Performance of liver surface nodularity quantification for the diagnosis of portal hypertension in patients with cirrhosis: comparison between MRI with hepatobiliary phase sequences and CT.

Abstract

To assess and compare the performance of liver surface nodularity (LSN) quantification using Gd-BOPTA-enhanced MRI and contrast-enhanced CT for the diagnosis of clinically significant portal hypertension (CSPH) in patients with cirrhosis. This retrospective study included 30 patients with compensated histologically proven cirrhosis who underwent hepatic venous pressure gradient (HVPG), abdominal CT and Gd-BOPTA-MRI within a 60-day interval during pre-surgery workup for hepatocellular carcinoma (HCC) between January 2016 and August 2018. LSN score was derived from CT portal venous phase (PVP), axial T2- and T1-weighted PVP and hepatobiliary phase (HBP). Accuracy for the detection of CSPH was evaluated for each set of images by ROC curve analysis. Intra-observer, inter-observer and inter-method reproducibilities were assessed by the intraclass correlation coefficient (ICC) and coefficient of variation (CV). Thirty patients were analysed (23 men [77%], mean age 60 ± 11years old), including 15 (50%) with CSPH. All CT- and MRI-derived LSN quantifications were correlated to HVPG (CT-PVP: r = 0.63, p = 0.001, AUROC = 0.908 ± 0.06; T1-w-PVP: r = 0.43, p = 0.028, AUROC = 0.876 ± 0.07; T1-w-HBP: r = 0.50, p = 0.012, AUROC = 0.823 ± 0.08; T2-w: r = 0.51, p = 0.007, AUROC = 0.801 ± 0.09). There was no significant difference in AUROC pairwise comparisons (p = 0.12-0.88). Patients with CSPH had higher LSN than those without (CT-PVP: 3.2 ± 0.6 vs 2.4 ± 0.5, p < 0.001; T1-w-PVP: 2.7 ± 0.4 vs 2.2 ± 0.4, p = 0.002; T1-w-HBP: 3.0 ± 0.6 vs 2.3 ± 0.3, p < 0.001; T2-w: 3.0 ± 0.6 vs 2.2 ± 0.3, p = 0.001) and 86%, 82%, 85% and 82% of patients were correctly classified, respectively. Reproducibility of inter-image set comparisons was excellent (ICC = 0.84-0.96 and CV = 8.3-14.2%). The diagnostic performance of MRI-based LSN for detecting CSPH is strong and similar to that of CT-based LSN.