting seed point for fiber tracking, until all pixels have been traversed. Based on the brain region division of the human brain functional atlas, the whole brain is divided into multiple brain network nodes. The fiber bundle connection density between two brain network nodes is used as the value SC of the edge connecting the two brain network nodes, thus obtaining a healthy brain structure network. The calculation of the degree of brain metabolic impairment L includes: FDG-PET brain imaging data of patients and healthy subjects were acquired and converted into NIFTI format. The FDG-PET brain images were normalized to the MNI space using the DARTEL algorithm, and Gaussian smoothing was performed on the spatially normalized data using a Gaussian kernel. The normalized uptake fraction (SUVR) was obtained by dividing the pixel value of each pixel by a baseline value, and saved as an SUVR image with SUVR as the pixel value. The baseline value was equal to the average pixel value of brain regions unaffected by brain disease. The degree of brain metabolic impairment (l) at the i-th pixel was calculated. i ; Calculate the degree of brain metabolic damage value of pixels within the region corresponding to each brain network node l i The average value is used to obtain the brain metabolic damage value L for each brain network node.

Images