An individual difference brain network construction method based on sparse kernel matrix optimization

Abstract

This invention discloses a method for constructing individual-discretionary brain networks based on sparse kernel matrix optimization. Using brain network atlases, time-series brain region extraction is performed on normal and abnormal brain imaging data. Similarity matrices are calculated for the time-series data of normal and abnormal brain regions using a similarity metric. The absolute value of the difference between the similarity matrices of each abnormal subject and all normal subjects is calculated to obtain several brain region discrepancy matrices. The values ​​with the highest similarity in each row of each brain region discrepancy matrix are taken as sparse kernel matrices. The sparse kernel matrix is ​​multiplied by the average of the remaining brain region discrepancy matrices to optimize the brain region discrepancy matrices. The average of all optimized brain region discrepancy matrices is calculated to obtain the individual-discretionary brain network of the abnormal subject. A discrepancy matrix is ​​constructed between the abnormal and normal control groups. By multiplying the sparse kernel matrix of each brain region's discrepancy by the average of the remaining brain region discrepancy matrices, important features in each brain region's discrepancy matrix are fully extracted and fused.

Description

Technical Field

[0001] This invention relates to brain network construction, specifically to a method for constructing individual-differenced brain networks based on sparse kernel matrix optimization. Background Technology

[0002] The study of brain functional networks is crucial for understanding the complex mechanisms of the human brain. The human brain is a highly integrated network, not an isolated anatomical region, forming a complex and precise network of connections through synaptic interactions between neurons. Related research indicates that various higher cognitive functions depend on the coordinated cooperation of different brain regions, not just a single specific area. The pathogenesis of many mental illnesses can be attributed to, to some extent, to dysregulation of connectivity between the involved brain regions. In recent years, advancements in neuroimaging techniques have made it possible to study abnormal brain connectivity in mental illnesses. Functional magnetic resonance imaging (fMRI) has proven invaluable in revealing the brain's intrinsic functional connectivity. This technique reflects neuronal activity in different brain regions during resting or task-related states by measuring blood oxygen level-dependent (BOLD) signals. Functional connectivity reflects the coordinated activity or synchronicity between different brain regions over a specific time period, thus revealing the relationship between the organizational structure and function of brain networks. Therefore, by using fMRI to analyze differences in functional connectivity between patients and healthy controls, researchers can gain a deeper understanding of the pathological mechanisms of various brain diseases.

[0003] In clinical medicine, schizophrenia (SZ) is a chronic and highly disabling severe mental illness that typically manifests in adolescence or early adulthood, presenting with disturbances in multiple areas, including thought, emotion, and behavior. Schizophrenia possesses a variety of psychopathological features, with core characteristics including positive symptoms (such as delusions, hallucinations, thought disorder, and behavioral disturbances), negative symptoms (including impaired motivation, reduced spontaneous speech, and social withdrawal), and cognitive impairments (including deficits in attention, memory, learning, thinking, and executive function), affecting approximately 1% of the global population. Currently, there are as many as 26 million people worldwide suffering from schizophrenia, causing immense suffering not only to patients but also placing a heavy burden on families and society. Therefore, utilizing modern medical imaging techniques to study the pathological mechanisms of schizophrenia and to achieve early warning and clinical intervention is of great significance. On the other hand, individual differences in functional connectivity have important potential value in neuroscience and brain imaging. First, it helps to understand the neural diversity of the human brain, revealing differences in cognitive abilities, behavioral performance, and emotional states. Secondly, by comparing the functional connectivity patterns of patients and healthy individuals, biomarkers for specific diseases can be identified, advancing symptom prediction and classification research. Simultaneously, research on individual differences in functional connectivity can promote the development of personalized medicine, allowing for the formulation of individualized treatment plans based on each patient's unique functional connectivity characteristics, thereby improving treatment outcomes. Furthermore, studying individual differences in functional connectivity helps reveal the fundamental mechanisms of changes in brain networks, promoting a deeper understanding of brain function. These studies not only advance neuroscience but also demonstrate broad application prospects in personalized clinical diagnosis and treatment.

[0004] However, in current research, researchers mostly calculate functional connectivity matrices based on the correlation of time-series data of brain regions in subjects, comparing the differences between the average functional connectivity matrix of each mental illness patient and the average functional connectivity matrix of a healthy control group to construct an individual difference matrix. However, due to individual differences and the unique brain activity patterns of mental illness patients, accurate analysis and fusion of these complex brain networks remain challenging. Using the average functional connectivity matrix of the normal group may mask subtle differences between individuals, failing to fully reflect the unique pathological characteristics of each patient, leading to the loss of individual characteristics and thus limiting the accuracy of personalized diagnosis. Summary of the Invention

[0005] Purpose of the invention: To address the above-mentioned shortcomings, this invention provides a method for constructing individual-differenced brain networks based on sparse kernel matrix optimization to improve the accuracy of brain networks.

[0006] Technical Solution: To solve the above problems, this invention employs a method for constructing individual-discretionary brain networks based on sparse kernel matrix optimization, comprising the following steps:

[0007] (1) Using brain network atlas, brain region time series were extracted from normal brain imaging data of several normal subjects and abnormal brain imaging data of several abnormal subjects to obtain brain region time series of several normal subjects and brain region time series of several abnormal subjects.

[0008] (2) Calculate the similarity matrix of time series of brain regions of several normal subjects and the similarity matrix of time series of brain regions of several abnormal subjects by using similarity measurement.

[0009] (3) Calculate the absolute value of the difference between the similarity matrix of each abnormal subject and the similarity matrix of all normal subjects to obtain the difference matrix of several brain regions of the abnormal subject.

[0010] (4) For the vth brain region difference matrix of each abnormal subject, arrange the similarity of each row of the brain region difference matrix from large to small, take the first N similarity values, and set the other similarity values ​​to zero to obtain the vth sparse kernel matrix of the abnormal subject. Multiply the vth sparse kernel matrix by the average value of the remaining brain region difference matrices to obtain the vth optimized brain region difference matrix. Through this step, all optimized brain region difference matrices of all abnormal subjects are obtained.

[0011] (5) Calculate the average value of the optimized brain region difference matrix of all brain regions for each abnormal subject to obtain the individual difference brain network of the abnormal subject.

[0012] Furthermore, in step (4), the average value of the vth sparse kernel matrix and the other normalized brain region difference matrices is multiplied to obtain the optimized sparse kernel matrix of the vth brain region.

[0013] Furthermore, the v-th optimized brain region difference matrix P (v) for:

[0014]

[0015] Among them, S (v) Let D′ represent the v-th sparse kernel matrix. (k) Let m represent the difference matrix of the k-th normalized brain region, and m represent the total number of normal subjects.

[0016] Furthermore, the similarity calculation formula is as follows:

[0017]

[0018] Where W(i,j) represents the brain region x of the subject. i and brain region x j The similarity value between time series, ρ 2 (x i ,x j ) represents the brain region x of the subject.i and brain region x j The square of the Euclidean distance between time series, μ represents the hyperparameter used for empirical settings, and ε i,j This represents the hyperparameters used to eliminate scaling issues.

[0019] Furthermore, step (4) specifically includes:

[0020]

[0021] Where S(i,j) represents the value in the i-th row and j-th column of the sparse kernel matrix S; D(i,j) is the value in the i-th row and j-th column of the brain region difference matrix D, representing the absolute value of the difference between W(i,j) of the abnormal subject and W(i,j) of the normal subject; and D(i,k) is the value in the i-th row and k-th column of the brain region difference matrix D.

[0022] Furthermore, the brain region difference matrix is ​​normalized:

[0023]

[0024] Where D′ represents the normalized matrix of brain region difference matrix D, and D′(i,j) represents the value in the i-th row and j-th column of the normalized brain region difference matrix D′.

[0025] Furthermore, the normal brain imaging data and the abnormal brain imaging data are structural magnetic resonance imaging data from a publicly available data center. They are preprocessed to obtain preprocessed images of normal brain region data and abnormal brain region data. The obtained preprocessed images are divided into brain regions, and the average time series of all voxels in each brain region in each data is calculated.

[0026] Furthermore, the preprocessed brain imaging data is divided into brain regions based on brain atlases, which include AAL atlases and BA atlases.

[0027] The present invention also employs a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the above method when executing the computer program.

[0028] The present invention also employs a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the above method.

[0029] Beneficial Effects: Compared with existing technologies, the significant advantage of this invention is the construction of a difference matrix between abnormal and normal control brain regions. By multiplying the sparse kernel matrix of each brain region's difference with the average of the difference matrices of the remaining brain regions, important features in each brain region's difference matrix are fully extracted and fused. Using the constructed individual difference matrix as a priori guide for feature selection in the prediction and classification of related mental illness symptoms and cognition, the accuracy of symptom and cognitive prediction for mental illnesses is improved to a certain extent, while also distinguishing patients. Furthermore, the embodiments of this invention have good robustness and can be used to extract important features from functional networks related to complex brain diseases, revealing the physiological and pathological mechanisms associated with symptoms and cognitive impairment. Attached Figure Description

[0030] Figure 1 This is a flowchart illustrating the brain network construction method of the present invention.

[0031] Figure 2 This is a schematic diagram illustrating the accuracy of the brain network construction method based on AAL maps in predicting positive / negative symptoms, total symptom scores, and cognitive scores for schizophrenia on the BSNIP1 dataset.

[0032] Figure 3 This diagram illustrates the accuracy and confusion matrix of the brain network construction method based on AAL maps in this invention for distinguishing between schizophrenia and normal individuals on the BSNIP1 dataset.

[0033] Figure 4 This is a schematic diagram illustrating the accuracy of the brain network construction method based on BA atlas in predicting positive / negative symptoms, total symptom scores, and cognitive scores of schizophrenia on the BSNIP1 dataset.

[0034] Figure 5 This diagram illustrates the accuracy and confusion matrix of the brain network construction method based on BA graphs in this invention for distinguishing between schizophrenia and normal individuals on the BSNIP1 dataset. Detailed Implementation

[0035] like Figure 1 As shown in this embodiment, a method for constructing individual-differenced brain networks based on sparse kernel matrix optimization includes the following steps:

[0036] Step S1: Using brain network atlases, extract the brain region time series from the functional magnetic resonance imaging (fMRI) images of the subjects.

[0037] In this step, the brain atlas used can be a Brainnetome Atlas (BA), an Automated Anatomy Labeling Atlas (AAL), etc. In practice, any one of these or other brain atlases can be selected.

[0038] This embodiment uses functional magnetic resonance imaging (fMRI) data of schizophrenia as an example for illustration, but the scope of protection of this invention is not limited to this. The data can be fMRI data of all mental disorders, including schizophrenia, depression, and other mental illnesses, as well as fMRI data of non-mental illnesses such as bisexuality.

[0039] In this embodiment, the preprocessing and feature extraction of functional magnetic resonance imaging (fMRI) brain imaging data are as follows: (1) head movement correction; (2) inter-slice time correction; (3) normalization to Montreal standard space, preferably resampling to 3×3×3mm; (4) regression of 6 head movement parameters and white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF); (5) smoothing using an 8mm full width half max (FWHM) Gaussian filter. Then, the brain is divided into specific regions using BA or AAL brain atlases, and the average time series of all voxels in each region is calculated to obtain the time series features of each node.

[0040] Step S2: Use similarity measurement methods to calculate the similarity of time series of brain regions of patients with mental illness and normal people respectively.

[0041] Similarity measurement methods include, but are not limited to, Euclidean distance, as well as cosine distance and other correlation calculation methods.

[0042]

[0043] Specifically, the similarity calculation process can be as follows: according to formula (1), calculate the brain region x from the time series feature matrix obtained after step S1. i and x j The square of the Euclidean distance between them is calculated, and then a scaling exponential similarity kernel is used to determine the weights of brain region similarity, where μ represents an empirically set hyperparameter, and ε... i,j This represents the hyperparameters used to eliminate scaling problems;

[0044]

[0045] Where, mean(ρ(x) i ,N)) represents x iThe average distance to each point in its neighborhood N, mean(ρ(x) j ,N)) represents x j The average distance to each point in its neighborhood N, ρ(x) i ,x j ) is x i and x j The Euclidean distance between the time series; by introducing the local scale parameter ε i,j It can more accurately reflect x i and x j This method considers the local distribution of points when calculating similarity weights within their respective local environments, thus improving the robustness and accuracy of the similarity measure. The smaller the distance, the larger the similarity weight W(i,j); the larger the distance, the smaller the similarity weight W(i,j). The exponential function ensures that the weights always remain between 0 and 1. Performing the above calculations on all schizophrenic patients and normal individuals yields the brain region similarity matrix for each subject.

[0046] Step S3: Calculate the absolute value of the difference between the brain region similarity matrix of each mental illness patient and all normal people, and use it as the brain region difference matrix D between patients and healthy control groups.

[0047] Specifically, the brain region similarity matrix of each mental illness patient is subtracted from the brain region similarity matrix of m normal individuals in turn. The absolute values ​​of the resulting m brain region difference matrices are then taken to obtain the m brain region difference matrix D between each mental illness patient and the m normal individuals.

[0048] Step S4: Calculate the top N values ​​with high similarity in each row of the v-th brain region difference matrix, and set the values ​​with low similarity to zero, as the sparse kernel matrix in the fusion process; then multiply the sparse kernel matrix by the average of the remaining m-1 brain region difference matrices, and repeat this process until all brain region difference matrices of each mental illness patient have been traversed.

[0049] Specifically, based on the brain region difference matrix obtained in step S3, according to formula (2), the sparse kernel matrix with large differences between patients and healthy controls is first calculated, that is, the brain region x is calculated. i The values ​​of the top N brain regions with high similarity are set to zero, and the sparse kernel matrix is ​​normalized.

[0050]

[0051] Then, the brain region difference matrix is ​​normalized according to formula (3), and the diagonal elements are set to 0.5, while ensuring that x i Since the sum of the similarities of all brain regions is 1, this normalization can eliminate the influence of unstable values.

[0052]

[0053] Based on the sparse kernel matrix and brain region difference matrix obtained by formulas (2) and (3), they are fused according to formula (4).

[0054]

[0055] The sparse kernel matrix contains the connections between the top N most similar brain regions in each brain region difference matrix; This represents averaging the m-1 normalized brain region difference matrices; multiplying the resulting sparse kernel matrix by the average of the remaining m-1 brain region difference matrices yields the v-th optimized brain region difference matrix P. (v) The calculation process is repeated m times until all m brain region difference matrices of each mental illness patient are traversed.

[0056] Step S5: Calculate the average value of the optimized brain region difference matrix obtained in step S4, and use it as the individual difference matrix for patients with mental illness, thereby realizing the construction of individual difference functional brain network.

[0057] This invention utilizes the Bipolar and Schizophrenia Network for Intermediate Phenotypes (BSNIP) dataset 1 (containing fMRI data from 168 schizophrenia patients and 219 normal control subjects) and dataset 2 (containing fMRI data from 397 normal control subjects) for testing. The constructed individual-difference brain network is used as a priori guide for feature selection to predict schizophrenia symptoms and cognitive scores on the BSNIP1 dataset, while simultaneously classifying patients and normal individuals.

[0058] like Figure 2 , 4 The results show a comparison of the accuracy of individual-difference functional brain network features constructed based on different brain atlases in predicting positive / negative symptoms, total symptom scores, and cognitive scores for schizophrenia on the BSNIP1 dataset. This method, based on individual-difference functional brain network features constructed from AAL and BA brain atlases, can reliably predict the multidimensional clinical manifestations of schizophrenia to a certain extent.

[0059] like Figure 3 , 5As shown, the accuracy and confusion matrix of individual-difference functional brain network features constructed based on different brain atlases in distinguishing between schizophrenia and normal individuals on the BSNIP1 dataset were compared. Each point represents the accuracy of 10-fold cross-validation, and the black dashed line represents the average accuracy of 10-fold cross-validation. In each confusion matrix, the colored cells on the diagonal (true positive rate) represent the matched true and predicted labels; the cells outside the diagonal represent the percentage of misclassification. Overall, the individual-difference functional brain network features constructed based on the AAL and BA brain atlases showed similar classification accuracy in distinguishing between schizophrenia and normal individuals.

Claims

1. A method for constructing individual-discretionary brain networks based on sparse kernel matrix optimization, characterized in that, Includes the following steps: (1) Using brain network atlas, brain region time series were extracted from normal brain imaging data of several normal subjects and abnormal brain imaging data of several abnormal subjects to obtain brain region time series of several normal subjects and brain region time series of several abnormal subjects. (2) Calculate the similarity matrix of the time series of brain regions of several normal subjects and the similarity matrix of the time series of brain regions of several abnormal subjects by using similarity measurement; (3) Calculate the absolute value of the difference between the similarity matrix of each abnormal subject and the similarity matrix of all normal subjects to obtain the difference matrix of several brain regions of the abnormal subject. (4) For each abnormal subject, the first... A set of brain region difference matrices is generated. Each row of the brain region difference matrix is ​​arranged in descending order. The first N values ​​are taken, and the rest are set to zero. This yields the Nth value of the abnormal subject. The sparse kernel matrix will be the th sparse kernel matrix. Multiplying the sparse kernel matrix by the average of the remaining brain region difference matrices yields the th . This step yields an optimized brain region difference matrix for all abnormal subjects. (5) Calculate the average value of all optimized brain region difference matrices for each abnormal subject to obtain the individual difference brain network of the abnormal subject.

2. The method for constructing individual-differenced brain networks according to claim 1, characterized in that, In step (4), the first The sparse kernel matrix is ​​multiplied by the average of the other normalized brain region difference matrices to obtain the _th ... The optimized brain region difference matrix for each brain region.

3. The method for constructing individual-differenced brain networks according to claim 2, characterized in that, The first An optimized brain region difference matrix for: in, Indicates the first A sparse kernel matrix, Indicates the first A normalized brain region difference matrix, This represents the total number of normal subjects.

4. The method for constructing individual-differenced brain networks according to claim 3, characterized in that, The similarity calculation formula is as follows: in, Indicates the brain regions of the subjects and brain regions Similarity values ​​between time series Indicates the brain regions of the subjects and brain regions The square of the Euclidean distance between time series. This represents the hyperparameters used for empirical settings. This represents the hyperparameters used to eliminate scaling issues.

5. The method for constructing individual-differenced brain networks according to claim 4, characterized in that, Step (4) specifically includes: S in, Represents the sparse kernel matrix S with the th Line 1 The value of the column; Brain region difference matrix The Middle Line 1 The values ​​in the column represent the abnormal subjects. Compared with normal subjects The absolute value of the difference between them; Brain region difference matrix The Middle Line 1 The value of the column.

6. The method for constructing individual-differenced brain networks according to claim 5, characterized in that, Normalize the brain region difference matrix: in, This represents the normalized matrix after applying the brain region difference matrix D. Represents the normalized brain region difference matrix The Middle Line number The value of the column.

7. The method for constructing individual-differenced brain networks according to claim 1, characterized in that, The normal brain imaging data and abnormal brain imaging data are structural magnetic resonance imaging data from a publicly available data center. They are preprocessed to obtain preprocessed images of normal brain region data and abnormal brain region data. The obtained preprocessed images are then divided into brain regions, and the average time series of all voxels in each brain region is calculated.

8. The method for constructing individual-differenced brain networks according to claim 1, characterized in that, Brain regions are divided based on preprocessed brain imaging data using brain atlases, which include AAL atlases and BA atlases.

9. A computer 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 according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

Images