Radiomics model with reduced field-of-view DWI for predicting LVSI and high-risk histological type in endometrial cancer

Motivation: Although reduced field of view (rFOV) DWI is essential for preoperative diagnosis and staging of endometrial cancer (EC), few studies have investigated its potential for predicting EC prognostic factors using radiomics analysis. Goal(s): Our study aims to predict substantial/extensive LVSI and high-risk histological type in EC using radiomics analysis with rFOV DWI. Approach: We analyzed 206 patients from 2 hospitals, extracting radiomic features from multiple sequences, including rFOV DWI. We utilized a Bayesian model incorporating radiomic and clinical data, including CA125 and hospital differences. Results: The Bayesian model showed high accuracy for LVSI (93.5%) and moderate for high-risk histology (80.6%). Impact: This study impacts endometrial cancer care by radiomics analysis using Bayesian modeling with MRI-derived radiomic features to predict lymphovascular space invasion and high-risk histology, supporting personalized treatment. Findings suggest the potential of rFOV DWI.