Multi-site benchmark classification of major depressive disorder using machine learning on cortical and subcortical measures

Vladimir Belov, Tracy Erwin-Grabner, Moji Aghajani, Andre Aleman, Alyssa R. Amod, Zeynep Basgoze, Francesco Benedetti, Bianca Besteher, Robin Bülow, Christopher R.K. Ching, Colm G. Connolly, Kathryn Cullen, Christopher G. Davey, Danai Dima, Annemiek Dols, Jennifer W. Evans, Cynthia H.Y. Fu, Ali Saffet Gonul, Ian H. Gotlib, Hans J. GrabeNynke Groenewold, J. Paul Hamilton, Ben J. Harrison, Tiffany C. Ho, Benson Mwangi, Natalia Jaworska, Neda Jahanshad, Bonnie Klimes-Dougan, Sheri Michelle Koopowitz, Thomas Lancaster, Meng Li, David E.J. Linden, Frank P. MacMaster, David M.A. Mehler, Elisa Melloni, Bryon A. Mueller, Amar Ojha, Mardien L. Oudega, Brenda W.J.H. Penninx, Sara Poletti, Edith Pomarol-Clotet, Maria J. Portella, Elena Pozzi, Liesbeth Reneman, Matthew D. Sacchet, Philipp G. Sämann, Anouk Schrantee, Kang Sim, Jair C. Soares, Dan J. Stein, Sophia I. Thomopoulos, Aslihan Uyar-Demir, Nic J.A. van der Wee, Steven J.A. van der Werff, Henry Völzke, Sarah Whittle, Katharina Wittfeld, Margaret J. Wright, Mon Ju Wu, Tony T. Yang, Carlos Zarate, Dick J. Veltman, Lianne Schmaal, Paul M. Thompson, Roberto Goya-Maldonado

Research output: Contribution to journalArticlepeer-review

2 Scopus citations


Machine learning (ML) techniques have gained popularity in the neuroimaging field due to their potential for classifying neuropsychiatric disorders. However, the diagnostic predictive power of the existing algorithms has been limited by small sample sizes, lack of representativeness, data leakage, and/or overfitting. Here, we overcome these limitations with the largest multi-site sample size to date (N = 5365) to provide a generalizable ML classification benchmark of major depressive disorder (MDD) using shallow linear and non-linear models. Leveraging brain measures from standardized ENIGMA analysis pipelines in FreeSurfer, we were able to classify MDD versus healthy controls (HC) with a balanced accuracy of around 62%. But after harmonizing the data, e.g., using ComBat, the balanced accuracy dropped to approximately 52%. Accuracy results close to random chance levels were also observed in stratified groups according to age of onset, antidepressant use, number of episodes and sex. Future studies incorporating higher dimensional brain imaging/phenotype features, and/or using more advanced machine and deep learning methods may yield more encouraging prospects.

Original languageEnglish (US)
Article number1084
JournalScientific reports
Issue number1
StatePublished - Dec 2024

Bibliographical note

Publisher Copyright:
© 2024, The Author(s).


Dive into the research topics of 'Multi-site benchmark classification of major depressive disorder using machine learning on cortical and subcortical measures'. Together they form a unique fingerprint.

Cite this