Neuroimaging-based classification of PTSD using data-driven computational approaches: A multisite big data study from the ENIGMA-PGC PTSD consortium.
Year: 2026
Type: article
Abstract: 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.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.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.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.
DOI: 10.17615/5a0s-a564
Authors: Qi, Rongfeng, Hudson, Anna R, May, Geoffery, King, Anthony, Guenette, Jeffrey P, Wang, Li, Lee, Seonjoo, Densmore, Maria, van Zuiden, Mirjam, Neria, Yuval, Gordon, Evan M, Simons, Raluca M, Kim, Yoojean, Bruce, Steven E, Zhu, Xi, Walter, Henrik, Mueller, Sven C, Cao, Zhihong, Maron-Katz, Adi, Nelson, Steven M, Sponheim, Scott R, Zilcha-Mano, Sigal, Sun, Delin, Averill, Christopher L, Larson, Christine L, Rosso, Isabelle M, Forster, Gina L, Davenport, Nicholas D, van der Werff, Steven J A, Nawijn, Laura, Jahanshad, Neda, Xie, Hong, DeBellis, Michael D, Wang, Xin, Angstadt, Michael, Abdallah, Chadi G, Lu, Guang Ming, Ressler, Kerry, Kennis, Mitzy, Peverill, Matthew, Geuze, Elbert, Olff, Miranda, Ipser, Jonathan C, Wall, Melanie, Koopowitz, Sheri M, Stein, Murray B, Thomopoulos, Sophia I, Lazarov, Amit, Grupe, Daniel W, Fercho, Kelene A, Cotton, Andrew S, Olatunji, Bunmi O, Gruber, Staci, Li, Gen, Gonenc, Atilla, Varkevisser, Tim, deRoon-Cassini, Terri, Fani, Negar, Manthey, Antje, Jovanovic, Tanja, O'Leary, Erin M, Shenton, Martha E, Ross, Marisa, Magnotta, Vincent A, van der Wee, Nic J A, Veltman, Dick J, Kaufman, Milissa L, Choi, Kyle, Disner, Seth G, Théberge, Jean, Herringa, Ryan J, Shih, Chiahao, Baugh, Lee A, Haswell, Courtney C, Koerte, Inga K, Suarez-Jimenez, Benjamin, Simons, Jeffrey S, Sheridan, Margaret, Cisler, Josh, Neufeld, Richard W J, Winternitz, Sherry, Vermeiren, Robert R J M, He, Xiaofu, Ravid, Orren, Bryant, Richard A, Lebois, Lauren A M, Sicorello, Maurizio, Sambrook, Kelly, Fitzgerald, Jacklynn, Nitschke, Jack B, Kinzel, Philipp, Jia-Richards, Meilin, Daniels, Judith K, Stevens, Jennifer S, Baird, C Lexi, Hage, Wissam El, Hofmann, David Bernd, Liberzon, Israel, Morey, Rajendra A, Stein, Dan J, Krystal, John H, Koch, Saskia B J, Quidé, Yann, Herzog, Julia, Sierk, Anika, van Erp, Theo G M, Lanius, Ruth, Rektor, Ivan, Straube, Thomas, Frijling, Jessie L, Etkin, Amit, van Rooij, Sanne J H, Davidson, Richard J, Hosseini, Bobak, Zhu, Ye, Weis, Carissa, Luo, Yifeng, Blackford, Jennifer U, Bomyea, Jessica, Simmons, Alan N, Huggins, Ashley A, Salminen, Lauren E, Phan, K Luan, Schmahl, Christian, McLaughlin, Katie, Dennis, Emily L, Thompson, Paul M, Korgaonkar, Mayuresh S, Riha, Pavel
Institution: University of North Carolina at Chapel Hill
Subfield: Clinical Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
Funders: Vanderbilt University, McLean Hospital, National Alliance for Research on Schizophrenia and Depression, Deutsche Forschungsgemeinschaft, National Institutes of Health, Medical Research Council, National Health and Medical Research Council, National Institute of Mental Health, National Institute of Biomedical Imaging and Bioengineering
Awards: G1434541397, G1786539460, Enhancing transdiagnostic mechanisms of cognitive dyscontrol using computer-based training, G205869997, G2112762742, G2140932772, G2402445671, G4421633565, Redefining Large-Scale Functional Network Models of Posttraumatic Stress Disorder through Resolution Limit-Free Community Detection, Neural correlates of PTSD treatment outcome: an fMRI study, Identification of Distinct Multimodal Biotypes of PTSD Using Data Driven Approach: A Multisite Big Data Study, ENIGMA Center for Worldwide Medicine, Imaging & Genomics, Translating Science into Better Posttraumatic Mental Health, G7873196974, A global alliance to unlock brain mechanisms influencing suicidal behaviors through the ENIGMA Consortium, G9000043281, G9000055034, G9000129531, G9000210308, G9000223705, G9000259532, G9000261263, G9000307569, G9000333532
OpenAlex ID: W7140277593
FWCI:
Cites:
Cited by:
Related to: