A study by Human Brain Project (HBP) researchers demonstrates how combining empirical data and simulation results can be useful for the classification of Parkinson’s disease patients.

The team of researchers from Forschungszentrum Jülich and Heinrich-Heine University Düsseldorf (Germany) applied machine learning to empirical and simulated whole-brain connectomes – maps of neural connections – to advance the classification of clinical data. The study was published in Brain Communications.

With this work, the researchers introduced a new approach to model validation, which they refer to as “behavioural model fitting,” and found that complementing the empirical data with simulated data can lead to a better differentiation between patients and healthy groups, as well as to a better classification of these patients by machine learning approaches.

These findings contribute to a better understanding of empirical and simulated whole-brain dynamics and their relationship to disease. The results also suggest an avenue for applying brain simulations to investigate disease diagnosis and inter-individual differences in the brain.

Text by Helen Mendes

Reference: Kyesam Jung, Esther Florin, Kaustubh R Patil, Julian Caspers, Christian Rubbert, Simon B Eickhoff, Oleksandr V Popovych, Whole-brain dynamical modelling for classification of Parkinson’s disease, Brain Communications, Volume 5, Issue 1, 2023, fcac331, https://doi.org/10.1093/braincomms/fcac331