Functional connectivity signatures of major depressive disorder: machine learning analysis of two multicenter neuroimaging studies

This study used machine learning algorithms to differentiate patients with major depressive disorder (MDD) from healthy controls using resting-state functional magnetic resonance imaging (fMRI) data from two large consortia, REST-meta-MDD (N=2338) and PsyMRI (N=1039). Support vector machines (SVM) and graph convolutional neural networks (GCN) achieved a mean classification accuracy of 61%. Sex classification accuracy was significantly higher (73-81%). Visualization revealed that classifications were driven by stronger thalamic connections, suggesting thalamic hyperconnectivity as a potential neurophysiological signature of depression.

Selene Gallo, Ahmed El-Gazzar, Paul Zhutovsky, Rajat M. Thomas, Nooshin Javaheripour, Meng Li, Lucie Bartova, Deepti Bathula, Udo Dannlowski, Christopher Davey, Thomas Frodl, Ian Gotlib, Simone Grimm, Dominik Grotegerd, Tim Hahn, Paul J. Hamilton, Ben J. Harrison, Andreas Jansen, Igor Nenadić, Sebastian Olbrich, Elisabeth Paul, Lukas Pezawas, Matthew D. Sacchet, Philipp Sämann, Gerd Wagner, Henrik Walter, Martin Walter, Guido van Wingen