Toward trustworthy clinical AI for obsessive-compulsive disorder: reliability, generalizability, and interpretability of a transformer model across the ENIGMA-OCD consortium.
Обработка статьи · этапы, ошибки и управление
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Pak Maria, Ryu Youngchan, Bae Sangyoon, Anticevic Alan, Costa Ana Daniela, Thorsen Anders L, van der Straten Anouk L, Couto Beatriz, Vai Benedetta, Hansen Bjarne, Soriano-Mas Carles, Li Chiang-Shan R, Vriend Chris, Lochner Christine, Pittenger Christopher, Moreau Clara A, Rodriguez-Manrique Daniela, Vecchio Daniela, Shimizu Eiji, Stern Emily R, Muñoz-Moreno Emma, Nurmi Erika L, Piras Fabrizio, Colombo Federica, Piras Federica, Jaspers-Fayer Fern, Benedetti Francesco, Venkatasubramanian Ganesan, Eng Goi Khia, Simpson H Blair, Ruan Hanyang, Hu Hao, van Marle Hein J F, Tomiyama Hirofumi, Martínez-Zalacaín Ignacio, Feusner Jamie, Narayanaswamy Janardhanan C, Yun Je-Yeon, Sato Joao R, Ipser Jonathan, Pariente Jose C, Menchón Jose M, O'Neill Joseph, Kwon Jun Soo, Koch Kathrin, Hagen Kristen, Backhausen Lea, Waller Lea, Lazaro Luisa, Batistuzzo Marcelo C, Hoexter Marcelo Q, Picó-Pérez Maria, Kim Minah, Dzinalija Nadza, Beyer Nicole, Vetter Nora C, Gruner Patricia, Morgado Pedro, Szeszko Philip R, Alonso Pino, Zhao Qing, Marsh Rachel, Stewart S Evelyn, Bertolín Sara, Brem Silvia, Thomopoulos Sophia I, Balachander Srinivas, Walitza Susanne, Yoshida Tokiko, Nakao Tomohiro, Shivakumar Venkataram, van Leeuwen Wieke, Reddy Y C Janardhan, Abe Yoshinari, Hirano Yoshiyuki, Cho Youngsun, van der Werf Ysbrand D, Ikemizu Yuki, Sakai Yuki, Wang Zhen, ENIGMA-OCD Working Group, Thompson Paul M, Bruin Willem, van Wingen Guido, Stein Dan J, van den Heuvel Odile A, Cha Jiook
Аннотация
METHODS: We applied a transformer-based deep learning model, the Multi-Band Brain Net, to the ENIGMA-OCD cohort - the largest available resting-state functional magnetic resonance imaging (rs-fMRI) dataset in OCD with 1,706 participants (869 cases with OCD, 837 controls) across 23 sites worldwide. We evaluated model reliability by calculating calibration - the model's ability to "know what it doesn't know". We assessed generalizability using leave-one-site-out validation to test performance on unseen sites with different scanners, acquisition protocols, and patient populations. Finally, we examined interpretability by analyzing model attention weights to identify the neural connectivity patterns that influence model predictions.
RESULTS: The model achieved modest but competitive classification performance (AUROC = .653 ± .039). Crucially, while large-scale pretraining on the UK Biobank (N = 40,783) did not boost accuracy, it significantly enhanced model calibration by reducing overconfident predictions. Leave-one-site-out validation showed a generalization gap across sites (AUROC = .427-.819). Pretraining did not close this gap but removed scanner manufacturer bias. Finally, attention-based mapping identified biologically plausible patterns of widespread hypoconnectivity in OCD relative to healthy controls, particularly in low-frequency bands involving the default mode, salience, and somatomotor networks. These findings aligned with known OCD neurobiology.
CONCLUSIONS: This study provides a framework for developing more reliable and trustworthy clinical artificial intelligence for OCD.