Answers are grounded in this paper and cite their sources.
Objective To investigate the value of a combined model based on T1-weighted enhancement (T1CE), T2-weighted imaging (T2WI), and diffusion-weighted imaging (DWI) sequences of lymph node radiomics features combined with clinical features in the preoperative prediction of lymph node metastasis (LNM) in rectal cancer. Methods A retrospective cohort from two hospitals (June 2019–Jan 2025) included rectal cancer patients with preoperative MRI. Lymph nodes from Yangpu Hospital (135 nodes, 40 patients) were split into training (n=108) and internal test (n=27) sets; 51 nodes from Shanghai Fourth People’s Hospital served as external test set. Radiomic features from segmented T1CE/T2WI/DWI images and clinical risk factors were used to develop radiomics, clinical, and combined clinical–radiomics models, evaluated for discrimination and calibration. Results The standalone clinical model exhibited suboptimal performance in the training, test, and external test sets, with AUCs of 0.699, 0.779, and 0.663, respectively. Although each individual radiomics model achieved relatively high AUCs in the training set, a general decline in performance was observed in the external test set. The combined model maintained high AUCs across all datasets (training set: 1.000; test set: 0.888; external test set: 0.848). Calibration curve and decision curve analyses also indicated that the combined model had better consistency in predicted probabilities and greater clinical net benefit. Conclusions The combined model integrating lymph node radiomic features from T1CE, T2WI, and DWI with clinical variables demonstrated potential for the preoperative prediction of LNM in rectal cancer. By providing additional quantitative information on nodal status, the model may complement established clinical and MRI risk factors and support more refined preoperative risk stratification.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: