A clinical-radiomics nomogram for the preoperative prediction of lung metastasis in colorectal cancer patients with indeterminate pulmonary nodules

A clinical-radiomics nomogram for the preoperative prediction of lung metastasis in colorectal cancer patients with indeterminate pulmonary nodules

European Radiology, June 2018

29948074

TingDan Hu, ShengPing Wang, Lv Huang, JiaZhou Wang, DeBing Shi, Yuan Li, Tong Tong, Weijun Peng

To develop and validate a clinical-radiomics nomogram for preoperative prediction of lung metastasis for colorectal cancer (CRC) patients with indeterminate pulmonary nodules (IPN). 194 CRC patients with lung nodules were enrolled in this study (136 in the training cohort and 58 in the validation cohort). To evaluate the probability of lung metastasis, we developed three models, the clinical model with significant clinical risk factors, the radiomics model with radiomics features constructed by the least absolute shrinkage and selection operator algorithm, and the clinical-radiomics model with significant variables selected by the stepwise logistic regression. The Akaike information criterion (AIC) was used to compare the relative strength of different models, and the area under the curve (AUC) was used to quantify the predictive accuracy. The nomogram was developed based on the most appropriate model. Decision-curve analysis was applied to assess the clinical usefulness. The clinical-radiomics model (AIC = 98.893) with the lowest AIC value compared with that of the clinical-only model (AIC = 138.502) or the radiomics-only model (AIC = 116.146) was identified as the best model. The clinical-radiomics nomogram was also successfully developed with favourable discrimination in both training cohort (AUC = 0.929, 95% CI: 0.885-0.974) and validation cohort (AUC = 0.922, 95% CI: 0.857-0.986), and good calibration. Decision-curve analysis confirmed the clinical utility of the clinical-radiomics nomogram. In CRC patients with IPNs, the clinical-radiomics nomogram created by the radiomics signature and clinical risk factors exhibited favourable discriminatory ability and accuracy for a metastasis prediction. • Clinical features can predict lung metastasis of colorectal cancer patients. • Radiomics analysis outperformed clinical features in assessing the risk of pulmonary metastasis. • A clinical-radiomics nomogram can help clinicians predict lung metastasis in colorectal cancer patients.