Multiparametric radiomics nomogram may be used for predicting the severity of esophageal varices in cirrhotic patients

Background: To explore whether a multiparametric radiomics nomogram on computed tomography (CT) images based on radiomics and relevant parameters of esophageal varices (EV) can be used for predicting the EV severity in patients with cirrhotic livers. Methods: From January 2016 to August 2018, 136 consecutive patients with clinicopathologically confirmed liver cirrhosis were included for the development of a predictive model. The patients were then divided into two groups, including non-conspicuous EV group (mild-to-moderate EV, n=30) and conspicuous EV group (severe EV, n=106) by using the endoscopic validation as the reference standard. The radiomic scores (Rad scores) were constructed using the binary logistic regression model from the radiomics features of regions of interest (ROIs) in the left liver (LL) and right liver (RL), respectively. The multiparametric nomogram combined the best performance Rad-score and EV-relevant factors, and the calibration, discrimination, and clinical usefulness of developed nomogram were evaluated using calibration curves, decision curve analysis (DCA) and net reclassification index (NRI) analysis respectively. Results: The LL Rad-score calculated from radiomics features was selected with a relatively higher area under the curve (AUC) (AUC; 0.88, training cohort; 0.87, the validation cohort) compared with RL Radscore (AUC; 0.86, training cohort; 0.83, the validation cohort). In addition, cross-sectional surface area (CSA) was identified as the important predictor (P<0.05), the multiparametric nomogram containing LL Rad-score and CSA was shown to have a better predictive performance and good calibration in the training model (C-index, 0.953, 95% CI, 0.892 to 0.973) and the validation cohort (C-index, 0.938, 95% CI, 0.841 to 0.961), resulting in an improved NRI (categorical NRI of 25.9%, P=0.0128; continuous NRI of 120%, P<0.001) and integrated discriminatory improvement (IDI) (IDI =13.9%, P<0.001). DCA demonstrated that the multiparametric radiomics nomogram was clinically useful. Conclusions: A multiparametric radiomics nomogram, which incorporates the liver radiomics signature and EV-relevant indices, is a useful tool for noninvasively predicting EV severity and may complement the standard endoscopy for evaluating EV severity in patients with cirrhosis.