Deep Learning Diagnoses Extracranial Carotid Plaques Effectively

A systematic review and meta-analysis in J Med Internet Res (2026) evaluated radiomics and deep learning models for diagnosing extracranial carotid plaques, searching five databases through September 24, 2025, and including 40 studies with 17,246 patients. Pooled sensitivity was 0.88, specificity 0.89, and area under the SROC curve 0.95; transfer learning and larger sample sizes improved performance, while external validation showed lower accuracy. Authors report high heterogeneity and call for multicenter studies and standardized designs to improve generalizability.