The Classification Research Of Resting-state FMRI Data In Major Depressive Disorder Based On ReHo And ALFF

With the accelerated pace of urban development, people suffered a great number of pressure that comes anywhere. So, more and more people suffering from depression which is extremely harmful to society. The current defect is lack of quantitative diagnosis of physiological indicators. Doctors mostly dependent on the information from patients, families and clinicians inquire symptoms. This method mostly depend on subjective approach, so it always delay the time of the best treatment. With the continuous development of medical imaging technology in recent years, especially in resting state functional magnetic resonance imaging technology matures, it provides the necessary theoretical basis for future research. Using resting state functional magnetic resonance imaging technology, advanced data analysis and cognitive modeling techniques, we make a multi-angle analysis of magnetic resonance imaging data in patients with depression that include regional homogeneity, amplitude of low frequency fluctuation and fractional amplitude of low frequency fluctuation. It is realized this three physiological indicators in brain imaging research and application of machine learning. The main work of this paper is as follows:Firstly, according to the experimental design, we collected experimental subjects in resting state through carefully and usefully method, and carry a series of data pretreatment process thus we get credible experimental data.Secondly, according to the definition of the adjacent voxels, the obtained data are processed separately it is include regional homogeneity, amplitude of low frequency fluctuation and fractional amplitude of low frequency fluctuation.Thirdly, through the use of model-driven analysis methods and theoretical knowledge, we apply the experimental data to the analysis. By using machine learning methods (SVM-RBF) and building classification model, this paper mainly explore the correct rate of the classification and the optimal number of features for the regional homogeneity, amplitude of low frequency fluctuation and fractional amplitude of low frequency fluctuation. And also using feature importance to analyze the selected features degree of contribution in the classification process and tune in the feature selection process, optimizing the classification model.Fourthly, in this paper the vast majority of the correct classification rate are more than75%, some even up to90%which fully explained in the text of the experimental method is reliable and effective. It can be effective distinguish in patients with depression and healthy controls. It indicates that depression through the use of resting state functional magnetic resonance imaging as a classification index played a relevant role, which also for doctors to provide a better supplementary clinical diagnosis of depression and psychological treatment methods.Finally, the current study is the main component of the National Natural Science Foundation Project "The study of fMRI data analysis methods and diagnosis treatment models in major depressive disorder (No.61170136)", and is also supported by the University Science Research and Development Project of Shanxi Province (No.20121003) and the Special/Youth Foundation of Taiyuan University of Technology (No.2012L014).