Reproducibility and accuracy of non-contrast abbreviated magnetic resonance imaging of the liver in surveillance for early recurrence for hepatocellular carcinoma in a Western population: a multi-reader study

Purpose : Abbreviated MRI (AMRI) protocols may represent an alternative to conventional MRI (CMRI) for surveillance of hepatocellular carcinoma (HCC). We aimed to compare the inter-reader agreement and sensitivity of AMRI versus CMRI for HCC recurrence <2 years after curative ablation in at risk-population. Methods : Eight radiologists (4 with <5 years’ and 4 with ≥5 years’ experience) from three institutions independently reviewed 143 CMRI and AMRI image sets from 75 consecutive cirrhotic patients (84% HCV and/or alcohol-related) undergoing secondary screening after HCC ablation with ≥1 month between readings. We calculated the intra and inter-reader agreement by means Gwet’s AC1 for detection of local recurrence at the ablation site, new intrahepatic, and any type of recurrence (either local and/or new intrahepatic) with CMRI and AMRI. Reference diagnoses of recurrent HCC were based on histological or imaging-based criteria. Results : Early HCC recurrence was detected in 37 patients (49.3%). AC1 agreement was similarly high for AMRI and CMRI for local recurrence [0.87 (0.83‒0.90) vs. 0.87 (0.83‒0.92)], but higher for AMRI than for CMRI for new intrahepatic [0.85 (0.81‒0.9) vs. 0.6 (0.52‒0.67)] and any type [0.73 (0.67‒0.78) vs. 0.56 (0.49‒0.64)] recurrences. Sensitivity for detecting any type of HCC recurrence was higher for CMRI [0.83 (0.78‒0.87) vs. 0.39 (0.33‒0.45) for AMRI, p<0.0001]. Conclusions: For early detection of HCC recurrence in a cohort with predominantly HCV and/or alcohol-related cirrhosis, non-contrast AMRI yielded better interobserver agreement but lower sensitivity than CMRI.