Skip to main content

Computer Science Department, PhD Proposal Defense - Ermal Toto " Towards Instantaneous Mental Health Screening "

Friday, November 22, 2019
9:00 am to 10:00 am
Floor/Room #: 
Morgan Conference Room


Committee members:

Prof.  Elke A. Rundensteiner, Advisor,  WPI Computer Science  

Prof.  Carolina Ruiz,  WPI-Computer Science .

Prof.  Lane T. Harrison, Assistant Professor, WPI- Computer Science  

Prof.  Francis (Lee) Stevens, Psychologist, Reliant Medical Group.


The World Health Organization (WHO) has identified mental health disorders as a serious epidemic affecting up to a quarter of the population in some countries. In particular, depression is the leading cause of disability in developed countries, with a tremendous mental and economic burden on patients and family members. The global economic impact of depression is predicted to be approximately US$5.36 trillion between 2011 and 2030.  Although early diagnosis and prevention are crucial to achieving a positive outcome, a fourth of the patients with major depression remain undiagnosed. The explosion of new data sources, such as smartphone logs and social media, offers tremendous opportunities to develop novel instantaneous mental health screening methods. Toward this goal, we propose the following three tasks.

First, we propose the Sub-clip boosting algorithm (SCB), a novel algorithm for emotion detection from easy to obtain short voice clips. 

Second, we propose to investigate the viability of retroactive smart-phone data for mental health screening, and especially depression screening. Retroactive data are generated prior to the user having any knowledge of the study, and in contrast to prospective data, they offer the opportunity for instantaneous mental health screening. In addition, we propose to use retroactive data to generate charts that illustrate the patients’ social interactions. These charts have the potential to be a useful tool for mental health professionals.

Finally, we propose Complex Multi-Gradient Behavior Prediction (COMB), a deep learning method to improve voice classification for the purpose of mental health screening. These steps complete a comprehensive framework that encompasses several aspects of automated mental health screening, from data collection to results delivery.