Project Detail |
Psychotic disorders rank among the top 25 global causes of disability, impacting individuals, their families, and society. Much of this burden stems from early onset and poor clinical outcomes. While approximately 30% of individuals experience persistent symptoms, illness trajectories following a first psychotic episode vary widely and remain poorly understood. Cumulative genetic and environmental risk exposures and underlying white matter (WM) microstructure abnormalities have been associated with clinical outcomes during the first few years following psychosis onset. Still, it remains unexplored how different combinations of risk exposures are linked via WM microstructure with short- and long-term psychosis outcomes. Thus, RISK-MATTER aims to identify biologically and clinically meaningful subgroups of individuals with recent-onset psychosis based on their different genetic and environmental risk profiles by employing advanced clustering approaches. I will assess whether stratifying participants into these more similar subgroups enhances the effectiveness of supervised machine learning models that utilize WM microstructure and baseline clinical features to predict traditional one-year clinical outcomes. Yet, traditional retrospective assessments to evaluate clinical outcomes are prone to memory and assessor bias. Hence, I will further re-contact a subset of individuals up to eight years after psychosis onset to collect traditional and smartphone-based assessments to determine whether my models’ predictive accuracy is robust when evaluating longer-term and real-life outcomes.
During this project, I will build expertise in diffusion-weighted imaging, unsupervised machine learning, and smartphone-based assessments, mentored by Prof. Pasternak (Harvard Medical School) and Prof. Koutsouleris (Ludwig-Maximilian-University). These skills will bolster my academic profile as a neuroscientific psychologist and facilitate my growth as an independent research group leader. |