A Multisequence MRI-based Deep Learning Radiomics Nomogram for Predicting Ki-67 Expression and Prognosis in Locally Advanced Rectal Cancer

Motivation: Accurately identifying Ki-67 expression and prognosis is crucial for guiding treatment decisions in locally advanced rectal cancer (LARC) patients who decline preoperative chemo-/radiotherapy. Goal(s): To develop a deep learning (DL) radiomics nomogram using multisequence MRI for predicting Ki-67 expression and prognosis in LARC patients who decline preoperative chemo-/radiotherapy. Approach: Radiomics and DL features from multisequence MRI were used to develop and evaluate six machine learning models. The optimal DL and radiomics models were combined into a nomogram to predict Ki-67 expression and disease-free survival (DFS). Results: The nomogram can effectively predict Ki-67 expression and DFS in LARC who decline preoperative chemo-/radiotherapy. Impact: This study highlights the potential of a nomogram that combines radiomics and DL to assess Ki-67 expression and prognosis in LARC patients who decline preoperative chemo-/radiotherapy, offering a promising tool for individualized treatment planning in this patient group.