Radiomics Analysis on Noncontrast CT for Distinguishing Hepatic Hemangioma (HH) and Hepatocellular Carcinoma (HCC)

BACKGROUND: To form a radiomic model on the basis of noncontrast computed tomography (CT) to distinguish hepatic hemangioma (HH) and hepatocellular carcinoma (HCC). METHODS: In this retrospective study, a total of 110 patients were reviewed, including 72 HCC and 38 HH. We accomplished feature selection with the least absolute shrinkage and operator (LASSO) and built a radiomics signature. Another improved model (radiomics index) was established using forward conditional multivariate logistic regression. Both models were tested in an internal validation group (38 HCC and 21 HH). RESULTS: The radiomic signature we built including 5 radiomic features demonstrated significant differences between the hepatic HH and HCC groups (P < 0.05). The improved model demonstrated a higher net benefit based on only 2 radiomic features. In the validation group, radiomics signature and radiomics index achieved great diagnostic performance with AUC values of 0.716 (95% confidence interval (CI): 0.581, 0.850) and 0.870 (95% CI: 0.782, 0.957), respectively. CONCLUSIONS: Our developed radiomics-based model can successfully distinguish HH and HCC patients, which can help clinical decision-making with lower cost.