High-order individual difference brain network construction method based on triple
Through the triple-based high-order individual differential brain network construction method, the relative distance between brain intervals and the weighted network is calculated, which solves the problem of insufficient accuracy of brain networks in the existing technology, and achieves higher individualized diagnosis and treatment Prediction accuracy.
Patent Information
- Application Number
- CN202411003589.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-25
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-07-25
AI Technical Summary
The prior art is difficult to accurately construct a brain network that reflects individual differences, and cannot effectively capture higher-order information about brain area interactions, resulting in insufficient accuracy in individualized diagnosis and treatment prediction.
The high-order individual differential brain network construction method based on triplets is adopted. By obtaining brain area data, calculating the relative distance between the two brain areas, constructing a reference network and an disturbance network, and weighted combinations are performed to obtain the high-order individual differential brain network. The accuracy of the network is improved by using the European distance formula and z-fraction standardization.
It significantly improves the accuracy of the brain network, can better reflect the functional attributes of the brain's multi-brain area interaction, enhances the accuracy of individualized diagnosis and treatment prediction, simulates real brain activities and effectively reduces dimensionality.
Smart Images

Figure CN119027680A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to brain network construction, and in particular to a method for constructing a high-order individual difference brain network based on triples. Background Art
[0002] The human brain is an extremely complex system in which neurons, neuronal clusters or regions are interconnected to form a complex network. The execution of various advanced cognitive functions of the brain depends on the cooperation of different brain regions, and the pathogenesis of many mental illnesses is also related to the connectivity disorders of brain regions. Therefore, neuroscientists try to use brain imaging data of different modalities to draw brain connectivity maps. Commonly used brain imaging data include structural magnetic resonance imaging (sMRI) that can measure the organizational structure information of the brain such as gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF), diffusion magnetic resonance imaging (dMRI) that describes fiber connections, and functional magnetic resonance imaging (fMRI) that reflects brain activity. Compared with fMRI and dMRI data, sMRI is relatively low in cost, more suitable for large-scale research, and has more consistent image quality. At the same time, previous studies have found that sMRI data can provide brain connectivity information. The inter-regional cortical thickness correlation map obtained using sMRI is very similar to the fiber bundle imaging obtained by dMRI, indicating that morphological correlation may be related to the anatomical connectivity of the brain. Traditional group-level research can only construct a unified network for the entire subject group and is limited to statistical analysis at the group level. Although this method can reveal statistical differences in the mean of brain networks, it cannot capture differences between individuals, which are ubiquitous in brain development. In the field of precision medicine, building brain networks that reflect individual differences will help to deeply explore biological characteristics at the individual level and improve the accuracy of personalized diagnosis and treatment prediction for individual subjects.
[0003] In order to study individual variability and identify abnormal brain morphology in patients, more and more studies have begun to focus on morphological networks constructed at the individual level. These methods can be divided into individual morphological brain networks based on single metrics and multiple metrics, such as: dividing three-dimensional regions to calculate similarities, constructing individual brain morphological networks based on similarities in gray matter morphology, mainly by dividing the individual brain into multiple small cubes and calculating the similarity of gray matter density between these cubes, and discovering significant small-world properties in the network. However, the nodes in this network have changed from atlas-based regions to voxel-based cubes, and these strictly extracted cubes may not correspond well to specific areas of brain anatomy or function. A set of normal control structural covariance networks are constructed by calculating the Pearson correlation coefficient between the gray matter volumes of each region, and then a perturbation structural covariance network is constructed by adding one patient to the control group each time; the individual structural covariance network of each patient is represented by the difference between the perturbation network and the reference network.
[0004] Based on multiple metrics-based individualized morphological brain networks, individual morphological networks based on multiple indicators can be divided into unimodal and multimodal networks according to the imaging method. For example, unimodal networks: In the prior art, a method based on Pearson correlation is used to estimate the similarity between seven morphological measurement features of different regions. This method concatenates the seven indicators of the number of vertices, mean curvature and Gaussian curvature, surface area, gray matter volume, and mean and standard deviation of cortical thickness of each region into a feature vector, and uses the Pearson correlation coefficient between the feature vectors of any pair of regions to estimate the morphological connectivity between regions. The network has a repeatable small-world property. For example, multimodal networks: Combining structural magnetic resonance imaging with diffusion magnetic resonance imaging, a multimodal brain morphological network based on three morphometric features of cortical curvature, myelin markers, and gray matter morphology is constructed to provide a multidimensional description of brain morphological features. However, the above individual brain network construction methods only calculate the first-order correlation between two brain regions, and cannot reflect the high-order information of brain region interaction. The brain is a highly complex network system. The interactions between different brain regions are not just simple two-by-two interactions. The above brain network construction method is not accurate enough. Summary of the invention
[0005] Purpose of the invention: In view of the above shortcomings, the present invention provides a method for constructing a high-order individual difference brain network based on triples to improve the accuracy of the brain network.
[0006] Technical solution: To solve the above problems, the present invention adopts a method for constructing a high-order individual difference brain network based on triples, comprising the following steps:
[0007] (1) Obtain brain region data from several subjects;
[0008] (2) Obtain the relative distances between each pair of brain regions in the subject's brain region data, use different brain regions as rows and columns, and construct a reference network based on the triplets of rows, columns, and relative distances between each pair of brain regions;
[0009] (3) Eliminate subject G, obtain the relative distances between each brain region in the subject's brain region data after excluding subject G, and construct the perturbation network of subject G based on the rows and columns of different brain regions and the triplets of relative distances;
[0010] (4) Perform a weighted combination of the reference network and the perturbation network of subject G to obtain the high-order individual difference brain network of subject G.
[0011] Furthermore, the relative distances between each pair of brain regions in the subjects' brain region data were calculated according to the Euclidean distance formula.
[0012] Furthermore, the calculation formula of the relative distance in the reference network is:
[0013]
[0014] in, represents the distance between brain region i and brain region v relative to brain region u, x im represents the i-th brain region of the m-th subject, x vm represents the vth brain region of the mth subject, x um represents the u-th brain region of the m-th subject, and N represents the number of subjects; e (ω) (x i ,x v ) is the distance between brain region i and brain region v relative to the k brain regions closest to brain region i. The values of v and u are the numbers of the first k brain regions closest to brain region i. The reference network e is constructed based on the distances of all brain regions. (ω) .
[0015] Furthermore, the calculation formula for the relative distance in the disturbance network is:
[0016]
[0017] in, represents the distance between brain area i and brain area v relative to brain area u after subject G is eliminated, e (ω-G) (x i ,x v ) represents the distance between brain region i and brain region v relative to the k brain regions closest to brain region i after subject G is removed. The perturbation network e is constructed based on the distances of all brain regions. (ω-G) .
[0018] Furthermore, the high-order individual difference brain network e of subject G (G) for:
[0019] e (G) =β(e (ω) -e (ω-G) )+α(e (ω-G) )
[0020] Among them, e (ω) represents the reference network, e (ω-G) represents the perturbation network of subject G, and β and α represent weight coefficients.
[0021] Furthermore, the step (4) also includes the high-order individual difference brain network e( G ) to perform z-score standardization:
[0022]
[0023] Among them, e (G) (x i ,x v ) is the high-order individual difference brain network e (G) The edge from midbrain region i to brain region v, Z iv for e (G) (x i ,x v ) after z-score standardization, high-order individual difference brain network e (G) The network after z-score standardization is Z, μ v Representing high-order individual difference brain networks (G) The mean of the vth brain region, σ v Representing high-order individual difference brain networks (G) The standard deviation of the vth brain region.
[0024] Furthermore, the brain region data includes normal brain region data and abnormal brain region data, and the normal brain region data and the abnormal brain region data are structural magnetic resonance imaging data from a public data center, which are preprocessed separately to obtain preprocessed images of the normal brain region data and the abnormal brain region data; the obtained preprocessed images are divided into brain regions, each divided into N brain regions, and the gray matter value feature vector of each brain region in each data is calculated.
[0025] Furthermore, the preprocessed brain image data is divided into brain regions according to a brain atlas, and the brain atlas includes an AAL atlas.
[0026] The present invention also adopts 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.
[0027] The present invention also adopts a computer-readable storage medium on which a computer program is stored, characterized in that the computer program implements the steps of the above method when executed by a processor.
[0028] Beneficial effects: Compared with the prior art, the significant advantage of the present invention is that it obtains high-order interactions between brain regions through triplets of brain region distances, better reflecting the characteristics of multi-brain region interactions in the real brain to achieve its functional attributes, thereby better understanding the working principles of the brain and improving the accuracy of the constructed brain network. The use of perturbation networks to construct individualized difference networks helps to explore biological characteristics at the individual level and improve the accuracy of personalized diagnosis and treatment predictions for single subjects. Since brain regions usually interact with their neighbors rather than distant brain regions, only brain regions close to the brain region in question are considered for triple calculations, which not only simulates real brain activity but also effectively reduces the dimensionality of features. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 Schematic diagram of the process of constructing a midbrain network according to the present invention.
[0030] Figure 2 This is a specific flow chart of the method for constructing a midbrain network of the present invention.
[0031] Figure 3 This is a scatter plot of the actual score and the predicted score obtained by the midbrain network construction method of the present invention on the BSNIP1 and BSNIP2 data sets after partial least squares regression of the total cognitive score. DETAILED DESCRIPTION
[0032] like Figure 1 As shown, a method for constructing a high-order individual difference brain network based on triples in this embodiment includes the following steps:
[0033] Step S1: Extracting gray matter values of sMRI brain regions of healthy subjects and schizophrenia subjects based on brain maps.
[0034] The brain atlas used may be a Brainnetome Atlas, an Automated Anatomy Labeling Atlas (AAL), an Allen Brain Atlas, etc. In this embodiment, an AAL template that divides the brain into 90 brain regions (excluding the cerebellum) is used to construct an individual difference brain network, and in actual applications, one or more atlases can be selected for implementation according to needs.
[0035] In this embodiment, only sMRI data of schizophrenia and its corresponding normal person are used as examples for illustration, but the protection scope of the present invention is not limited thereto. The data can be sMRI data of all mental disorders, including schizophrenia, depression and other mental illnesses, as well as sMRI data of non-mental diseases such as bipolar disorder.
[0036] In this embodiment, gray matter is extracted from the structural magnetic resonance image sMRI as its feature. The preprocessing and feature extraction of sMRI are specifically as follows: (1) standardization to the Montreal standard space, preferably, resampling to 3×3×3 mm; (2) segmentation into white matter, gray matter, and cerebrospinal fluid; (3) gray matter is smoothed using an 8 mm half-maximum full-width Gaussian filter kernel; (4) subject field points are detected to ensure that all subjects are correctly segmented; (5) the standard spatial coordinates of the AAL template are obtained, and the preprocessed gray matter image is aligned to the AAL template space using a linear alignment method; (6) the aligned gray matter image is divided into 90 brain regions to ensure that each voxel is correctly assigned to the corresponding brain region; (7) the voxel values in each brain region are averaged to obtain the gray matter value feature vector for each subject.
[0037] After the above processing, the structural magnetic resonance imaging (sMRI) data was converted into gray matter value features, and a 90×N feature matrix A was obtained, where N is the number of subjects. The feature matrix of healthy subjects is recorded as A1, and the number of subjects is recorded as N1; the feature matrix of schizophrenia subjects is recorded as A2, and the number of subjects is recorded as N2.
[0038] Step S2: Calculate the relative distances between the two groups of brain regions according to the Euclidean distance formula to obtain the reference networks corresponding to the two groups of subjects.
[0039] Specifically, taking healthy subjects as an example, the process of calculating the reference network includes:
[0040] (1) Calculate the Euclidean distance between any two rows of the feature matrix and sort them in ascending order to obtain a brain region number matrix N of size 90 × 90 k . Select the sequence matrix N k The first k elements in the i-th row (i=1,2,...,90) are denoted as n k In this embodiment, k is 5;
[0041] (2) Select the i-th row vector, u-th row vector, and v-th row vector in the gray matter value feature matrix A1, and denote them as x i ,x u ,x v , where u,v∈n k . i ,x u ,x v Substitute the following formula:
[0042]
[0043] The distance between brain area i and brain area v relative to brain area u can be obtained Among them, x *m represents the *th brain region of the mth subject;
[0044] (3) The triple matrix S is transformed into (ω) Summing over the dimension u:
[0045]
[0046] The reference network e1 corresponding to the healthy subjects can be obtained (ω) Among them, e1 (ω) (x i ,x v ) represents the distance between brain region i and brain region v relative to the k brain regions closest to brain region i.
[0047] The same operation was performed on the schizophrenia subjects to obtain the reference network e2 corresponding to the schizophrenia subjects. (ω) , where the brain region number matrix N is used to calculate the reference network corresponding to the schizophrenia subjects. k The matrix is the same as that used for healthy subjects and will not be recalculated.
[0048] Step S3: First, for the healthy subject group, remove subject G1 (G1=1, 2, ..., N1) in turn, and repeat the operation in step S2 to obtain the perturbation network corresponding to each healthy subject. Then, for the schizophrenia group, remove subject G2 (G2=1, 2, ..., N2) in turn, and repeat the operation in step S2 to obtain the perturbation network corresponding to each schizophrenia patient.
[0049] Taking healthy subjects as an example, the process of calculating the perturbation network is as follows: first, delete the G1th column in the feature matrix A1, where G1 = 1, 2, ..., N1, and obtain the matrix Will Substituting into step S2(2)(3), we can get the perturbation network corresponding to each healthy subject. The above operation was also performed on the schizophrenia subjects to obtain the perturbation network corresponding to the schizophrenia subjects.
[0050] Step S4: The reference networks of the two groups of subjects are weightedly combined with the perturbation network of each subject to obtain a triplet-based high-order individual difference brain network.
[0051] Taking healthy subjects as an example, the specific steps of constructing a high-order individual difference brain network based on triples include: first, the reference network e1 (ω) With perturbation network Substitute the following formula:
[0052]
[0053] The individual difference triplet high-order brain network corresponding to each healthy subject can be obtained. represents the individual difference network corresponding to the G1th subject; β represents the reference network e1 (ω) With perturbation network The weight coefficient of the difference; α represents the perturbation network The weight coefficients are, in this embodiment, β is 15 and α is 0.5.
[0054] Finally, the scores of the obtained individual difference high-order brain networks were standardized according to the following formula:
[0055]
[0056] in, Brain Network The edge from midbrain region i to brain region v, μ v Individual Difference Network The mean of the vth brain region; v Individual Difference Network The standard deviation of the vth brain region. Z 1 Represents the standardized individual difference triplet high-order brain network corresponding to healthy subjects, Representing individual differences in triplet high-order brain network Z 1 edge.
[0057] The above operation was also performed on the schizophrenia subjects, and finally the individual difference triplet high-order brain network Z corresponding to each schizophrenia subject was obtained. 2 .
[0058] The embodiment of the present invention is tested using the Bipolar and Schizophrenia Network for Intermediate Phenotypes (BSNIP) dataset (BSNIP1 contains structural magnetic resonance imaging (sMRI) data of 163 schizophrenia subjects and 211 normal subjects, and BSNIP2 contains structural magnetic resonance imaging (sMRI) data of 282 schizophrenia subjects and 421 normal subjects). Partial least squares regression is used as the prediction model, and the total cognitive score, digital sorting task score, symbol encoding task score, marking movement task score, Tower of London task score, language fluency task score, and language memory task score are used as prior information.
[0059] like Figure 3 As shown in the figure, on the BSNIP1 and BSNIP2 datasets, the scatter plots of the true score and the predicted score are obtained after partial least squares regression of the total cognitive score. The results show that the proposed method has good robustness for predicting the total cognitive score of different datasets.
[0060] As shown in Table 1, the method proposed in this embodiment is compared with the whole-brain voxel dimensionality reduction features, brain region average gray matter value features, and KL divergence features based on the BSNIP1 dataset to predict the accuracy of seven cognitive scores: total cognitive score, digital sorting task, symbol encoding task, marking movement task, Tower of London task, language fluency task, and language memory.
[0061] Table 1 Comparison of the prediction accuracy of the four methods for seven cognitive scores
[0062]
[0063] The results show that this method has the highest prediction accuracy among the four methods, indicating that the high-order individual difference brain network features proposed by this method can better capture the characteristic information related to cognitive scores.
[0064] In summary, this method has high prediction accuracy on different data sets, and in comparison with different brain network construction methods, this method has the highest prediction accuracy for different cognitive scores, indicating that the high-order individual difference brain network features proposed by this method have good robustness on different data sets and can more accurately reveal the characteristic information related to cognitive scores.
Claims
1. A method for constructing a high-order individual difference brain network based on triples, It is characterized in that The following steps are involved: (1) Obtain brain region data from several subjects; (2) Obtain the relative distances between each pair of brain regions in the subject's brain region data, use different brain regions as rows and columns, and construct a reference network based on the triplets of rows, columns, and relative distances between each pair of brain regions; (3) Eliminate subject G, obtain the relative distances between each brain region in the subject's brain region data after excluding subject G, and construct the perturbation network of subject G based on the rows and columns of different brain regions and the triplets of relative distances; (4) Perform a weighted combination of the reference network and the perturbation network of subject G to obtain the high-order individual difference brain network of subject G.
2. The method for constructing a high-order individual difference brain network according to claim 1, It is characterized in that The relative distance between each pair of brain regions in the subjects' brain region data was calculated according to the Euclidean distance formula.
3. The method for constructing a high-order individual difference brain network according to claim 2, It is characterized in that The calculation formula of the relative distance in the reference network is: in, represents the distance between brain region i and brain region v relative to brain region u, x im represents the i-th brain region of the m-th subject, x vm represents the vth brain region of the mth subject, x um represents the u-th brain region of the m-th subject, and N represents the number of subjects; e (ω) (x i ,x v ) is the distance between brain region i and brain region v relative to the k brain regions closest to brain region i. The values of v and u are the numbers of the first k brain regions closest to brain region i. The reference network e is constructed based on the distances of all brain regions. (ω) .
4. The method for constructing a high-order individual difference brain network according to claim 3, It is characterized in that The calculation formula of the relative distance in the perturbation network is: in, represents the distance between brain area i and brain area v relative to brain area u after subject G is eliminated, e (ω-G) (x i ,x v ) represents the distance between brain region i and brain region v relative to the k brain regions closest to brain region i after subject G is removed. The perturbation network e is constructed based on the distances of all brain regions. (ω-G) .
5. The method for constructing a high-order individual difference brain network according to claim 4, It is characterized in that The high-order individual difference brain network e( G )for: And (G) =β(e (ω) -And (ω-G) )+α(e (ω-G) ) Among them, e (ω) represents the reference network, e (ω-G) represents the perturbation network of subject G, and β and α represent weight coefficients.
6. The method for constructing a high-order individual difference brain network according to claim 5, It is characterized in that The step (4) also includes the high-order individual difference brain network e( G ) to perform z-score standardization: Among them, e (G) (x i ,x v ) is the high-order individual difference brain network e (G) The edge from midbrain region i to brain region v, Z iv for e (G) (x i ,x v ) after z-score standardization, high-order individual difference brain network e (G) The network after z-score normalization is Z, μ v Representing high-order individual difference brain networks (G) The mean of the vth brain region, σ v Representing high-order individual difference brain networks (G) The standard deviation of the vth brain region.
7. The method for constructing a high-order individual difference brain network according to claim 1, It is characterized in that The brain region data include normal brain region data and abnormal brain region data, and the normal brain region data and the abnormal brain region data are structural magnetic resonance imaging data from a public data center, which are preprocessed respectively to obtain preprocessed images of the normal brain region data and the abnormal brain region data; the obtained preprocessed images are divided into brain regions, each divided into N brain regions, and the gray matter value feature vector of each brain region in each data is calculated.
8. The method for constructing a high-order individual difference brain network according to claim 1, It is characterized in that The preprocessed brain image data is divided into brain regions according to a brain atlas, which includes an AAL atlas.
9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.
Citation Information
Patent Citations
Multi-modal brain network feature classification method based on multiple data centers
CN116433964A
Class general perturbation face image privacy protection method based on triple constraint
CN117763612A
Feature classification method and system based on individual difference brain network
CN118038176A
Unified representation calculation method and apparatus for brain network, and electronic device and storage medium
WO2024119337A1
