Post doc Tobias Kaufmann
Utilizing machine learning to discover functional brain connectivity abnormalities in schizophrenia
Tobias Kaufmann, PhD in Psychology
Functional dysconnectivity in schizophrenia
When two brain regions show the same activity patterns over time, we consider them as functionally linked. We study such functional connectivity using temporal signal variations of functional magnetic resonance imaging data.
Utilizing advanced network modeling, we build the functional brain networks on data from patients with schizophrenia and healthy controls. Next, we assess the differences between these networks using uni - and multivariate statistics.
Application of machine learning
The phenomenology of schizophrenia may be as diverse that two patients may not share one common symptom. In a similar vein, the variation in functional network connectivity between schizophrenic patients is high, thus hampering a clear picture of the underlying network pathophysiology.
Machine learning can help us identifying reliable patterns in order to disentangle origins of the disease from multiple subject-specific network alterations. In short, we train a machine learning classifier on a set of patients and healthy controls to identify patterns across their functional connectivity networks. Next, we test the classifier on single subject level to assess sensitivity of our methods.
From these results we can learn where in the brain we find reliable effects of schizophrenia on functional brain network connectivity thereby gaining novel insights into the disease pathophysiology.