| Healthy elderly people experience a decline in cognitive ability during the process of brain aging.To prevent their cognitive level from continuously declining,which can lead to mild cognitive impairment(MCI)or even Alzheimer’s disease(AD)patients,providing appropriate early intervention for healthy elderly people with poor cognitive level can slow down their cognitive decline,thereby reducing the incidence of MCI and even AD,avoiding creating burdens on society and families.Therefore,early effective assessment of the cognitive level of healthy elderly people is of great significance.Research has shown that during aging,changes in brain structure and function occur before significant cognitive impairment,and building brain networks can characterize the functional connections of the brain.Therefore,this dissertation constructs brain network based on resting-state f MRI data,establishes multivariate pattern analysis(MVPA)method to identify changes in brain functional connection,and explores effective methods for assessing the cognitive level of healthy elderly people.The specific research content is as follows:(1)Construct static brain network based on resting-state f MRI data and establish an MVPA method for assessing the cognitive level of healthy elderly people.Specific MVPA process: use Gaussian Copula mutual information(GCMI),which is sensitive to nonlinear,to build brain network and filter features,and combine support vector machine(SVM)to achieve effective assessment of cognitive level of healthy elderly people.In addition,the existing MVPA methods are used to assess the cognitive level of healthy elderly people for comparative analysis.The MVPA method established in this study achieved a classification accuracy of 77.22%,with sensitivity,specificity,and AUC of 81.82%,72.09%,and 0.77,respectively.However,the classification results of the comparative methods were all inferior to the assessment model established in this study.Further analysis of consistent functional connections using independent-sample -tests suggests that the decrease in cognitive level may be caused by the strength weakening of functional connections with significant differences(P<0.05).Especially the functional connection between ACG.L and STG.L,between INS.L and OLF.L,and between INS.R and SFGmed.L has the potential as a clinical imaging marker.(2)Due to the inability of static brain networks to reflect the dynamic changes in brain functional connections during data scanning.This dissertation is based on resting-state f MRI data,and further uses sliding window technology combined with Pearson correlation coefficient to construct dynamic brain network to assess the cognitive level of healthy elderly people.Specific process: Using Pearson correlation coefficient to construct a dynamic brain network based on non-overlapping sliding window,Kendall correlation coefficient is used to filter features for each window’s brain network,and then SVM-based classifiers are trained on the data after completing feature selection based on each window.Soft voting algorithm is used to integrate all base classifiers,achieving effective assessment of the cognitive level of healthy elderly people.In addition,in the experiment,overlapping window dynamic brain networks and static brain network as control methods were used to assess the cognitive level of healthy elderly people.The results showed that the classification based on dynamic brain networks were generally better than based on static brain networks.The dynamic brain network with nonoverlapping windows achieved the best classification results,with a classification accuracy of77.78%,sensitivity,specificity,and AUC of 90.91%,60.47%,and 0.79,respectively.Further analysis of the dynamic changes in brain functional connection related to cognitive levels revealed that different cognitive levels may be related to the dynamic change patterns within sensory motor network(SMN)modules or between with other functional network modules. |
PDF Full Text Request