LAUSR

PURPOSE To identify the value of discrepancies in the central scar (CS)-like structure between dynamic CT and gadoxetate disodium-enhanced MRI for differentiating FNH from HCA. METHODS This retrospective study included 113 patients with pathologically-diagnosed FNH (n = 80) or HCA (n = 37). CS-like structures were evaluated on arterial phase (AP) CT and hepatobiliary phase (HBP) MRI. Presence of the CS-like structure, its discrepancy in visibility or size between AP CT and HBP MRI and between AP and HBP MRI, and features of non-scarred tumor portion were evaluated by two radiologists. Inter-observer agreement was evaluated by intraclass correlation coefficients (ICCs) and weighted kappa. Univariable and multivariable logistic regression and ROC analysis were performed to explore features differentiating FNH from HCA. RESULTS Inter-observer agreement was moderate-to-excellent (ICCs≥0.74, kappa≥0.65). On univariable analysis, presence of CS-like structures (P < 0.001), discrepancy of the CS-like structures between AP CT and HBP MRI (73.8 % in FNH; 16.2 % in HCA, P < 0.001) and between AP and HBP MRI (70.0 % in FNH; 16.2 % in HCA, P < 0.001), and the features of non-scarred tumor portion (P ≤ 0.011) were significantly different between FNH and HCA. On multivariable analysis, the discrepancy of CS-like structures between AP CT and HBP MRI, and the absence of low SI of the non-scarred tumor portion on HBP MRI, were suggestive of FNH (P = 0.036 and P < 0.001, respectively; area under the ROC curve, 0.96 [95 % CI, 0.93-0.99]). CONCLUSION Evaluation of discrepancy in the visibility or size of CS-like structures between dynamic CT and gadoxetate disodium-enhanced MRI may facilitate the differentiation of FNH from HCA.