• magnetic resonance
• neoadjuvant therapy
• radiomics
• rectal cancer

Background and Objective: Rectal cancer represents a significant clinical challenge, particularly in patients with locally advanced rectal cancer (LARC). Despite advances in neoadjuvant therapy aimed at improving surgical outcomes and reducing the risk of local recurrence, predicting a pathological complete response (pCR) continues to be a challenge. This study proposes a radiomics-based machine learning model to identify patients achieving pCR after neoadjuvant therapy by analysing radiomic features extracted from post-treatment MRI scans. Methods: This retrospective study included 86 patients treated at Pisa University Hospital between 2017 and 2022, using oblique axial T2-weighted images for manual tumor segmentation. A total of 107 radiomic features were extracted and selected using LASSO and logistic regression with ElasticNet regularization. Results: The model achieved a mean AUC-ROC of 74% and an accuracy of 67.4%, demonstrating high reliability in identifying non-responders (TNR 71%). Conclusions: These findings support the use of radiomics as a complementary tool for personalizing therapeutic strategies.