Integrative intratumoral and peritumoral radiomic model for predicting endometrial cancer grade.

Non-invasive preoperative assessment of tumor grade serves an important role for clinical management. The aim of the present study was to assess the value of an intratumoral and an peritumoral radiomic model in predicting the histological grade of endometrial cancer (EC). A total of 107 patients with EC were retrospectively enrolled and randomly divided into the training (n=74) and test cohorts (n=33). Radiomic features were extracted from intratumoral and peritumoral regions (RT) with different expansion regions (1, 2 and 3 mm) using T2-weighted and the 5 to 8th phase of contrast-enhancement images. The diagnostic performance of several peritumoral features was compared with the maximum area under the curve (AUC) value. These intratumoral features were combined with peritumoral features to construct a fusion model. The AUCs for the RT model were 0.879 [95% confidence interval (CI), 0.797-0.962] for the training cohort and 0.869 (95% CI, 0.590-1.000) for the test cohort. The peritumoral model with a 3-mm expansion (RT-3) demonstrated superior performance, yielding AUCs of 0.934 (95% CI, 0.875-0.994) in the training cohort and 0.875 (95% CI, 0.744-1.000) in the test cohort. The fusion model incorporating features from both RT and RT-3 achieved the highest diagnostic performance for distinguishing low-grade from high-grade EC, with AUCs of 0.955 (95% CI, 0.910-1.000) and 0.885 (95% CI, 0.771-1.000) in the training and test cohorts, respectively. In conclusion, the results of the present study indicate that radiomic features from magnetic resonance images incorporating both intratumoral and peritumoral regions can effectively predict low-and high-grade EC.