China Oncology ›› 2020, Vol. 30 ›› Issue (1): 49-56.doi: 10.19401/j.cnki.1007-3639.2020.01.006

• Article • Previous Articles     Next Articles

Development and validation of a clinical-radiomics nomogram for preoperative prediction of lymph node metastasis in colorectal cancer

LI Menglei 1 , ZHANG Jing 2 , DAN Yibo 2 , YANG Guang 2 , YAO Yefeng 2 , TONG Tong 1   

  1. 1. Department of Radiology, Fudan University Shanghai Cancer Center, Department of Oncology, Shanghai Medical College, Fudan University, Shanghai 200032, China; 2. Shanghai Key Laboratory of Magnetic Resonance, East China Normal University, Shanghai 200062, China
  • Online:2020-01-30 Published:2020-01-17
  • Contact: TONG Tong E-mail: t983352@126.com

Abstract:  Background and purpose: Accurate preoperative prediction of lymph node metastasis (LNM) is very important for the prognosis and recurrence of patients with colorectal cancer (CRC). The purpose of our study was to develop and validate a clinical-radiomics nomogram for preoperative prediction of LNM for patients with CRC. Methods: We enrolled 767 patients treated in Fudan University Shanghai Cancer Center (537 in the primary cohort and 230 in the validation cohort) with clinicopathologically confirmed CRC. We included nine significant clinical risk factors [age, gender, preoperative carcinoembryonic antigen (CEA) level, preoperative carbohydrate antigen 19-9 (CA19-9) level, grade, histological type, tumor location, tumor size and M stage] to build the clinical model. We used ANOVA, Relief and recursive feature elimination (RFE) for feature selection (including clinical risk factors, imaging features of primary lesions and peripheral lymph nodes), established the classification models through logistic regression analysis and selected respective optimal models by one-standard-error rule. Then we combined the clinical risk factors, the primary lesion radiomics features and the peripheral lymph node radiomics features of the optimal models to establish combined prediction models. The performance of the model was assessed by area under curve (AUC) of the receiver operating characteristic (ROC). Finally, decision curve analysis (DCA) and nomogram were applied to assess the clinical usefulness. Results: The clinical-primary lesion radiomics-peripheral lymph node radiomics model with the highest AUC (0.743 0) was identified as the best model. This optimal clinical-radiomics model also showed good discrimination and calibration in both primary cohort and validation cohort. DCA demonstrated that the clinical-radiomics nomogram was useful for preoperative prediction in clinical practice. Conclusion: The present study proposed a clinical-radiomics nomogram created by the radiomics signature and clinical risk factors, which can be potentially applied in the individual preoperative prediction of LNM in patients with CRC.

Key words: Colorectal cancer, Lymph node metastasis, Preoperative prediction, Nomograms, Radiomics