Machine-learning models in the context of physiological state transitions

Advances in wearable technology allow for the continuous assessment of physiological data to get a detailed picture of daily-life dynamics associated with the development of mental illness. Specific physiological state transitions may hold prognostic value in the course of development, treatment, and relapse of mental disorders. Computational modelling of the physiological data is a promising approach in predicting this course. Physiological state transitions can be observed in different time-windows. On a momentary level, minute-to-minute changes in physiology, such as in the case of acute stress and recovery from acute stress, may predict specific illness-related behavior and symptoms. On the longer term, more structural alterations in physiology such as chronic stress or patterns in circadian rhythm may signal important phases in illness progression, treatment effects, or relapse. I will develop new computational models to detect daily life markers of state transitions related to mental health. Therefore, I will work on existing datasets that have been collected from healthy volunteers and individuals with psychiatric complaints.

Keywords:Machine learning, Computational modeling, Formal theory, Physiological states, Stress

Disciplines:Biological and physiological psychology not elsewhere classified, Mathematical psychology, Artificial intelligence, Psychopathology

Project type:PhD project