The use of artificial intelligence to predict depression through thermal imaging

Depression is a mood disorder characterized by a deep feeling of sadness and a sense of indifference. Thus, it become a serious health condition if it happens recurrently within a certain intensity. In that case, early detected or prediction is considered as the most effective method to overcome this crucial disorder. Therefore, this research proposes a Deep Neural Network (DNN) which is one of models of artificial intelligence to predict whether a person as normal, mild, moderate, or severe depression based on Beck's Depression Inventory (BDI) data and by utilizing thermal imaging of the subjects because thermal images are positively possible to mapping the condition of the inner tissues of the human body. The DNN model used is Convolution Neural Network (CNN) with ResNet101 and MobileNetV2 architecture. The test results showed that the proposed model outlines positive performance, under an accuracy of 96.61% for ResNet101 and 96.55% for MobileNetV2. Moreover, MobileNetV2 shown another positive advantage in terms of speed in the training phase. However, there are possible obstacles during the experiment, the external factors, the environmental temperature which might affect the results.