Tumor ecology model based on quantitative parameters of DCE-MRI for predicting lymph node metastasis in rectal cancer

Motivation: Lymph node metastasis (LNM) significantly impacts treatment decisions and prognosis in rectal cancer (RC). Goal(s): This study investigates the construction of a landscape index model of the RC sub-habitat using DCE-MRI parameter maps to predict LNM in RC prior to treatment. Approach: DCE-MRI-derived Ktrans maps segmented each tumor into two habitat subregions based on Ktrans values. Landscape pattern indices were introduced to quantify the spatial features of habitat imaging, and a diagnostic model for LNM was constructed using machine learning. Results: In the validation cohort, the landscape index model at the subregion 1achieved an AUC of 0.840 and demonstrated biological interpretability. Impact: Accurately identifying the LN status of RC patients prior to initial treatment is crucial for determining treatment strategies. The proposed quantitative DEC-MRI tumor ecology model is a promising tool for predicting LNM in RC.