Zhu, XiKim, YoojeanRavid, OrrenHe, XiaofuSuarez-Jimenez, BenjaminZilcha-Mano, SigalLazarov, AmitLee, SeonjooAbdallah, Chadi GAngstadt, MichaelAverill, Christopher LBaird, C LexiBaugh, Lee ABlackford, Jennifer UBomyea, JessicaBruce, Steven EBryant, Richard ACao, ZhihongChoi, KyleCisler, JoshCotton, Andrew SDaniels, Judith KDavenport, Nicholas DDavidson, Richard JDeBellis, Michael DDennis, Emily LDensmore, MariadeRoon-Cassini, TerriDisner, Seth GHage, Wissam ElEtkin, AmitFani, NegarFercho, Kelene AFitzgerald, JacklynnForster, Gina LFrijling, Jessie LGeuze, ElbertGonenc, AtillaGordon, Evan MGruber, StaciGrupe, Daniel WGuenette, Jeffrey PHaswell, Courtney CHerringa, Ryan JHerzog, JuliaHofmann, David BerndHosseini, BobakHudson, Anna RHuggins, Ashley AIpser, Jonathan CJahanshad, NedaJia-Richards, MeilinJovanovic, TanjaKaufman, Milissa LKennis, MitzyKing, AnthonyKinzel, PhilippKoch, Saskia BJKoerte, Inga KKoopowitz, Sheri MKorgaonkar, Mayuresh SKrystal, John HLanius, RuthLarson, Christine LLebois, Lauren AMLi, GenLiberzon, IsraelLu, Guang MingLuo, YifengMagnotta, Vincent AManthey, AntjeMaron-Katz, AdiMay, GeofferyMcLaughlin, KatieMueller, Sven CNawijn, LauraNelson, Steven MNeufeld, Richard WJNitschke, Jack BO'Leary, Erin MOlatunji, Bunmi OOlff, MirandaPeverill, MatthewPhan, K LuanQi, RongfengQuidé, YannRektor, IvanRessler, KerryRiha, PavelRoss, MarisaRosso, Isabelle MSalminen, Lauren ESambrook, KellySchmahl, ChristianShenton, Martha ESheridan, MargaretShih, ChiahaoSicorello, MaurizioSierk, AnikaSimmons, Alan NSimons, Raluca MSimons, Jeffrey SSponheim, Scott RStein, Murray BStein, Dan JStevens, Jennifer SStraube, ThomasSun, DelinThéberge, JeanThompson, Paul MThomopoulos, Sophia Ivan der Wee, Nic JAvan der Werff, Steven JAvan Erp, Theo GMvan Rooij, Sanne JHvan Zuiden, MirjamVarkevisser, TimVeltman, Dick JVermeiren, Robert RJMWalter, HenrikWang, LiWang, XinWeis, CarissaWinternitz, SherryXie, HongZhu, YeWall, MelanieNeria, YuvalMorey, Rajendra A2024-07-172024-07-172023-121053-81191095-9572https://hdl.handle.net/10161/31283<h4>Background</h4>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.<h4>Methods</h4>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.<h4>Results</h4>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.<h4>Conclusion</h4>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.https://creativecommons.org/licenses/by-nc/4.0BrainHumansMagnetic Resonance ImagingReproducibility of ResultsStress Disorders, Post-TraumaticNeuroimagingBig DataJournal article