Purpose. Different types of malignant hepatic lesions are difficult to diagnosis through imaging modalities depending on visual assessment. To study the potential of metabolic and radiomics parameters of 18F-fluorodeoxyglucose–positron emission tomography/computed tomography (FDG-PET/CT) in the classification of malignant hepatic lesions including hepatocellular carcinoma (HCC) and liver metastasis (LM) and cholangiocarcinoma (CC).

Materials and methods. Thirty-three primary tumors from 33 patients with HCC, 34 lesions of LM from 15 patients, and13 lesions of cholangiocarcinoma from 9 patients were included in this study. The lesions have been segmented from the PET images after the CT correction, then 7 metabolic features, 2 shape indices feature, and 35 radiomics features were extracted from images through semiautomatic method in LIFEx package. Kruskal-Wallis test was used to determine the significant different features. Finally, the area under the curve (AUC) was calculated to find the features with high sensitivity to characterize the tumor types.

Results. According to Kruskal-Wallistest,4 metabolic parameters, 1 shape indices features, and 22 radiomics features were significantly different among tumor types. The most differentiating radiomics and metabolic features were GLRLM-GLNU and shape compactiy. Most of the features show high sensitivity to distinguish (HCC/LM) and (HCC/CC) according to the area under the curve (AUC).

Conclusions. Radiomics features in addition to metabolic parameters and shape indices from FDG-PET/CT studies may be of value in characterizing different malignant hepatic tumors.