Research On FMRI Data Classification Based On Independent Component Analysis And Ensemble Learning

The brain is the central system in which the human body stores,processes,and integrates all kinds of signals,in the study of brain science,functional magnetic resonance imaging is an effective method to study the cognitive process of human brain,it can not only achieve non-invasive and accurate localization,but also has the advantages of high resolution,repeatable research and so on.A large number of data of brain activity can be collected by fMRI technology,it is one of the important means in brain science research to obtain effective information from these data and identify them effectively.This thesis studies fMRI data from the perspectives of feature extraction and classification recognition,and describes the importance of fMRI data features and classification models in brain cognitive research,the common feature extraction methods of fMRI data,the basic process of fMRI data classification and the commonly used classification algorithms are briefly introduced.On the basis of fast independent component analysis,an ensemble learning model based on support vector machine was proposed to analyze fMRI data.The main research contents of this paper are as follows:(1)According to the feature extraction method of fMRI data,the FastICAalgorithm based on the three negative entropy approximation functions of logcosh,exp,cube was used to extract the features,the components extracted under these three functions were used as the input of SVM for classification and recognition respectively.the experimental results show that the FastICA algorithm based on logcosh negative entropy approximation function has certain advantages in feature extraction of fMRI data.(2)In order to solve the high-dimensional problem of fMRI data,an ensemble classification model was constructed by using SVM,and combined with FastICA feature extraction algorithm based on logcosh negative entropy approximation function,a new fMRI data analysis model was obtained.The features extracted by FastICA algorithm were input into the ensemble classification model and SVM for classification and recognition,the experimental results show that the ensemble classification model can better capture the useful information contained in the features and show better classification performance for fMRI data.It enriches the analysis method of fMRI data.