Added Value of CT Arterial Subtraction Images in Liver Imaging Reporting and Data System Treatment Response Categorization for Transcatheter Arterial Chemoembolization-Treated Hepatocellular Carcinoma

Objectives The aim of this study was to assess the benefit of adding arterial subtraction images from computed tomography (CT) to the Liver Imaging Reporting and Data System (LI-RADS) v2018 treatment response (LR-TR) categorization in patients treated with transcatheter arterial chemoembolization (TACE) for hepatocellular carcinoma (HCC). Materials and Methods This retrospective study included 115 patients with 151 HCCs treated by TACE using an emulsion of doxorubicin and iodized oil who underwent multiphasic CT protocol that additionally generated arterial subtraction images based on nonrigid anatomic correction algorithm. Of 151 HCCs, 67 (44.4%) were viable and 84 (55.6%) were nonviable. Two independent readers assessed the per-lesion LR-TR categories in set 1 of multiphasic CT images alone and set 2 including both set 1 and CT arterial subtraction images, besides diagnostic confidence, and the quality of subtraction images. The sensitivity and specificity of LR-TR viable category between the sets were compared using the generalized estimating equation. Interobserver agreements of LR-TR categorization in each set and the quality of subtraction images were assessed by Cohen kappa. Results The quality of subtraction images was mostly good to perfect (98.7%) with good interobserver agreement (kappa = 0.71), and none were nondiagnostic. For detecting viable HCC, LR-TR viable category showed sensitivity of 53.7% to 56.7% and specificity of 96.4% to 98.8% in set 1. In comparison, set 2 showed significantly higher sensitivity of 88.1% to 89.6% (P < 0.002) and equivalent specificity of 94% to 95.2% (P > 0.13) for the same category. In sets 1 and 2, 31.3% to 34.3% and 9% to 10.4% of viable HCC were miscategorized as LR-TR nonviable, respectively. LR-TR equivocal category was less assigned in set 2 (1.3%) than in set 1 (6.6%-7.9%). Set 2 showed slightly higher level of confidence for LR-TR categorization compared with set 1 (3.4 +/- 0.8 vs 3.8 +/- 0.5). Interobserver agreement was excellent in both sets (kappa = 0.85 in set 1 and 0.97 in set 2). Conclusions The LR-TR viable category is highly specific but inadequately sensitive for detecting viable tumor in TACE-treated HCC on conventional multiphasic CT. Adding arterial subtraction images to the conventional CT images significantly increases sensitivity without compromising the specificity and improves the diagnostic confidence of LR-TR viable category.