Liver imaging reporting and data system (LI-RADS) v2018: Reliability and agreement for assessing hepatocellular carcinoma locoregional treatment response

•Liver imaging reporting and data system (LI-RADS) treatment response algorithm (LR-TR) v2018 has excellent diagnostic performance and reliability for the diagnosis of residual hepatocellular carcinoma after locoregional treatment.•LR-TR enhancement characteristics and final category show interobserver agreement of 0.815 and 0.795, respectively.•The LR-TR algorithm v2018 is recommended for reporting hepatocellular carcinoma treatment response after locoregional therapy. The purpose of this study was to determine the reliability and interobserver agreement of the liver imaging reporting and data system (LI-RADS) treatment response algorithm (LR-TR) v2018 using dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and the added value of diffusion-weighted imaging (DWI). A total of 54 patients who underwent DCE-MRI and DWI after locoregional treatment of 81 hepatocellular carcinoma (HCC) lesions from September 2020 to July 2021 were included. There were 47 men and 7 women, with a mean age of 63.9 ± 9.2 (SD) years (age range: 23–77 years). Locoregional treatments included transarterial chemoembolization (TACE) (53/81; 65.4%), radiofrequency ablation (RFA) (25/81; 30.9%) and microwave ablation (MWA) (3/81; 3.7%). Two independent radiologists retrospectively evaluated DCE-MRI examinations obtained after locoregional treatment using LR-TR, and then three months later both radiologists reevaluated DCE-MRI examinations with DWI. Interobserver agreement was assessed using intraclass correlation coefficient (ICC) and Kappa test. Diagnostic performances were evaluated in term of sensitivity, specificity, and area under ROC curve (AUC) using a composite standard of reference that included results of histopathological examinations and follow-up findings. Using DCE-MRI alone, observer 1 had 83.9% sensitivity (26/31; 95% confidence interval [CI]: 66–95%), 88% specificity (44/50; 95% CI: 76–95%) and 86.4% accuracy (70/81; 95%CI: 77–93%), and observer 2 had 71% sensitivity (22/31; 95% CI: 52–86%), 92% specificity (46/50; 95% CI: 81–98%) and 83.9% accuracy (68/81; 95% CI: 74–91%). For the diagnosis of viable tumors using DCE-MRI with DWI, observer 1 and observer 2 had 87.1% (27/31; 95% CI: 70–96%) and 74.2% (23/31; 95% CI: 55–88%) sensitivity, respectively. The diagnostic performance of DCE-MRI with DWI yielded an AUC (0.875; 95% CI: 0.789–0.962) not different from that of DCE-MRI without DWI (0.859; 95% CI: 0.768–0.951) (P = 0.317). Interobserver agreement for arte