CT Radiomics Predicts Capsular Invasion And NI Risk

In a 2026 retrospective cohort of 111 thyroid carcinoma patients, researchers extracted CT radiomic features and developed machine-learning models to predict capsular invasion and stratify neural invasion (NI) risk. Clinical and radiomic nomograms achieved internally estimated AUCs of 0.9418 and 0.9334 respectively, while a multimodal neural network reached AUC 0.775. Radiomic-only models show potential for preoperative NI risk stratification, though clinical data improve postoperative assessment.