Radiomics using CT images for preoperative prediction of lymph node metastasis in perihilar cholangiocarcinoma: a multi-centric study
Peng-chao Zhan,Ting Yang,Yuan Zhang,Ke-yan Liu,Zhen Li,Yu-yuan Zhang,Xing Liu,Na-na Liu,Hui-xia Wang,Bo Shang,Yan Chen,Han-yu Jiang,Xiang-tian Zhao,Jing-hai Shao,Zhe Chen,Xin-dong Wang,Kang Wang,Jian-bo Gao,Pei-jie Lyu
DOI: https://doi.org/10.1007/s00330-023-10108-1
IF: 7.034
2023-08-17
European Radiology
Abstract:ObjectivesTo develop a CT-based radiomics model for preoperative prediction of lymph node (LN) metastasis in perihilar cholangiocarcinoma (pCCA).MethodsThe study enrolled consecutive pCCA patients from three independent Chinese medical centers. The Boruta algorithm was applied to build the radiomics signature for the primary tumor and LN. The k-means algorithm was employed to cluster the selected LNs based on the radiomics signature LN. Support vector machines were used to construct the prediction models. The diagnostic efficiency was measured by the area under the receiver operating characteristic curve (AUC). The optimal model was evaluated in terms of calibration, clinical usefulness, and prognostic value.ResultsA total of 214 patients were included in the study (mean age: 61.6 years ± 9.4; 130 male). The selected LNs were classified into two clusters, which were significantly correlated with LN metastasis in all cohorts (p < 0.001). The model incorporated the clinical risk factors, radiomics signature primary tumor, and the LN cluster obtained the best discrimination, with AUC values of 0.981 (95% CI: 0.962–1), 0.896 (95% CI: 0.810–0.982), and 0.865 (95% CI: 0.768–0.961) in the training, internal validation, and external validation cohorts, respectively. High-risk patients predicted by the optimal model had shorter overall survival than low-risk patients (median, 13.7 vs. 27.3 months, p < 0.001).ConclusionsThe study proposed a radiomics model with good performance to predict LN metastasis in pCCA. As a noninvasive preoperative prediction tool, this model may help in patient risk stratification and personalized treatment.Clinical relevance statementA CT-based radiomics model accurately predicts lymph node metastasis in perihilar cholangiocarcinoma patients. This noninvasive preoperative tool can aid in patient risk stratification and personalized treatment, potentially improving patient outcomes.Key Points• The radiomics model based on contrast-enhanced CT is a useful tool for preoperative prediction of lymph node metastasis in perihilar cholangiocarcinoma.• Radiomics features extracted from lymph nodes show great potential for predicting lymph node metastasis.• The study is the first to identify a lymph node phenotype with a high probability of metastasis based on radiomics.
radiology, nuclear medicine & medical imaging