To investigate the feasibility of ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) positron emission tomography with computed tomography (PET-CT) radiomics in the differential diagnosis of moderately- or poorly-differentiated hepatocellular carcinoma (HCC) from intrahepatic cholangiocarcinoma (ICC).

In this prospective study, 36 patients with primary liver cancer receiving ¹⁸F-FDG PET-CT in the Third Affiliated Hospital of Sun Yat-sen University from June 2015 to April 2018 were recruited. The informed consents of all patients were obtained and the local ethical committee approval was received. Among them, 31 patients were male and 5 female, aged 21-74 years with a median age of 52 years. 26 patients were diagnosed with moderately- or poorly-differentiated HCC and 10 of ICC. The volume of region of interests on the ¹⁸F-FDG PET-CT images was delineated by using 3D Slicer software. The radiomic features of each lesion were extracted. A total of 105 radiomic features were screened and optimized by LASSO regression model to construct a diagnostic model with radiomics signatures. The efficacy of this model in differential diagnosis from moderately- or poorly-differentiated HCC and ICC was evaluated by the receiver operating characteristic (ROC) curve.

In total, 105 PET-CT radiomic features were obtained for each lesion. After screening and optimization, 2 PET-CT radiomic features, Sphericity and ZoneVariance, were sorted and included into the Logistic regression model. The diagnostic model with radiomics signatures was logit (P) =-8.984+10.506×Sphericity+61.341×ZoneVariance. The area under ROC curve of this model was 0.923, 95%CI: 0.84-1.00, with the sensitivity 1.00 and specificity 0.73.

¹⁸F-FDG PET-CT radiomics can be adopted to differentiate moderately- or poorly-differentiated HCC from ICC with a high sensitivity, specificity and diagnostic value.