Prediction of Tumor Response Grading in Rectal Cancer Neoadjuvant Chemoradiotherapy: An Imaging Biomarker Analysis of Habitat Radiomics

Background: Tumor regression grading (TRG) is a critical predictor of treatment outcomes in rectal cancer. Conventional assessment methods may fail to capture the full complexity of intratumoral heterogeneity. Advances in imaging, particularly radiomics and habitat-based analysis, offer the potential to enhance TRG prediction by characterizing subregional tumor features. This study evaluates the effectiveness of these techniques in improving TRG prediction. Methods: Computed tomography (CT) images were analyzed to comparethe predictive performance of conventional radiomics features and habitat-based analysis. Tumor regions of interest (ROIs) were segmented, extracting local imaging features. Voxel-level clustering was employed to identify distinct intratumoral subregions. Machine learning algorithms, including ExtraTrees, support vector machine (SVM), and Random Forest, were applied to predict TRG. Results: The ExtraTrees algorithm achieved superior performance, with AUCs of 0.912 and 0.817in training and testing cohorts, respectively, outperforming SVM and Random Forest. In contrast, the conventionalradiomics model had an AUC of 0.504. Integrating habitat features with clinical variables further enhanced predictive accuracy, yielding AUCs of 0.916 (training) and 0.833 (testing), with good calibration confirmed by the Hosmer–Lemeshow test. Conclusion: By integrating imaging and clinical data, habitat-based radiomics significantly improves TRG prediction and advances personalized medicine. Future research should validate these approaches in larger, independent cohorts to support clinical translation.