Machine Learning in Clinical Brain Research

Prof. Dr. Kerstin Ritter, Junior Professor for Computational Neuroscience, Charité Berlin

In this presentation, I will delve into the potential of machine learning techniques for analyzing brain imaging data to assess individual brain health. I will cover the spectrum from classical to advanced methods, like deep (representational) learning, highlighting their use in diagnosing neurological and psychiatric conditions, and forecasting disease progression and treatment outcomes. Besides discussing methodological challenges, including dealing with small sample sizes and noisy labels, ensuring model explainability, correcting confounding variables, and conducting causal inference, I will show these techniques' applicability in various clinical settings including the diagnosis of Alzheimer's disease, assessing brain age as a marker of brain health, predicting the disease course in multiple sclerosis, predicting binge drinking in adolescents and forecasting treatment outcome in internalizing disorders such as depression and anxiety. I will underscore the importance of integrating data from different modalities and the need for a more comprehensive characterization of individuals on different phenotypical levels.