Poster Presentation DGPPN 2019


Introduction. Deviation between chronological age and brain-based age prediction has frequently been used as an indicator of disease risk and severity in psychiatric and neurological disorders. Here, in a sample of patients with major depressive disorder, we aim to understand to what extent the severity of depressive symptoms is associated with a deviation from the trajectory of normal brain aging. We employ an approach to normative brain age modelling which uses machine learning models to predict chronological age from structural neuroimaging data in healthy individuals. Using these normative models on a sample of depressive patients, we infer the degree of acceleration in brain aging and associate this deviation with the severity of depressive symptoms. Methods. We trained regional whole-brain machine-learning models on N = 1,753 healthy adults (17-66 years) to predict chronological age based on structural neuroimaging data using the Harvard Oxford atlas. We then computed individual brain predicted age differences (brain-PAD) for every ROI in an independent sample of depressive patients (N = 513, 18-65 years). Beck’s Depression Inventory (BDI) was used to measure symptom severity in these subjects (3). The region-specific impact of BDI on brain-PAD was estimated using a multi-level Bayesian modelling approach. Results. The brain age model predicted chronological age in the healthy reference sample with high accuracy (cross-validated mean absolute error = 4.07 years; Pearson’s r = .93). The Bayesian multi-level model revealed a substantial positive relationship between BDI scores and brain-PAD (slope b = 0.40, CI(95%) = 0.02, 0.78) in 31 of the 69 Harvard Oxford regions. Conclusions. These findings suggest a substantial association between depressive symptom severity and accelerated brain aging in multiple regions of the brain. Future studies might address the question whether accelerated brain aging is stable across time.

Nov 29, 2019 1:30 PM — 3:00 PM
Messedamm 26, Berlin, 14055
Nils R Winter
PhD Student in Machine Learning and Neuroimaging

My research interests include machine learning, statistical modelling, neuroimaging and research on mental disorders.