Multi-site benchmark classification of major depressive disorder using machine learning on cortical and subcortical measures
-
Published:2024-01-11
Issue:1
Volume:14
Page:
-
ISSN:2045-2322
-
Container-title:Scientific Reports
-
language:en
-
Short-container-title:Sci Rep
Author:
Belov Vladimir, Erwin-Grabner Tracy, Aghajani Moji, Aleman Andre, Amod Alyssa R., Basgoze Zeynep, Benedetti FrancescoORCID, Besteher Bianca, Bülow Robin, Ching Christopher R. K., Connolly Colm G., Cullen KathrynORCID, Davey Christopher G., Dima Danai, Dols Annemiek, Evans Jennifer W., Fu Cynthia H. Y., Gonul Ali Saffet, Gotlib Ian H.ORCID, Grabe Hans J., Groenewold Nynke, Hamilton J PaulORCID, Harrison Ben J., Ho Tiffany C., Mwangi Benson, Jaworska Natalia, Jahanshad Neda, Klimes-Dougan Bonnie, Koopowitz Sheri-Michelle, Lancaster ThomasORCID, Li Meng, Linden David E. J.ORCID, MacMaster Frank P., Mehler David M. A., Melloni Elisa, Mueller Bryon A.ORCID, Ojha Amar, Oudega Mardien L., Penninx Brenda W. J. H., Poletti SaraORCID, Pomarol-Clotet Edith, Portella Maria J., Pozzi Elena, Reneman Liesbeth, Sacchet Matthew D., Sämann Philipp G., Schrantee Anouk, Sim KangORCID, Soares Jair C., Stein Dan J.ORCID, Thomopoulos Sophia I., Uyar-Demir Aslihan, van der Wee Nic J. A., van der Werff Steven J. A., Völzke Henry, Whittle Sarah, Wittfeld KatharinaORCID, Wright Margaret J., Wu Mon-Ju, Yang Tony T., Zarate Carlos, Veltman Dick J., Schmaal LianneORCID, Thompson Paul M., Goya-Maldonado Roberto,
Abstract
AbstractMachine 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.
Publisher
Springer Science and Business Media LLC
Subject
Multidisciplinary
Reference81 articles.
1. Kessler, R. C. & Bromet, E. J. The epidemiology of depression across cultures. Annu. Rev. Public Health 34, 119–138 (2013). 2. Cho, Y. et al. Factors associated with quality of life in patients with depression: A nationwide population-based study. PLOS ONE 14, e0219455 (2019). 3. Cai, H. et al. Prevalence of suicidality in major depressive disorder: A systematic review and meta-analysis of comparative studies. Front. Psychiatry 12, (2021). 4. Cleare, A. F. S. W. C. D. K. M. B. M. L. P. A. J. A multidimensional tool to quantify treatment resistance in depression: The Maudsley staging method. J. Clin. Psychiatry 70, 12363 (2009). 5. Han, L. K. M. et al. Brain aging in major depressive disorder: Results from the ENIGMA major depressive disorder working group. Mol. Psychiatry https://doi.org/10.1038/s41380-020-0754-0 (2020).
|
|