Soumyashree Sahoo, assistant professor of computer science at Quinnipiac University, has co-authored a new study exploring the integration of artificial intelligence, wearable technology, and mental health treatment. This interdisciplinary research includes collaboration with teams from the University of Connecticut and UConn Health, aiming to develop more continuous methods for monitoring depression treatment progress.
Read More
Sahoo's expertise lies in machine learning and data modeling, particularly in relation to multimodal and time-series data. Her research focuses on improving health systems and understanding human behavior, emphasizing interpretable and practical AI applications.
The study analyzed data from 49 patients undergoing pharmacological treatment for depression, utilizing daily step counts and sleep patterns collected via Fitbit devices. Conducted over a period of 12 weeks, the research indicated that traditional methods of gauging patient progress, such as questionnaires, often rely on self-reports influenced by recall bias. Sahoo stated that passive sensing through everyday devices provides an alternative by collecting behavioral data with minimal patient effort.
The findings suggested that behavior signals from wearable technology, when assessed with advanced AI methods, can effectively provide insights into treatment progress. Sahoo noted that clinical evaluations were used to determine treatment improvement instead of relying solely on patient self-reports.
To enhance the analysis despite limited data, the research team created a self-supervised contrastive learning approach using a transformer-based model, PatchTST, to identify behavioral patterns from the collected data. Sahoo emphasized the importance of learning from simple, passively gathered signals.
The study reported an F1 score of up to 0.74 when combining step-count data with a baseline depression assessment, which improved to 0.77 when including sleep data. Sahoo highlighted that even basic physical activity or sleep changes can provide valuable insights when analyzed through machine learning.
The research team's paper, titled "Predicting Depression Treatment Outcome Using Daily Step Count Sensory Data," has been accepted for presentation at the 2026 conference by the Institute of Electrical and Electronics Engineers (IEEE) and the Association for Computing Machinery (ACM). Sahoo recently shared the findings at the IEEE/ACM Conference on Connected Health in Pittsburgh, Pennsylvania.
Sahoo emphasized the potential for this research to alleviate the burden on patients by reducing the frequency of questionnaires. Wearable devices enable continuous data collection on physical activity and sleep, paving the way for more personalized mental health care. She concluded that such approaches could allow clinicians to monitor behavioral changes over time and adjust treatments for individual patients more effectively.