We found that MVPA of parcellated structural neuroimaging data is unable to provide accurate distinction between OCD cases and HC. Classification of the complete sample using site-stratified CV ranged between an AUC of 0.57 and 0.61, which is not sufficient for clinical application. Differences in performance between classifiers were minimal. Similar results were obtained for classifications performed separately on pediatric and adult samples. When validated on completely new data from other sites using LOSO-CV, model performance hardly exceeded chance-level (0.5 AUC).
