To construct a preoperative prediction model for proliferative hepatocellular carcinoma(HCC) based on enhanced CT image features, and to analyze the prognosis of the disease.

A retrospective case-control study was conducted on 603 patients with pathologically confirmed HCC. Among them, 519 cases from the First Affiliated Hospital of Army Medical University were randomly divided into a training group(n=363) and an internal verification group(n=156) in a ratio of 7:3. Another 84 patients from the Second Affiliated Hospital of Chongqing Medical University served as an external validation group. All patients underwent abdominal CT scan with contrast before surgery. The clinical data and CT imaging characteristics of proliferative and non-proliferative HCC patients were observed. Binary logistic regression analysis was used to identify the independent risk factors of proliferative HCC, and a nomogram prediction model was constructed. Receiver operating characteristic(ROC) curve was plotted to evaluate its diagnostic performance, and calibration curve and decision curve analysis(DCA) were applied to evaluate its calibration performance and clinical application value. The model was validated in both the internal and external validation groups. Kaplan-Meier survival curves were employed to compare the prognosis between proliferative and non-proliferative HCC.

Multivariate analysis showed that negative result of HBV-DNA quantification, incomplete tumor capsule, intratumoral necrosis or ischemia(≥20%), and annular hyperenhancement in arterial phase were independent predictors in predicting proliferative HCC(P<0.05). Our nomogram model for predicting proliferative HCC based on the above clinical imaging features had an AUC value of 0.805(95%CI: 0.756~0.854), a sensitivity of 83.19% and a specificity of 64.80% in the training group. For the internal validation group, the AUC value was 0.793(95%CI: 0.721~0.854), the sensitivity was 67.86%, and the specificity was 75.00%. In the external validation group, the AUC value was 0.842(95%CI: 0.746~0.912), the sensitivity was 72.41%, and the specificity was 90.91%. Calibration curve and DCA showed that the model had good calibration performance and clinical applicability. The disease free survival(DFS) of the patients with proliferative HCC diagnosed by pathologically confirmed/predictive models was significantly shorter than that of non-proliferative HCC patients(training group: P=0.005, P<0.001; internal validation group: P=0.006, P=0.006; external validation group: P=0.002, P=0.015).

Our prediction model based on clinical and imaging features can accurately diagnose proliferative HCC before surgery, and the prognosis of proliferative HCC is generally poor.