Prediction of Ki-67 expression in hepatocellular carcinoma with a computed tomography (CT) extracellular volume-derived nomogram

AIM

To investigate and verify the ability of the extracellular volume fraction (fECV) during the delayed computed tomography (CT) phase to noninvasively predict the preoperative Ki-67 expression level in hepatocellular carcinoma (HCC).

MATERIALS AND METHODS

The clinical and imaging data of 94 patients with HCC, pathologically diagnosed according to preoperative enhanced CT at our hospital, were retrospectively analysed The patients were randomly divided into a training group (66 patients) and a validation group (28 patients) at a 7:3 ratio. Univariable and multivariable logistic regression analyses were used to identify clinical risk factors, which were integrated with the fECV model to generate a joint nomogram model, whose performance was assessed using the Akaike information criterion (AIC), area under the curve (AUC), accuracy, sensitivity, and specificity. The clinical utility of the models was assessed via decision curve analysis (DCA).

RESULTS

In multivariate analysis, tumour capsule (OR = 0.178, P = 0.013) and fECV (OR = 1.282, P < 0.001) were independent predictors of high Ki-67 levels. The AUCs of the joint nomogram model constructed from these predictors and the fECV model were greater than those of the fECV model alone in the training and test sets, but the differences were not significant (P > 0.05, DeLong test). Moreover, the nomogram model had the lowest AIC value (21.09), indicating that it was the best model, and showed good clinical utility in both the training and validation sets.

CONCLUSION

The combined nomogram model based on the delayed-phase fECV has potential value in predicting individualised preoperative Ki-67 expression levels in HCC patients.