10 July 2022>: Clinical Research
Automatic Identification of Depression Using Facial Images with Deep Convolutional Neural Network
Xinru Kong 1CE* , Yan Yao 1BC* , Cuiying Wang 1BC , Yuangeng Wang 1BF , Jing Teng 2DG , Xianghua Qi 2AG*DOI: 10.12659/MSM.936409
Med Sci Monit 2022; 28:e936409
Figure 1 The collected datasets of patients with depression and healthy participants were divided into a training set, test set, and validation set at the ratio of 7: 2: 1. The study used the convolutional neural network to construct a complete connection layer, which was added to the current advanced attention mechanism model and named FCN. The FCN model had 11 layers (1) convolutional layer (7,7,64); (2) convolutional layer (3,3,64)×2; (3) convolutional layer (3,3,64)×2; (4) convolutional layer (3,3,128)×2; (5) convolutional layer (3,3,128)×2; (6) convolutional layer (3,3,256)×2; (7) convolutional layer (3,3,256)×2; (8) convolutional block attention module; (9) convolutional layer (3,3,512)×2; (10) convolutional layer (3,3,512)×2; and (11) fully connected Layer (512,2). (Because the validation set is also used to test the training model, the validation set was organized into the validation set and was not reflected in the flowchart separately.)