Sea member occupational plastic brain network identification method based on effective connection bias

Abstract

The invention relates to a method for identifying a seaman occupational plastic brain network based on effective connection bias, and the method comprises the following steps: S1, obtaining and preprocessing brain image data, and extracting a time sequence of a large-scale brain network corresponding to each subject; s2, calculating an effective connection matrix between the brain networks; s3, calculating left half brain network bias, right half brain network bias and whole brain network effective connection bias; s4, performing inter-group comparison to obtain a large-scale whole-brain network and half-brain network set with significant difference; and S5, taking an intersection of the large-scale whole-brain network and the half-brain network set with significant difference to obtain a psychological sub-health network and a seaman occupational plastic brain network. Compared with the prior art, the method has the advantages that the recognition precision of the occupational plasticity brain network and the psychological sub-health state brain network of the seamen is improved, and the like.

Description

Technical Field

[0001] The present invention relates to, in particular to, a method for identifying seafarers' occupational plasticity brain network based on effective connection lateralization. Background Art

[0002] In recent years, functional magnetic resonance imaging (fMRI) has gradually become an emerging neuroimaging technology. Using this technology, brain activity can be monitored to determine the cognitive, emotional and behavioral characteristics of individuals, and to explore the differences in brain function between seafarers and non-seafarers. The differences in brain function among seafarers are based on the influence of long-term work and environment. The working conditions of seafarers are significantly different from those of the environment and land, so they are easily affected by complex factors such as the natural environment and work pressure, resulting in adverse psychological conditions, which in turn affects the safety of navigation operations. In addition, this special profession has the characteristics of high risk, great hardship, great responsibility, strong professionalism and strict discipline, which places higher demands on the quality of seafarers. Therefore, the differences in seafarers' brain functional connectivity patterns are affected by the combined effects of mental health status and occupational plasticity.

[0003] Lateralization of brain function refers to the incomplete symmetry of left and right brain functions. Since the two hemispheres of the brain have a certain division of labor in function, there is a phenomenon of specialization or asymmetry. Common lateralization calculation methods are generally based on functional connectivity or brain entropy calculation, and have not yet adopted effective connections with directionality that can represent causal effects for analysis. The traditional Granger causal algorithm predicts the time series of regions of interest (ROIs) based on the vector autoregression method, which has the disadvantages of being unable to capture the nonlinear relationship between ROIs, long running time and low efficiency. In addition, in fMRI data, the number of ROIs is generally greater than the length of the time series, which will greatly affect the vector autoregression solution.

[0004] Existing technologies are unable to determine whether the differences in seafarers' brain functional connectivity patterns are entirely due to occupational plasticity or whether the differences in connectivity patterns are indeed caused by poor mental health. The recognition accuracy of seafarers' occupational plasticity brain networks and psychological sub-health brain networks is not high. Summary of the invention

[0005] The purpose of the present invention is to provide a method for identifying seafarers' occupational plasticity brain networks based on effective connection lateralization in order to improve the recognition accuracy of seafarers' occupational plasticity brain networks and psychological sub-health state brain networks.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A method for identifying seafarers' occupational plasticity brain network based on effective connection lateralization, the method comprising the following steps:

[0008] S1. Acquire brain imaging data, the data including three groups: a seaman subject group, a non-seaman brain network normal control group, and a non-seaman brain network abnormal control group; preprocess the brain imaging data, and extract the brain network time series of each subject from the preprocessed brain imaging data according to a large-scale brain network template;

[0009] S2. Divide the time series into a test set and a training set. Train the Transformer neural network on the training set, input the test set into the trained neural network, output the predicted value, and calculate the difference between the predicted value and the original time series. Combined with the Granger causal calculation formula, the matrix of effective connections between each brain network is obtained.

[0010] S3, taking the mean of the effective connection matrices within the three groups as the threshold, thresholding the effective connection matrix, calculating the left brain network lateralization, the right brain network lateralization and the whole brain network effective connection lateralization based on the thresholded effective connection matrix, and taking the left brain lateralization, the right brain lateralization and the whole brain network effective connection lateralization as features;

[0011] S4. Based on the results of the normality test and variance homogeneity test between the groups, a two-sample T test / Wilcoxon rank sum test and multiple comparison correction were performed, and inter-group comparison was performed to obtain two brain networks with significant differences;

[0012] S5. The intersection of the two significantly different brain networks was taken to obtain the psychological sub-health network, and based on the psychological sub-health network, the seafarers' occupational plasticity brain network was obtained.

[0013] Furthermore, the elements of the effective connection matrix are:

[0014]

[0015] Among them, G i→j represents the effective connection value of brain network i to j, where i and j are different brain networks of the same subject, σ represents the variance of δ, and δ represents the estimation error.

[0016] Furthermore, the effective connection matrices in the three groups are averaged, and a threshold is determined based on the average, and elements in the effective connection matrix that are smaller than the threshold are set to 0 to obtain a thresholded effective connection matrix.

[0017] Further, the left hemisphere lateralization is:

[0018]

[0019] Among them, η is the number of brain networks, e is the left hemisphere;

[0020]

[0021] Among them, SumLL(e) represents the effective connection sum of all left hemisphere brain networks between e and itself, EC represents the effective connection matrix, and α is the standard deviation of each group of effective connections.

[0022] Further, the right hemisphere lateralization is:

[0023]

[0024] Among them, brain networks For the right brain:

[0025]

[0026] Furthermore, the whole-brain network effective connection lateralization is:

[0027]

[0028] in, is the sum of the effective connections between the left brain network e and all brain networks, For the right brain network The sum of effective connections to all brain networks.

[0029] Furthermore, the specific steps of performing the two-sample T test / Wilcoxon rank sum test based on the results of the normality test and the variance homogeneity test between the characteristics of the groups are as follows:

[0030] The characteristics between groups were tested for normality and variance homogeneity. If they met the normal distribution and variance homogeneity, the two-sample T test was used. Otherwise, the Wilcoxon rank sum test was used.

[0031] Furthermore, the specific steps of performing inter-group comparison to obtain two brain networks with significant differences are:

[0032] Inter-group comparison was performed based on the sample differences obtained after two-sample T test or Wilcoxon rank sum test and multiple comparison correction. The brain networks with significant differences between the seafarer subject group and the non-seafarer normal brain network control group were recorded as SailROIs, and the brain networks with significant differences between the non-seafarer abnormal brain network control group and the non-seafarer normal brain network control group were recorded as DepROIs.

[0033] Furthermore, the psychological sub-health network is PsyROIs=SailROIs∩DepROIs.

[0034] Furthermore, the seafarer career plasticity brain network is CareerROIs=SailROIs-PsyROIs.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The present invention can effectively capture the nonlinear relationship between brain network time series by introducing the Transformer model, and the addition of the self-attention mechanism in the model allows the model to capture dependencies in input sequences at different positions, and can effectively process time series. At the same time, it constructs a lateralization index suitable for effective connection and analyzes the lateralization between the left and right brain networks, which can more accurately identify the seafarers' occupational plasticity brain network and psychological sub-health state brain network, thereby quantitatively and objectively evaluating mental health earlier, and accurately evaluating the quality and plasticity of seafarers, and guiding selection and training. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0038] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0039] At present, it is still impossible to determine whether the differences in seafarers' brain functional connectivity patterns are entirely due to occupational plasticity. In order to further determine the differences, the present invention proposes a method for identifying seafarers' occupational plasticity brain networks based on effective connection lateralization, introduces data from a non-seafarer brain network abnormality control group, and subdivides seafarers' differential brain function areas by matching the differential brain networks related to the non-seafarer brain network abnormality control group, thereby identifying psychological sub-health brain networks and occupational plasticity brain networks. This allows for earlier quantitative and objective assessments of mental health, as well as accurate assessments of seafarers' qualities and plasticity, to guide selection and training. The flow chart of the method of the present invention is shown below. Figure 1 shown.

[0040] The method of the present invention comprises the following steps:

[0041] S1. Acquire brain imaging data, the data including three groups: a seaman subject group, a non-seaman brain network normal control group, and a non-seaman brain network abnormal control group; pre-process the brain imaging data, and extract the brain network time series of each subject according to a large-scale brain network template;

[0042] S2. Divide the time series into a test set and a training set. Train the Transformer neural network on the training set, input the test set into the trained neural network, output the predicted value, and calculate the difference between the predicted value and the original time series. Combined with the Granger causal calculation formula, the matrix of effective connections between each brain network is obtained.

[0043] S3, taking the mean of the effective connection matrices within the three groups as the threshold, thresholding the effective connection matrix, calculating the left brain network lateralization, the right brain network lateralization and the whole brain network effective connection lateralization based on the thresholded effective connection matrix, and taking the left brain lateralization, the right brain lateralization and the whole brain network effective connection lateralization as features;

[0044] S4. Based on the results of the normality test and variance homogeneity test between the groups, a two-sample T test / Wilcoxon rank sum test and multiple comparison correction were performed, and inter-group comparison was performed to obtain two brain networks with significant differences;

[0045] S5. The intersection of the two significantly different brain networks was taken to obtain the psychological sub-health network, and based on the psychological sub-health network, the seafarers' occupational plasticity brain network was obtained.

[0046] In S1, brain imaging data are preprocessed to extract large-scale brain network time series;

[0047] In S2, the Transformer-Granger causal model was constructed to calculate the effective connections between large-scale brain networks for each subject;

[0048] In S3, the lateralization values ​​between whole-brain networks and between hemi-brain networks are calculated as features;

[0049] In S4, the features are used to conduct inter-group significance tests and corrections on the subjects to obtain brain networks with significant differences;

[0050] In S5, the brain networks with differences between the groups were further compared, similar brain networks were matched, and large-scale occupational plasticity brain networks were further identified.

[0051] Step S2 specifically includes:

[0052] S21, divide the time series of each brain network into a test set and a training set;

[0053] S22. Use Transformer neural network to replace vector autoregression and train on the time series training set:

[0054] x j (t)=Transformer(X(t-lag))+δ j (t)

[0055] x j\i (t) = Transformer(X \i (t-lag))+δ j\i (t)

[0056] Among them, x j= {x(1), x(2), ..., x(t), ..., x(N)} represents the time series of brain network j, X = {x1, ..., x i ,...,x j ,...,x η} includes all brain network time series, \i means excluding x i , Transformer(·) is the Transformer model, lag is the time delay, and δ(t) is the estimation error.

[0057] S23. On the test set, perform time series prediction based on the training model to obtain a predicted value, and calculate the difference between the predicted value and the original time series;

[0058] S24. Calculate the effective connection value between brain networks i and j:

[0059]

[0060] G i→j Represents the effective connection value of brain network i to j, σ represents the variance of δ, and the effective connection values ​​between all brain networks are saved in the matrix EC. When performing effective connection calculation in the present invention, the causal relationship between all brain networks of a subject is calculated, so i and j are two brain networks of the same subject, and the size of the effective connection matrix required is (η,η).

[0061] Step S3 specifically includes:

[0062] S31. Calculate the mean of all effective connections of all subjects in each group as the threshold, and threshold the effective connections to retain strong effective connections;

[0063] S32. To quantify the lateralization property of the brain, the lateralization index of the whole brain network and the lateralization index between the hemibrain networks were calculated based on the S31 thresholded effective connection matrix; the hemibrain network lateralization index and the whole brain network lateralization index were defined respectively to measure the lateralization of the brain's effective connections; among them, the lateralization formula of the left hemibrain network e is:

[0064]

[0065] Among them, η is the number of brain networks corresponding to each subject, and e is the left hemisphere brain network;

[0066] is the sum of the effective connections between e and all left hemisphere brain networks (excluding itself); to keep the sum consistent, the sum of the effective connections between e and all right hemisphere brain networks is excluded when calculating the sum of the effective connections between e and the corresponding right brain network The effective connection between α is the standard deviation of the effective connection of each group, aiming to reduce the differences between subjects within the group. is located in the right hemisphere, and right hemisphere lateralization is defined as:

[0067]

[0068] Similarly, there are:

[0069]

[0070] halfL(e) is a lateralization index, which is used to measure the lateralization degree of effective connections of brain network e, and -1≤halfL≤1. If halfL>0, it means that the function of the corresponding brain network has a left-lateralized performance, that is, the functional complexity of the left brain network domain is greater than that of the right side. HalfL<0 indicates that the functional complexity of the right brain is higher. In addition, if the value of halfL is low, it means that the corresponding asymmetry is low and the lateralization degree is small.

[0071] The lateralization of effective connections of the whole brain network is to regard the left and right hemisphere brain networks as the whole brain network E, so as to determine the lateralization degree of the brain network E in the whole brain. The calculation formula is:

[0072]

[0073] in, is the sum of the effective connections between the left brain network e and all brain networks, For the right brain network The sum of the effective connections between all brain networks. Like halfL, L<0 means left-lateralization, and vice versa, right-lateralization. Its value represents the degree of lateralization.

[0074] Step S4 specifically includes:

[0075] S41, normality test and variance homogeneity test were conducted on the characteristics of the subjects;

[0076] S42. Based on the test results, two-sample T test / Wilcoxon rank sum test and multiple comparison correction were used to explore the significant differences between the subjects in the large-scale whole-brain network and hemi-brain network groups;

[0077] Step S5 specifically includes:

[0078] S51, comparing the two different brain networks obtained after the significance test of the normal brain network control group of seafarers and non-seafarers, the abnormal brain network control group of non-seafarers and the normal brain network control group of non-seafarers, and taking their intersection, which is the psychological sub-health network;

[0079] S52. In the seafarers' difference network, the psychological sub-health network that is the same as the difference network of the non-seafarers' brain network abnormality control group can be considered as the seafarers' occupational plasticity brain network.

[0080] In the above steps, each of the three groups of data in the control group has multiple subjects, and each subject corresponds to η brain networks.

[0081] Compared with the prior art, the above invention has at least the following advantages or beneficial effects:

[0082] 1. A Transformer-based calculation method for Granger causality is proposed. By introducing the Transformer model, the nonlinear relationship between brain network time series can be effectively captured. The addition of the self-attention mechanism in the model allows the model to capture dependencies in input sequences at different positions, which can effectively process time series.

[0083] 2. Based on the traditional lateralization calculation method, lateralization indicators suitable for effective connections are constructed in two areas: the hemispheric brain network and the whole brain network. The lateralization between the left and right brain networks provides a new perspective for analyzing the functional connectivity pattern of seafarers' brains.

[0084] 3. The Transformer-Granger model and the calculation method of the effective connection lateralization index are not only applicable to the exploration of seafarers' brain connectivity patterns, but also to the identification of connectivity patterns in other occupations.

[0085] The following is an actual experimental analysis. Based on the above method, the following steps are performed:

[0086] Step 1: preprocess the obtained brain image data of the subjects and extract the large-scale brain network time series;

[0087] In step 1, a resting state functional magnetic resonance imaging data set is collected, including 87 seafarer group subjects, 113 non-seafarer brain network normal control group and 51 non-seafarer brain network abnormal control group patients. The data is preprocessed, including removing the first several time points, head motion correction, time layer correction, spatial standardization to EPI template, smoothing and other reasonable preprocessing steps can be used. After data preprocessing, the time series of the brain network is extracted according to the large-scale brain network template (the template used in this example contains 14 brain networks, 7 for the left and right brains).

[0088] Step 2: According to the brain network time series extracted in step 1, the constructed Transformer-Granger model is used to calculate the effective connection matrix of the brain network;

[0089] In a specific implementation, step 2 includes:

[0090] Step 2.1, divide the time series length into training set: test set = 8:2;

[0091] Step 2.2: Build a Transformer network model with a batch size of 14, 8 heads of the multi-head attention mechanism, 2 encoder and decoder layers, a dropout value of 0.001, and a maximum number of iterations of 1000.

[0092] Step 2.3: According to the model trained in step 2.2, input the time series in the test set, obtain the predicted value, and calculate the error between the predicted value and the original value;

[0093] Step 2.4: Based on the above-mentioned error and Granger causality calculation formula, obtain the effective connection matrix of the causal relationship between the 14 brain networks.

[0094] Step 3, calculating the lateralization values ​​between whole-brain networks and between hemi-brain networks as features;

[0095] In a specific implementation, step 3 includes:

[0096] Step 3.1, average the effective connection matrices of the three groups of subjects and set them as the threshold m, retain the effective connection strength between brain networks G≥m, set G<m to 0, and obtain the threshold matrix EC(η×η), η=14;

[0097] Step 3.2, calculate the effective connection lateralization; according to the lateralization formula, calculate the hemibrain network lateralization and the whole brain network lateralization for each subject, and the results are η and

[0098] Step 4: Use the features to conduct inter-group significance tests on the subjects to obtain brain networks with significant differences;

[0099] In a specific implementation, step 4 includes:

[0100] Step 4.1, take the effective connection lateralization value of the subject as the feature, and perform normality test and variance homogeneity test;

[0101] Step 4.2: Based on the test results, two-sample T test / Wilcoxon rank sum test were used, followed by multiple comparison correction. The large-scale whole-brain network and half-brain network sets with significant differences between seafarers and non-seafarers with normal brain networks were recorded as SailROIs, and the brain networks with significant differences between non-seafarers with abnormal brain networks and non-seafarers with normal brain networks were recorded as DepROIs.

[0102] Step 5: Further compare the brain networks with differences between the groups, match similar brain networks, and further identify large-scale plasticity brain networks;

[0103] In a specific implementation, step 5 specifically includes:

[0104] Step 5.1. Compare the difference brain network sets SailROIs and DepROIs obtained after the significance test of the abnormal brain network control group of seafarers and non-seafarers with the normal brain network control group of non-seafarers, and take their intersection, which is the psychological sub-health network PsyROIs = SailROIs ∩ DepROIs, then the seafarers' career plasticity brain network is CareerROIs = SailROIs-PsyROIs.

[0105] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for identifying seafarers' occupational plasticity brain network based on effective connection lateralization, characterized in that: The method comprises the following steps: S1. Acquire brain imaging data, including three groups: a seaman subject group, a non-seaman brain network normal control group, and a non-seaman brain network abnormal control group; preprocess the brain imaging data, and extract the brain network time series of each subject according to a large-scale brain network template; S2. Divide the time series into a test set and a training set. Train the Transformer neural network on the training set, input the test set into the trained neural network, output the predicted value, and calculate the difference between the predicted value and the original time series. Combined with the Granger causal calculation formula, the matrix of effective connections between each brain network is obtained. S3, taking the mean of the effective connection matrices within the three groups as the threshold, thresholding the effective connection matrix, calculating the left brain network lateralization, the right brain network lateralization and the whole brain network effective connection lateralization based on the thresholded effective connection matrix, and taking the left brain lateralization, the right brain lateralization and the whole brain network effective connection lateralization as features; S4. Based on the results of the normality test and variance homogeneity test between the groups, a two-sample T test / Wilcoxon rank sum test and multiple comparison correction were performed, and inter-group comparison was performed to obtain two brain networks with significant differences; S5. The intersection of the two significantly different brain networks was taken to obtain the psychological sub-health network, and based on the psychological sub-health network, the seafarers' occupational plasticity brain network was obtained.

2. The method for identifying seafarers' occupational plasticity brain network based on effective connection lateralization according to claim 1 is characterized in that: The elements of the effective connection matrix are: Among them, G i→j represents the effective connection value of brain network i to j, where i and j are different brain networks of the same subject, σ represents the variance of δ, and δ represents the estimation error.

3. The method for identifying seafarers' occupational plasticity brain network based on effective connection lateralization according to claim 1 is characterized in that: The left hemisphere lateralization is: Among them, η is the number of brain networks, e is the left hemisphere; Among them, SumLL(e) represents the effective connection sum of all left hemisphere brain networks between e and itself, EC represents the effective connection matrix, and α is the standard deviation of each group of effective connections.

4. The method for identifying seafarers' occupational plasticity brain network based on effective connection lateralization according to claim 3 is characterized in that: The right hemisphere lateralization is: Among them, brain networks For the right brain:

5. The method for identifying seafarers' occupational plasticity brain network based on effective connection lateralization according to claim 1, characterized in that: The effective connectivity lateralization of the whole brain network is: in, is the sum of the effective connections between the left brain network e and all brain networks, For the right brain network The sum of effective connections to all brain networks.

6. The method for identifying seafarers' occupational plasticity brain network based on effective connection lateralization according to claim 1, characterized in that: The specific steps of performing the two-sample T test / Wilcoxon rank sum test based on the results of the normality test and variance homogeneity test between the group characteristics are as follows: The characteristics between groups were tested for normality and variance homogeneity. If they met the normal distribution and variance homogeneity, the two-sample T test was used. Otherwise, the Wilcoxon rank sum test was used.

7. The method for identifying seafarers' occupational plasticity brain network based on effective connection lateralization according to claim 6 is characterized in that: The specific steps of performing inter-group comparison to obtain two brain networks with significant differences are: Inter-group comparison was performed based on the sample differences obtained after two-sample T test or Wilcoxon rank sum test and multiple comparison correction. The brain networks with significant differences between the seafarer subject group and the non-seafarer normal brain network control group were recorded as SailROIs, and the brain networks with significant differences between the non-seafarer abnormal brain network control group and the non-seafarer normal brain network control group were recorded as DepROIs.

8. The method for identifying seafarers' occupational plasticity brain network based on effective connection lateralization according to claim 7 is characterized in that: The psychological sub-health network is PsyROIs=SailROIs∩DepROIs.

9. The method for identifying seafarers' occupational plasticity brain network based on effective connection lateralization according to claim 8, characterized in that: The seafarer career plasticity brain network is CareerROIs=SailROIs-PsyROIs.

Images