Su Y, Xu X, Zuo P, et al. - In this retrospective study involving 245 patients, researchers evaluated the performance of machine learning (ML)-based MRI radiomics analysis for distinguishing between uveal melanoma (UM) and other intraocular masses. The sample consisted of patients with intraocular masses (165 UMs and 80 other intraocular masses). Radiomics characteristics have been extracted from T1WI, T2WI, and contrast-enhanced T1-weighted images (CET1WI), respectively. The authors discovered that the optimal radiomics feature set was 10, 15, 15, and 24 for T1W, T2W, CET1W, and joint T2W and CET1W images, respectively. They observed that the accuracy of T1WI, T2WI, CET1WI, and the joint T2WI and CET1WI models ranged from 72.0 %–78.0 %, from 79.6 %–81.6 %, from 74.0 %–82.0 %, and from 76.0 %–86.0 % in the test set. The performance of ML classifiers in the combined model was better than the visual assessment performance in the training set and in all patients. The performance of the ML classifiers in the combined model was better than the visual assessment performance in the training set and in all patients.

Overall, they concluded that radiomics analysis represents a promising tool for separating UM from other intraocular masses.