Elena Brown
In order to predict high-risk Endometrial Cancer (EC) before surgery, it is necessary to develop and validate radiomics models based on Magnetic Resonance Imaging (MR) that can estimate Deep Myometrial Invasion (DMI) and Lymphovascular Space Invasion (LVSI), as well as distinguish between low-risk and other categories of risk as recommended by ESGO/ESTRO/ESP (European Society of Gynecological OncologyEuropean Society for Radiotherapy & Oncology and European Society of Pathology. The 96 women with EC who received 1.5-T MR imaging prior to surgical staging between April 2009 and May 2019 at two referral facilities were included in this retrospective analysis. The participants were split into training (T=73) and validation cohorts (V=23). The MODDICOM library was used to extract radiomics characteristics, and whole-tumor volume was manually delineated on MR images (axial T2- weighted). Using a subset of the most important texture features analyzed independently in univariate analysis using Wilcoxon-MannWhitney, the diagnostic abilities of radiomic models were assessed by area under the Receiver Operating Characteristic (ROC) Area Under Curve in Training (AUCT) and Area Under Curve Validation (AUCV) cohorts. After extracting a total of 228 radiomics characteristics, only 38 for DMI, 29 for LVSI, and 15 for risk-class prediction for logistic radiomic modelling remained. In DMI estimation, LVSI prediction, and separating low-risk from other risk classes (intermediate/high-intermediate/ high), whole-tumor radiomic models produced AUCT/AUCV values of 0.85/0.68, 0.92/0.81, and 0.84/0.76 respectively. In conclusion, enhanced prognostication in EC has a lot of potential for MRI-based radiomics.