NeuroImage (Dec 2023)

Neuroimaging-based classification of PTSD using data-driven computational approaches: A multisite big data study from the ENIGMA-PGC PTSD consortium

  • Xi Zhu,
  • Yoojean Kim,
  • Orren Ravid,
  • Xiaofu He,
  • Benjamin Suarez-Jimenez,
  • Sigal Zilcha-Mano,
  • Amit Lazarov,
  • Seonjoo Lee,
  • Chadi G. Abdallah,
  • Michael Angstadt,
  • Christopher L. Averill,
  • C. Lexi Baird,
  • Lee A. Baugh,
  • Jennifer U. Blackford,
  • Jessica Bomyea,
  • Steven E. Bruce,
  • Richard A. Bryant,
  • Zhihong Cao,
  • Kyle Choi,
  • Josh Cisler,
  • Andrew S. Cotton,
  • Judith K. Daniels,
  • Nicholas D. Davenport,
  • Richard J. Davidson,
  • Michael D. DeBellis,
  • Emily L. Dennis,
  • Maria Densmore,
  • Terri deRoon-Cassini,
  • Seth G. Disner,
  • Wissam El Hage,
  • Amit Etkin,
  • Negar Fani,
  • Kelene A. Fercho,
  • Jacklynn Fitzgerald,
  • Gina L. Forster,
  • Jessie L. Frijling,
  • Elbert Geuze,
  • Atilla Gonenc,
  • Evan M. Gordon,
  • Staci Gruber,
  • Daniel W Grupe,
  • Jeffrey P. Guenette,
  • Courtney C. Haswell,
  • Ryan J. Herringa,
  • Julia Herzog,
  • David Bernd Hofmann,
  • Bobak Hosseini,
  • Anna R. Hudson,
  • Ashley A. Huggins,
  • Jonathan C. Ipser,
  • Neda Jahanshad,
  • Meilin Jia-Richards,
  • Tanja Jovanovic,
  • Milissa L. Kaufman,
  • Mitzy Kennis,
  • Anthony King,
  • Philipp Kinzel,
  • Saskia B.J. Koch,
  • Inga K. Koerte,
  • Sheri M. Koopowitz,
  • Mayuresh S. Korgaonkar,
  • John H. Krystal,
  • Ruth Lanius,
  • Christine L. Larson,
  • Lauren A.M. Lebois,
  • Gen Li,
  • Israel Liberzon,
  • Guang Ming Lu,
  • Yifeng Luo,
  • Vincent A. Magnotta,
  • Antje Manthey,
  • Adi Maron-Katz,
  • Geoffery May,
  • Katie McLaughlin,
  • Sven C. Mueller,
  • Laura Nawijn,
  • Steven M. Nelson,
  • Richard W.J. Neufeld,
  • Jack B Nitschke,
  • Erin M. O'Leary,
  • Bunmi O. Olatunji,
  • Miranda Olff,
  • Matthew Peverill,
  • K. Luan Phan,
  • Rongfeng Qi,
  • Yann Quidé,
  • Ivan Rektor,
  • Kerry Ressler,
  • Pavel Riha,
  • Marisa Ross,
  • Isabelle M. Rosso,
  • Lauren E. Salminen,
  • Kelly Sambrook,
  • Christian Schmahl,
  • Martha E. Shenton,
  • Margaret Sheridan,
  • Chiahao Shih,
  • Maurizio Sicorello,
  • Anika Sierk,
  • Alan N. Simmons,
  • Raluca M. Simons,
  • Jeffrey S. Simons,
  • Scott R. Sponheim,
  • Murray B. Stein,
  • Dan J. Stein,
  • Jennifer S. Stevens,
  • Thomas Straube,
  • Delin Sun,
  • Jean Théberge,
  • Paul M. Thompson,
  • Sophia I. Thomopoulos,
  • Nic J.A. van der Wee,
  • Steven J.A. van der Werff,
  • Theo G.M. van Erp,
  • Sanne J.H. van Rooij,
  • Mirjam van Zuiden,
  • Tim Varkevisser,
  • Dick J. Veltman,
  • Robert R.J.M. Vermeiren,
  • Henrik Walter,
  • Li Wang,
  • Xin Wang,
  • Carissa Weis,
  • Sherry Winternitz,
  • Hong Xie,
  • Ye Zhu,
  • Melanie Wall,
  • Yuval Neria,
  • Rajendra A. Morey

Journal volume & issue
Vol. 283
p. 120412

Abstract

Read online

Background: Recent advances in data-driven computational approaches have been helpful in devising tools to objectively diagnose psychiatric disorders. However, current machine learning studies limited to small homogeneous samples, different methodologies, and different imaging collection protocols, limit the ability to directly compare and generalize their results. Here we aimed to classify individuals with PTSD versus controls and assess the generalizability using a large heterogeneous brain datasets from the ENIGMA-PGC PTSD Working group. Methods: We analyzed brain MRI data from 3,477 structural-MRI; 2,495 resting state-fMRI; and 1,952 diffusion-MRI. First, we identified the brain features that best distinguish individuals with PTSD from controls using traditional machine learning methods. Second, we assessed the utility of the denoising variational autoencoder (DVAE) and evaluated its classification performance. Third, we assessed the generalizability and reproducibility of both models using leave-one-site-out cross-validation procedure for each modality. Results: We found lower performance in classifying PTSD vs. controls with data from over 20 sites (60 % test AUC for s-MRI, 59 % for rs-fMRI and 56 % for d-MRI), as compared to other studies run on single-site data. The performance increased when classifying PTSD from HC without trauma history in each modality (75 % AUC). The classification performance remained intact when applying the DVAE framework, which reduced the number of features. Finally, we found that the DVAE framework achieved better generalization to unseen datasets compared with the traditional machine learning frameworks, albeit performance was slightly above chance. Conclusion: These results have the potential to provide a baseline classification performance for PTSD when using large scale neuroimaging datasets. Our findings show that the control group used can heavily affect classification performance. The DVAE framework provided better generalizability for the multi-site data. This may be more significant in clinical practice since the neuroimaging-based diagnostic DVAE classification models are much less site-specific, rendering them more generalizable.

Keywords