Research On FMRI Data Classification Method Based On Statistical Features

Functional magnetic resonance imaging(fMRI)is a technique that can detect changes in brain BOLD signals and has been widely used in medical diagnosis.Based on the analysis of the statistical characteristics of fMRI data,the influence of brain signals on various diseases can be explored.This paper studies the classification method of fMRI data based on statistical feature analysis.Specifically,it involves statistical analysis of fMRI data,feature selection based on statistical analysis,and classification methods based on statistical features.The specific research content includes the following three aspects:First of all,this article proposes to use the regional homogeneity analysis(ReHo)method and the voxel-mirrored homotopic connectivity(VMHC)method to analyze brain abnormalities in patients with chronic urticaria.Using the local consistency analysis method,we show that there are significant differences in brain regions between patients with chronic urticaria and healthy people.Further analysis show that many abnormal brain regions are located in symmetrical parts of the brain,and the voxel homotopy connection analysis is carried out to directly study the symmetrical voxels of the brain.Through the two-sample t-test,the brain areas that are significantly different between chronic urticaria patients and healthy people under the two indicators of Re Ho and VMHC are obtained.Secondly,in view of the limitation that the existing method of using the BOLD signal of the voxel in the brain area as the classification feature will cause the classification speed to be low,a method of feature selection based on the results of statistical analysis is proposed.The brain areas with significant differences obtained from the analysis results are extracted,and each different brain area is made into a mask,and the ReHo value and VMHC value of the different brain areas are extracted based on these masks,which are used as classification features.Finally,according to the limitation of the Naive Bayes algorithm that requires that each feature attribute must be independent of each other,the classification error is large,and a weighted naive Bayes classification algorithm based on Cramer's V coefficient is proposed.By weighting the attributes,the degree of influence of each attribute on the decision is calculated,thereby improving the original algorithm.Finally,a comparative experiment was carried out for the three classification algorithms.The results show that the classification accuracy of the classification algorithm proposed in this paper is higher,and the classification error of the naive Bayes algorithm is largely reduced.