A multi-lesion radiomics model for predicting progression-free survival in patients with unresectable colorectal liver metastases undergoing chemotherapy

To develop and validate a machine learning model based on clinical variables and multi-lesion, multiphase contrast-enhanced CT radiomics for predicting progression-free survival (PFS) in patients with unresectable colorectal liver metastases (CRLM). This retrospective cohort study included 159 patients with unresectable CRLM who received first-line chemotherapy. Patients were divided into a training cohort ( n = 127) and an internal hold-out set ( n = 32) using stratified sampling. All patients underwent contrast-enhanced CT before treatment. For each patient, all measurable liver metastases were segmented separately on arterial-phase (A) and portal venous-phase (V) images, and radiomic features were extracted. A feature selection pipeline based on the Criteria Importance Through Intercriteria Correlation (CRITIC) objective weighting method and multi-criteria decision making (MCDM) was applied. Multiple machine learning classifiers were trained on unimodal data, and the base learners were integrated using a performance-weighted soft-voting ensemble strategy. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, calibration curves, and decision curve analysis (DCA). Model interpretability was assessed using Shapley additive explanations (SHAP). The trimodal arterial-phase, portal venous-phase, and clinical-variable ensemble model (AVC) showed encouraging discriminative performance in the internal hold-out set, with an AUC of 0.792 (95% CI: 0.625–0.933), although the study was underpowered for formal model comparisons. DCA suggested a favorable net benefit within part of the clinically relevant threshold range, although this finding requires confirmation in larger external cohorts. The proposed CRITIC_MCDM feature selection method achieved the highest numerical AUC among the compared feature-selection methods, although the differences were not statistically significant. Key predictive features included arterial-phase first-order (FO) statistics (e.g., A_FO_10Pct_sum), portal venous-phase FO skewness and gray-level dependence matrix (GLDM) features (e.g., V_FO_Skew_mean), and clinical biomarkers such as CA19-9, CEA, and primary tumor resection status. A multi-lesion, multiphase CT-based radiomics framework combined with CRITIC_MCDM feature selection and performance-weighted soft voting showed promise for stratifying 12-month PFS risk in patients with unresectable CRLM undergoing first-line chemotherapy. This non-invasive and interpretable approach may support individualized risk assessment, but prospective external validation is needed before clinical application, given the current limitation of being underpowered for model comparisons.