Sara Bertolín

Background. Studies applying machine learning to obsessive-compulsive disorder (OCD) typically report accuracy in homogeneous samples but rarely assess model reliability, generalizability, and interpretability needed for clinical use. 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, SD = .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.

Objective Cognitive behavioral therapy (CBT) is an effective first-line treatment for obsessive-compulsive disorder (OCD), yet it remains difficult to predict who will respond to this intervention. This study investigates associations between neural activity during inhibitory control tasks and CBT outcomes, and whether task-based fMRI data could serve as a predictive marker of individual CBT response. Methods Using fMRI data from individuals performing an inhibitory control task across five samples (n=130, age range=8-57, 54% female) of the ENIGMA-OCD consortium, univariate associations were analyzed between activity during response inhibition and error processing and three CBT outcomes: response, remission, and pre-post treatment change in symptom severity. Random forest and support vector machine models using leave-one-sample-out cross-validation were used for prediction of CBT response and remission from fMRI activity and clinical data. Results Remission after CBT was associated with weaker activity in default mode regions during response inhibition and in the right supramarginal gyrus during error processing. Greater symptom reduction was linked to weaker pre-treatment activity across frontoparietal, dorsal attention, visual, and subcortical regions during response inhibition, but to stronger default mode activity during error processing. Despite these robust group-level effects, machine learning models failed to predict individual outcomes above chance level with either neuroimaging or clinical data. Conclusion Weaker activity during response inhibition in a widespread network, as well as stronger activity in default mode regions during error processing before treatment, appear beneficial to CBT response. However, these findings cannot yet be translated into individually predictive markers of CBT outcome.

Brain disorders are increasingly understood as disorders of distributed brain circuits, yet functional connectivity (FC), the dominant framework for mapping them, treats the brain as a collection of pairwise relationships between regions and cannot represent pathology distributed across coordinated sets of connections. We introduce a Hodge-Laplacian topological framework that localizes higher-order “loop” (1-cycle) organization within functional connectome, maps each loop to specific edges and networks, and yields a subject-level measure of loop expression. Applied to resting-state fMRI from the ENIGMA-OCD consortium (1,024 patients and 1,028 controls across 28 sites), the framework identified 93 loop-level abnormalities in obsessive–compulsive disorder (OCD), concentrated in frontoparietal and somatomotor systems. The edges forming these loops largely showed no significant differences between groups, indicating that the abnormalities were invisible to conventional FC analysis. The frontoparietal and somatomotor loop clusters recurred across the clinical subgroups, suggesting convergence on a shared higher-order phenotype. Robustness analyses showed the loop signal reflected higher-order organization rather than an artifact of individual edges, the network backbone, or any single site. These results indicate that coordinated, multi-edge pathology exists and can be localized even when pairwise analyses fail to detect it, positioning higher-order topology as a generalizable axis for mapping circuit pathology across psychiatric and neurological disorders.

White matter pathways, typically studied with diffusion tensor imaging (DTI), have been implicated in the neurobiology of obsessive-compulsive disorder (OCD). However, due to limited sample sizes and the predominance of single-site studies, the generalizability of OCD classification based on diffusion white matter estimates remains unclear. Here, we tested classification accuracy using the largest OCD DTI dataset to date, involving 1336 adult participants (690 OCD patients and 646 healthy controls) and 317 pediatric participants (175 OCD patients and 142 healthy controls) from 18 international sites within the ENIGMA OCD Working Group. We used an automatic machine learning pipeline (with feature engineering and selection, and model optimization) and examined the cross-site generalizability of the OCD classification models using leave-one-site-out cross-validation. Our models showed low-to-moderate accuracy in classifying (1) “OCD vs. healthy controls” (Adults, receiver operator characteristic-area under the curve = 57.19 ± 3.47 in the replication set; Children, 59.8 ± 7.39), (2) “unmedicated OCD vs. healthy controls” (Adults, 62.67 ± 3.84; Children, 48.51 ± 10.14), and (3) “medicated OCD vs. unmedicated OCD” (Adults, 76.72 ± 3.97; Children, 72.45 ± 8.87). There was significant site variability in model performance (cross-validated ROC AUC ranges 51.6–79.1 in adults; 35.9–63.2 in children). Machine learning interpretation showed that diffusivity measures of the corpus callosum, internal capsule, and posterior thalamic radiation contributed to the classification of OCD from HC. The classification performance appeared greater than the model trained on grey matter morphometry in the prior ENIGMA OCD study (our study includes subsamples from the morphometry study). Taken together, this study points to the meaningful multivariate patterns of white matter features relevant to the neurobiology of OCD, but with low-to-moderate classification accuracy. The OCD classification performance may be constrained by site variability and medication effects on the white matter integrity, indicating room for improvement for future research.

Current knowledge about functional connectivity in obsessive-compulsive disorder (OCD) is based on small-scale studies, limiting the generalizability of results. Moreover, the majority of studies have focused only on predefined regions or functional networks rather than connectivity throughout the entire brain. Here, we investigated differences in resting-state functional connectivity between OCD patients and healthy controls (HC) using mega-analysis of data from 1024 OCD patients and 1028 HC from 28 independent samples of the ENIGMA-OCD consortium. We assessed group differences in whole-brain functional connectivity at both the regional and network level, and investigated whether functional connectivity could serve as biomarker to identify patient status at the individual level using machine learning analysis. The mega-analyses revealed widespread abnormalities in functional connectivity in OCD, with global hypo-connectivity (Cohen’s d: -0.27 to -0.13) and few hyper-connections, mainly with the thalamus (Cohen’s d: 0.19 to 0.22). Most hypo-connections were located within the sensorimotor network and no fronto-striatal abnormalities were found. Overall, classification performances were poor, with area-under-the-receiver-operating-characteristic curve (AUC) scores ranging between 0.567 and 0.673, with better classification for medicated (AUC = 0.702) than unmedicated (AUC = 0.608) patients versus healthy controls. These findings provide partial support for existing pathophysiological models of OCD and highlight the important role of the sensorimotor network in OCD. However, resting-state connectivity does not so far provide an accurate biomarker for identifying patients at the individual level.

Background: Although the consequences of the COVID-19 pandemic on emotional health are evident, little is known about its impact on patients with obsessive-compulsive disorder (OCD). Methods: One hundred and twenty-seven patients with OCD who attended a specialist OCD Clinic in Barcelona, Spain, were assessed by phone from April 27 to May 25, 2020, during the early phase of the pandemic, using the Yale–Brown Obsessive–Compulsive Scale (Y-BOCS) and a structured interview that collected clinical and sociodemographic information. Results were compared with those for 237 healthy controls from the same geographic area who completed an online survey. Results: Although 65.3% of the patients with OCD described a worsening of their symptoms, only 31.4% had Y-BOCS scores that increased >25%. The risk of getting infected by SARS-CoV2 was reported as a new obsession by 44.8%, but this only became the main obsessive concern in approximately 10% of the patients. Suicide-related thoughts were more frequent among the OCD cohort than among healthy controls. The presence of prepandemic depression, higher Y-BOCS scores, contamination/washing symptoms, and lower perceived social support all predicted a significantly increased risk of OCD worsening. Conclusions: Most patients with OCD appear to be capable of coping with the emotional stress of the COVID-19 outbreak and its consequences during the initial phase of the pandemic. Nevertheless, the current crisis constitutes a risk factor for a significant worsening of symptoms and suicidal ideation. Action is needed to ensure effective and individualized follow-up care for patients with OCD in the COVID-19 era.