MentalHealthDaily

OCD Clinical AI Reached AUROC 0.653 but Failed Site Generalization

by GLOOM

MHD featured image for OCD clinical AI and ENIGMA fMRI generalization.

A 2026 ENIGMA-OCD preprint tested a transformer model on resting-state fMRI from 1,706 participants and reached AUROC = 0.653 ± 0.039 for OCD classification.1 The more important result was the failure mode: held-out-site performance ranged from AUROC 0.427 to 0.819, which is exactly the gap clinical AI has to close before brain-scan diagnosis is credible. …

Alzheimer’s Blood Biomarker Models Lose Rule-Out Power Across Cohorts

by GLOOM

Editorial card showing Alzheimer's blood biomarker model transfer across cohorts with PET scan and blood assay imagery.

A 2026 ADNI/A4 validation study found that Alzheimer’s plasma-biomarker machine-learning models still ranked amyloid PET status well across cohorts, but the practical rule-out number moved hard: negative predictive value fell from 0.831 inside ADNI to 0.644 when the ADNI-trained model was applied to A4.1 Research Highlights 1,707-person ADNI/A4 test: researchers trained amyloid PET prediction models …