Research On Recognition Of Depression Patients Under Musical And Nonmusical Stimulus Based On MVPA

Depression is a socially influential disease in recent years,and the development of functional magnetic resonance imaging has provided technical support for the brain research of patients with depression.Based on functional magnetic resonance imaging data,this paper aims to use machine learning methods to classify patients with major depressive disorder under auditory stimulation,and to map the brain function of characteristic voxels to provide a reference brain area for the study of the diagnosis,prognosis and the pathogenesis of depression.Based on the musical functional magnetic resonance imaging data,first of all,the image preprocessing of f MRI is completed with SPM,including three steps : slicing time correction,motion correction and spatial normalization.Then,the multivariate pattern analysis is completed with Python,which includes preliminary feature selection through univariate analysis of variance,feature selection through Lasso method,linear support vector classification based on Jackknife method,and verification of the significance of prediction results through permutation tests.It comes out that 83 features are chosen from the 170,000 and the prediction result of the classifier in test set is very good with the accuracy rate being 100%.Finally,Brodmann Areas are used to map the voxels obtained by feature selection to the brain function,and 10 possible depression-related Brain areas are obtained.In addition,the analysis process under musical stimulus is compared with that under nonmusical stimulus,which suggests that the classification results and brain function localization results under nonmusical stimulus are consistent with the results under musical stimulus.Combining the results under music stimulus and nonmusical stimulus,this paper obtains a total of 11 Brodmann Areas that may be related to depression,which are: BA 6,BA 8,BA 9,BA 10,BA 13,BA 24,BA 32,BA 44,BA 45,BA 46,BA 47.