The Study Of Functional Connectivity Analysis Methods Based On Resting State FMRI

With the accelerated pace of modern life, all aspects of the pressure increasing, more and more people are suffering from mental illness of depression. According to the survey of depression prevalence,it is very high and its suicide rate is located in all mental illness, which causes the suffering to the patients and their families and the loss to the society that other mental illness can not be compare. At present, the identify rate of this illness diagnosis method is low due to clinical experience rather than the quantitative physiological indicator, which delays the optimal duration of treatment and leads to serious consequences. In recent years, with the development of the medical imaging technology especially the resting state fMRI technology, it provides the material basis to investigate the brain function of depression and it is possible to develop the safe and efficient treatment for depression. In view of the urgent need of looking for the biological indicators for the diagnosis and treatment of depression, as well as the growing maturity of fMRI data analysis and pattern classification techniques, this paper studies the functional connectivity analysis and classification methods based on resting state fMRI between depression and health control. The main work of this paper is as follows.Firstly, according to the purpose of the experiment this paper collectes the resting state fMRI data of subjects by careful selection and compleats the data prepreocessing thus to get efficient and reliable experiment data.Secondly, In the data analysis methods, this paper focuses on the various model-driven methods of resting state fMRI functional connectivity analysis. It not only includes cross-correlation analysis and partial cross-correlation analysis in the time domain but also coherence analysis and mutual information analysis in the frequency domain.Thirdly, in the aspects of feature selection in this paper, it selects the parameters T test method and non-parameters K-S test method.Finally, Classification is SVM and three parameter optimization of the SVM algorithm to classify.The classification accuracy of this experiment is75%, far more than the random level of50%. This demonstrates that the selection methods in the experiment are feasible and can distinguish the depression and the healthy. It places a certain role to establish the physiological indicator for the diagnosis and treatment of depression, thus it can better aid the clinical diagnosis and treatment.