Discriminant Analysis In Depression Based On Resting State FMRI Data

Depression, also called depressive disorders, is mainly characterized by the significant and lasting low mood. The nosogenesis of depression remains unclear so far. There has been a lot of research on using the new technology with the potential of describing and distinguishing the brain diseases to objectively forecast the disease of depression in order to play a supplementary role in clinical diagnosis and reduce the impact of subjective factors since fMRI rose. Therefore, having collecting the data of resting-state fMRI of 39 patients with depression and 37 normal people in the control group, the paper hopes to use the data of function connectivity extracted as the feature to discriminate and analyze the patients and the normal people so as to find out the abnormal cerebral areas and functional network of the patients with depression.Firstly, the paper uses the non-parametric Mann-Whitney U test to reduce the dimension of 4005-dimension feature of 90 brain regions, and then utilizes these significantly different features between the depressed and the normal to discriminate and analyze, with the results: the highest classification of generalization based on decision tree models gets 65.79% correct while the one based on the Elman neural network model gets 76.62% correct; and the highest classification of generalization based on the AdaBoost model elevated by the decision trees gets 83.48% correct, while the one based on the AdaBoost model enhanced by Elman neural network gets 81.43% correct. The paper also males use of all the 4005-dimension features as an input, and the highest generalization using the Lasso-Logistic regression model gets 88.16% correct.Using the Lasso-Logistic regression model with the highest accuracy in classification of generalization, the paper reconstitutes all the participant regions of the brain and their functional connectivity, and finds out that the abnormal brain and functional connectivity of the patients with depression. The main brain regions contain the left temporal gyrus, left fusiform gyrus, right orbital frontal cortex and left globus pallidus lentiform nucleus, and the three main circuits of the abnormal functions are the circuit of “inferior temporal gyrus-middle temporal gyrus-middle frontal gyrus” and the limbic system of “temporal pole-orbital frontal cortex-insula- precentral gyrus” and the emotion regulation networks of “insula-superior frontal gyrus-shell lentiform nucleus-amygdaloid nucleus”.