Subjective Cognitive Decline FMRI Data Analysis And Study Based On Statistical Pattern Recognition Method

Subjective Cognitive Decline(SCD)is the earliest state of Alzheimer disease(AD).It has the characteristic of not easy to diagnosis,and easy to develop into AD state.Existing SCD diagnostic methods have remained at the level of purely medical clinical testing and traditional statistics and analysis.Aiming at the above problems,this paper proposes a functional magnetic resonance imaging(fMRI)statistical pattern recognition method based on model fusion and threshold network feature extraction.The main research contents are as follows.A Principal Components Analysis constrained average optimal interface distance(PCAC-AOID)model fusion algorithm based on principal component analysis constraints is proposed.First,PCA is used to reduce the data dimension and obtain linearly independent sample data between column dimensions.Secondly,use these data to construct different classifier models;then,use the distance from the sample points to the optimal interface to model fusion.The method combines the characteristics of sufficient dimension and linear independence after data dimensionality reduction,solves the redundancy problem in feature selection and feature disturbance,and avoids a large amount work of feature selection.A Cascade-based Global Sparse Threshold(CGST)algorithm is proposed.By studying the mean frequency histogram of the sample function connection values,it is found that there are differences in the frequency of functional connection values between SCD and NC samples.In order to study this difference and the impact on the classification of fMRI data,this paper cascades the network characteristics under different parameters based on the global sparse threshold method.The method combines features of multiple global sparse parameters to improve the classification effect..This paper studies the fMRI data classification problem of SCD samples from the perspective of optimization model fusion and network feature extraction.The algorithm of optimal interface distance model fusion method based on PCA constraint and Cascade-based Global Sparse Threshold are proposed respectively.The effectiveness of the proposed algorithm is verified by experiments.