Depression Recognition Method Based on Regional Homogeneity Features from Emotional Response fMRI Using Deep Convolutional Neural Network

The research of deep convolutions neural network (DNN) in the depression recognition has become a popular topic. In this paper, we propose a method for depression recognition based on the regional homogeneity (ReHo) in emotional task state functional magnetic resonance imaging (task-fMRI) using DNN. First, the task-fMRI is extracted by processing the fMRI of emotional stimulation tasks. And the task-fMRI with ReHo (ReHo-task-fMRI) is calculated based on task-fMRI. And then, convolutional networks of DNN (such as VGG16, etc.) pre-trained on ImageNet are used to automatically complete extracting the classification features from ReHo-taskfMRI. Finally, the Kernel Extreme Learning Machine (KELM) was used to classify the depression. The results of test set showing that for depression recognition, the sensitivity and specificity of ReHo-task-fMRI were 87.46% and 85.35%, however that of task-fMRI were only 67.69% and 55.44%. This study suggest that compared with emotional task-fMRI, ReHo-task-fMRI can better represent the characteristics of brain dysfunction for patients with depression.