An interpretable machine learning model combining MRI-DKI habitat radiomic features and clinical biomarkers for noninvasive prediction of lymphatic metastasis in rectal cancer: a prospective study

Accurate assessment of lymphatic metastasis risk in rectal cancer is crucial for clinical decision-making and personalized treatment optimization. Diffusion kurtosis imaging-derived parameters and habitat radiomic features can quantify and characterize intratumoral heterogeneity. The combined model provides higher predictive performance for LVI and LNM in rectal cancer.