Multi-Task Learning to Screen for Major Depressive Disorder and Post-Traumatic Stress Disorder

Shrestha, Avantika

Etd

Multi-Task Learning to Screen for Major Depressive Disorder and Post-Traumatic Stress Disorder

Public Deposited

Prevalent mental health conditions, major depressive disorder (MDD), and posttraumatic stress disorder (PTSD) have severe physical and social impacts. Detection is difficult and expensive, requiring substantial time from trained mental professionals. To counteract this issue, recent research has explored the diagnostic potential of deep learning models trained on modalities extracted from virtual agent-conducted clinical interview videos. However, deep learning models are challenging to train due to long sequences and the small number of participants that are common in the mental health community. In this thesis, a solution to combat these challenges is developed by leveraging a multi-task learning framework that uses temporal facial features as input to screen for MDD and PTSD. The multi-task framework is based on a bidirectional GRU model with self-attention. This thesis evaluates the multi-task model on temporal facial features extracted from the responses to 15 clinical interview questions conducted by a virtual agent. The results suggest that multi-task learning increases the generalization performance compared to single-task learning. For MDD screening, multi-task learning improved the balanced accuracy over single-task learning for 11 of the 15 datasets. The multi-task learning model increased the MDD screening ability by 25 percent to a balanced accuracy of 0.87 in some scenarios. This work provides valuable findings for the future of mental screening applications leveraging temporal facial features.

Creator