Intraindividual comparison of MRI-derived liver surface nodularity score at 1.5 T and 3 T

To compare the MRI-derived liver surface nodularity (LSN) scores acquired on both 1.5 T and 3 T.

Forty chronic liver disease patients who underwent gadoxetic acid-enhanced MRI at both 1.5 and 3 T were included. Axial hepatobiliary phase images with the same voxel size were used to calculate the LSN score in both liver lobes with a quantitative software. Rank correlation, Wilcoxon test, and Bland-Altman limits of agreement were used for statistical analysis.

There was a weak correlation between the right and left liver lobe on 1.5 T (r_s = 0.331, p = 0.037) and 3 T (r_s = 0.381, p = 0.015). The correlation between 1.5 T and 3 T on both liver lobes showed a very strong correlation (right, r_s = 0.927, p < 0.001; left, r_s = 0.845, p < 0.001). LSN scores differed significantly between both lobes on 1.5 T (median, 1.201 vs. 0.674, right vs. left) and 3 T (1.076 vs. 0.592) (all p < 0.001). LSN scores differed significantly between 1.5 T and 3 T on both lobes (all p < 0.001). The Bland-Altman plot comparing 1.5 T and 3 T on right and left liver lobes showed a systemic bias of 0.08 and 0.07, respectively.

LSN scores differed significantly on 1.5 T vs. 3 T and right vs. left liver lobe. Caution should be made when comparing LSN scores derived from different field strengths or the hepatic lobe. Interplatform, interlobar reproducibility should be resolved to use LSN scores, which is relatively easy to perform without additional hardware or images.