TUBITAK Destekli Projelerden Çıkan Yayınlar Topluluğu
Published January 1, 2024 | Version v1
Journal article Open

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

Creators

  • 1. Georg August Univ, Univ Med Ctr Gottingen UMG, Dept Psychiat & Psychotherapy, Lab Syst Neurosci & Imaging Psychiat SNIP Lab, Von Siebold Str 5, D-37075 Gottingen, Germany
  • 2. Univ Groningen, Univ Med Ctr Groningen, Dept Biomed Sci Cells & Syst, Groningen, Netherlands
  • 3. Univ Cape Town, Dept Psychiat & Mental Hlth, Cape Town, South Africa
  • 4. Univ Minnesota, Med Sch, Dept Psychiat & Behav Sci, Minneapolis, MN 55417 USA
  • 5. IRCCS San Raffaele Sci Inst, Div Neurosci, Milan, Italy
  • 6. Jena Univ Hosp, Dept Psychiat & Psychotherapy, Jena, Germany
  • 7. Univ Med Greifswald, Inst Radiol & Neuroradiol, Greifswald, Germany
  • 8. Univ Southern Calif, Mark & Mary Stevens Neuroimaging & Informat Inst, Keck Sch Med, Imaging Genet Ctr, Marina Del Rey, CA USA
  • 9. Florida State Univ, Dept Biomed Sci, Tallahassee, FL 32306 USA
  • 10. Univ Melbourne, Dept Psychiat, Melbourne Neuropsychiat Ctr, Parkville, Vic, Australia
  • 11. Vrije Univ Amsterdam, Amsterdam UMC, Amsterdam Neurosci, Dept Psychiat,Amsterdam Publ Hlth Res Inst, Amsterdam, Netherlands
  • 12. NIMH, Expt Therapeut & Pathophysiol Branch, NIH, Bethesda, MD 20892 USA
  • 13. Ege Univ, Sch Med, Dept Psychiat, SoCAT Lab, Izmir, Turkiye
  • 14. Stanford Univ, Dept Psychol, Stanford, CA 94305 USA
  • 15. Univ Med Greifswald, Dept Psychiat & Psychotherapy, Greifswald, Germany

Description

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.

Files

bib-f15a6436-9466-49b4-bccd-b8439d0e7e0b.txt

Files (413 Bytes)

Name Size Download all
md5:588df6075b92fa5b40948fb79072fc02
413 Bytes Preview Download