Supporting depression screening with multimodal emotion detection

Depression is the most common mental disorder in Italian population. The Beck Depression Inventory-II (BDI-II) questionnaire is generally adopted by clinicians as indicator of depression. In this paper we propose LieToMe, a mobile application aiming at collecting together with the BDI-II patient's answers also her images and voice. Emotional analysis is performed on the video and on the images by using deep Convolutional Neural Network. The correlation between the emotional results and the patient's DBI-II scores is provided to the clinician as indication of the relationship between the patient's emotional state and the result of the depression screening. We conduct a preliminary evaluation aiming at assessing the clinician satisfaction on the support offered by LieToMe. Participants were 8 clinicians. Results seem to support the acceptability of LieToMe in the clinical practice.