Assisted diagnosis of depression based on hybrid bilinear model

Depression, as a common chronic mental disorder, has complex causes and low recovery rates. A method for assisting the diagnosis of depression using a hybrid bilinear deep learning network based on scalp electroencephalography is proposed. Firstly, the spatial features extracted by the convolutional neural network and the spatiotemporal features extracted by the convolutional long short-term memory network are fused into second-order hybrid features through bilinear methods to construct a hybrid bilinear model. Then, the functional connectivity matrices of each frequency band of EEG signals are used to train the model, and different functional connectivity measurement methods are used to analyze the relationship between the functional connectivity of each frequency band of EEG signals and depression. Finally, this method is applied on the MODMA dataset. The experimental results showed that the hybrid bilinear model using second-order hybrid features achieved an accuracy of 99.38% on the Beta frequency band correlation functional connectivity matrix, which indicates the effectiveness of the second-order hybrid features of the Beta frequency band correlation functional connectivity matrix in the auxiliary diagnosis of depression. Compared with other methods, the proposed method achieves higher accuracy and has high application prospects.