Comparison of MRI pulse sequences for prediction of size of hepatocellular carcinoma at explant evaluation.

Abstract

The purpose of this study was to retrospectively compare the size of hepatocellular carcinoma (HCC) on images obtained using different MRI pulse sequences with the tumor size determined at pathologic evaluation of liver explant specimens. Ninety-two patients with HCC who underwent contrast-enhanced liver MRI within 90 days before liver transplant were included. A single pathologist measured the dominant HCC in each case. In different sessions, two abdominal radiologists (readers 1 and 2) aware only of the location of the dominant HCC independently measured lesion size on images obtained using the following sequences: T2-weighted imaging; b-500 diffusion-weighted imaging; and arterial, portal venous, and equilibrium phases of contrast enhancement. Size measurements on MR images were compared with explant measurements by use of Pearson correlation coefficients, paired t tests, and Bland-Altman plots. Correlation with pathologic findings was highest for reader 1 for portal venous (r = 0.890) and equilibrium (r = 0.828) phase images and for reader 2 for arterial, portal venous, and equilibrium phase images (r = 0.842-0.860). Absolute error relative to pathologic size was lowest for reader 1 using portal venous (4.3 mm) and for reader 2 using portal venous and arterial phase images (both 4.7 mm). Systematic error for both readers was lowest with portal venous and equilibrium phase images (reader 1, systematic under-measurement of 0.5 mm in both sequences; reader 2, systematic over-measurement of 0.1 mm with portal venous phase images and systematic under-measurement of 1.1 mm with equilibrium phase images). Sequences in which reader 1 made systematic over-measurements were diffusion-weighted images, arterial phase images, and T2-weighted images (by 3.5, 2.9, and 1.6 mm). Reader 2 made systematic over-measurements using arterial phase and T2-weighted images (by 1.5 and 0.4 mm). The data suggest the arterial phase may be suboptimal for measuring HCC at MRI. Portal venous phase acquisition warrants further investigation as a potential standard approach for such measurements.