Radiomic signature based on enhanced CT and 3T MRI for survival analysis in patients with esophageal squamous carcinoma

We enrolled 478 patients with esophageal squamous carcinoma (ESCC) split them into a training and a test cohort with a ratio of 7 to 3. Radiomic features were extracted from lesions on both MR and CT images and used to build models for predicting disease-free survival (DFS) and overall survival (OS). For both MR and CT images, the radiomic signature combined with clinical variables achieved performance comparable to radiological signature. Over the test cohort, MRI-based models achieved C-index values of 0.707 and 0.663 for DFS and OS predictions, respectively; CT-based models achieved 0.731 and 0.68 for DFS and OS, respectively.