Abstract
Original language | English |
---|---|
Article number | 120412 |
Journal | NEUROIMAGE |
Volume | 283 |
DOIs | |
Publication status | Published - 1 Dec 2023 |
Keywords
- Classification
- Deep learning
- Machine learning
- Multimodal MRI
- Posttraumatic stress disorder
Access to Document
Other files and links
Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver
}
In: NEUROIMAGE, Vol. 283, 120412, 01.12.2023.
Research output: Contribution to journal › Article › Academic › peer-review
TY - JOUR
T1 - Neuroimaging-based classification of PTSD using data-driven computational approaches
T2 - A multisite big data study from the ENIGMA-PGC PTSD consortium
AU - Zhu, Xi
AU - Kim, Yoojean
AU - Ravid, Orren
AU - He, Xiaofu
AU - Suarez-Jimenez, Benjamin
AU - Zilcha-Mano, Sigal
AU - Lazarov, Amit
AU - Lee, Seonjoo
AU - Abdallah, Chadi G.
AU - Angstadt, Michael
AU - Averill, Christopher L.
AU - Baird, C. Lexi
AU - Baugh, Lee A.
AU - Blackford, Jennifer U.
AU - Bomyea, Jessica
AU - Bruce, Steven E.
AU - Bryant, Richard A.
AU - Cao, Zhihong
AU - Choi, Kyle
AU - Cisler, Josh
AU - Cotton, Andrew S.
AU - Daniels, Judith K.
AU - Davenport, Nicholas D.
AU - Davidson, Richard J.
AU - DeBellis, Michael D.
AU - Dennis, Emily L.
AU - Densmore, Maria
AU - deRoon-Cassini, Terri
AU - Disner, Seth G.
AU - Hage, Wissam El
AU - Etkin, Amit
AU - Fani, Negar
AU - Fercho, Kelene A.
AU - Fitzgerald, Jacklynn
AU - Forster, Gina L.
AU - Frijling, Jessie L.
AU - Geuze, Elbert
AU - Gonenc, Atilla
AU - Gordon, Evan M.
AU - Gruber, Staci
AU - Grupe, Daniel W.
AU - Guenette, Jeffrey P.
AU - Haswell, Courtney C.
AU - Herringa, Ryan J.
AU - Herzog, Julia
AU - Hofmann, David Bernd
AU - Hosseini, Bobak
AU - Hudson, Anna R.
AU - Huggins, Ashley A.
AU - Ipser, Jonathan C.
AU - Jahanshad, Neda
AU - Jia-Richards, Meilin
AU - Jovanovic, Tanja
AU - Kaufman, Milissa L.
AU - Kennis, Mitzy
AU - King, Anthony
AU - Kinzel, Philipp
AU - Koch, Saskia B. J.
AU - Koerte, Inga K.
AU - Koopowitz, Sheri M.
AU - Korgaonkar, Mayuresh S.
AU - Krystal, John H.
AU - Lanius, Ruth
AU - Larson, Christine L.
AU - Lebois, Lauren A. M.
AU - Li, Gen
AU - Liberzon, Israel
AU - Lu, Guang Ming
AU - Luo, Yifeng
AU - Magnotta, Vincent A.
AU - Manthey, Antje
AU - Maron-Katz, Adi
AU - May, Geoffery
AU - McLaughlin, Katie
AU - Mueller, Sven C.
AU - Nawijn, Laura
AU - Nelson, Steven M.
AU - Neufeld, Richard W. J.
AU - Nitschke, Jack B.
AU - O'Leary, Erin M.
AU - Olatunji, Bunmi O.
AU - Olff, Miranda
AU - Peverill, Matthew
AU - Phan, K. Luan
AU - Qi, Rongfeng
AU - Quidé, Yann
AU - Rektor, Ivan
AU - Ressler, Kerry
AU - Riha, Pavel
AU - Ross, Marisa
AU - Rosso, Isabelle M.
AU - Salminen, Lauren E.
AU - Sambrook, Kelly
AU - Schmahl, Christian
AU - Shenton, Martha E.
AU - Sheridan, Margaret
AU - Shih, Chiahao
AU - Sicorello, Maurizio
AU - Sierk, Anika
AU - Simmons, Alan N.
AU - Simons, Raluca M.
AU - Simons, Jeffrey S.
AU - Sponheim, Scott R.
AU - Stein, Murray B.
AU - Stein, Dan J.
AU - Stevens, Jennifer S.
AU - Straube, Thomas
AU - Sun, Delin
AU - Théberge, Jean
AU - Thompson, Paul M.
AU - Thomopoulos, Sophia I.
AU - van der Wee, Nic J. A.
AU - van der Werff, Steven J. A.
AU - van Erp, Theo G. M.
AU - van Rooij, Sanne J. H.
AU - van Zuiden, Mirjam
AU - Varkevisser, Tim
AU - Veltman, Dick J.
AU - Vermeiren, Robert R. J. M.
AU - Walter, Henrik
AU - Wang, Li
AU - Wang, Xin
AU - Weis, Carissa
AU - Winternitz, Sherry
AU - Xie, Hong
AU - Zhu, Ye
AU - Wall, Melanie
AU - Neria, Yuval
AU - Morey, Rajendra A.
N1 - Funding Information: Dr. Thompson received partial grant support from Biogen, Inc., and Amazon, Inc., for work unrelated to the current study; Dr. Lebois reports unpaid membership on the Scientific Committee for International Society for the Study of Trauma and Dissociation (ISSTD), grant support from the National Institute of Mental Health, K01 MH118467 and the Julia Kasparian Fund for Neuroscience Research, McLean Hospital. Dr. Lebois also reports spousal IP payments from Vanderbilt University for technology licensed to Acadia Pharmaceuticals unrelated to the present work. ISSTD and NIMH were not involved in the analysis or preparation of the manuscript; Dr. Etkin reports salary and equity from Alto Neuroscience, equity from Mindstrong Health and Akili Interactive. Other authors have no conflicts of interest to declare.Dr. Zhu is supported by NIH K01MH122774 and by a NARSAD Young Investigator Grant from the Brain & Behavior Research Foundation 27040; Dr. Dennis is supported by NIH R61NS120249; Dr. Jahanshad is supported by NIH R01MH117601; Dr. Thompson is supported by NIH U54 EB020403; Dr. Fani is supported by NIH AT011267 and MH111671; Dr. Bomyea is suppprted by NIH R61MH127005 and CX001600; Dr. Lebois is supported by NIH K01MH118467; Dr. Daniels is supported by German Research Foundation DA 1222/4–1; Dr. Disner is supported by VA RR&D Award IK2RX002922; Dr. Bruce is supported by NIH K23 MH090366; Dr. Bryant is supported by National Health and Medical Research Council #1073041; Dr. Ross is supported by the NIH T32MH018931, F31MH122047 and T32GM007507; Dr. Cisler is supported by NIMH MH119132 and MH097784; Dr. Morey is supported by NIMH MH111671 and MH129832. Funding Information: Dr. Zhu is supported by NIH K01MH122774 and by a NARSAD Young Investigator Grant from the Brain & Behavior Research Foundation 27040 ; Dr. Dennis is supported by NIH R61NS120249 ; Dr. Jahanshad is supported by NIH R01MH117601 ; Dr. Thompson is supported by NIH U54 EB020403 ; Dr. Fani is supported by NIH AT011267 and MH111671 ; Dr. Bomyea is suppprted by NIH R61MH127005 and CX001600 ; Dr. Lebois is supported by NIH K01MH118467 ; Dr. Daniels is supported by German Research Foundation DA 1222/4–1 ; Dr. Disner is supported by VA RR&D Award IK2RX002922 ; Dr. Bruce is supported by NIH K23 MH090366 ; Dr. Bryant is supported by National Health and Medical Research Council #1073041 ; Dr. Ross is supported by the NIH T32MH018931 , F31MH122047 and T32GM007507 ; Dr. Cisler is supported by NIMH MH119132 and MH097784 ; Dr. Morey is supported by NIMH MH111671 and MH129832 . Publisher Copyright: © 2023 The Author(s)
PY - 2023/12/1
Y1 - 2023/12/1
N2 - 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.
AB - 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.
KW - Classification
KW - Deep learning
KW - Machine learning
KW - Multimodal MRI
KW - Posttraumatic stress disorder
UR - http://www.scopus.com/inward/record.url?scp=85174730385&partnerID=8YFLogxK
U2 - https://doi.org/10.1016/j.neuroimage.2023.120412
DO - https://doi.org/10.1016/j.neuroimage.2023.120412
M3 - Article
C2 - 37858907
SN - 1053-8119
VL - 283
JO - NEUROIMAGE
JF - NEUROIMAGE
M1 - 120412
ER -