3D spectral CT-based fusion model predicts prognosis and postoperative adjuvant chemotherapy benefit in locally advanced rectal cancer.

Early identification of high-risk recurrence patients is crucial as it can provide information for treatment decisions, achieve personalized treatment, and maximize the benefits for locally advanced rectal cancer (LARC). The goal of this study was to explore the impact of ROI selection on the repeatability of spectral detector CT (SDCT) imaging biomarkers and to investigate and visualize an interpretable multidimensional radiological-angiogenesis-clinicopathological integrated model (RACIM) for predicting disease-free survival (DFS) in LARC. 204 LARC patients who underwent SDCT scanning prior to any anticancer treatment were retrospectively included. Two observers independently measured the iodine concentrations and normalized iodine concentration (NIC) at venous/delayed phases (VP/DP) using two different ROI protocols (2D vs. 3D). Cox regression methods were applied to determine the independent risk predictors associated with DFS, which were used to develop a prediction model. Kaplan-Meier survival curves were used to explore the performance of RACIM for personalized risk stratification. The models' performance was assessed using the area under the receiver operating characteristic (ROC) curve (AUC). The SHapley Additive exPlanation (SHAP) algorithm was used to attribute interpretability to the optimal prediction model. The 3D-ROI method achieved satisfactory interobserver accuracy. Multivariate Cox regression analyses indicated that perineural invasion, CD105, histological grade and NICVP