Sequence and Observer Variability in Gadoxectic Acid-Enhanced MRI Lesion Measurements in Hepatocellular Carcinoma

Abstract

The purpose of the study was to investigate interobserver and intersequence variability in measuring hepatocellular carcinoma on magnetic resonance imaging (MRI). Twenty treatment-naïve lesions on Gadoxetic Acid enhanced MRI scans from 20 patients were retrospectively measured by six reviewers with different levels of experience, twice, six weeks apart, on eight different MRI sequences, in randomized order. The sequences include arterial, hepatobiliary, transitional, portal venous, T2, and diffusion weighted images. The single longest diameter (SLD) and longest diameter perpendicular to the longest overall diameter were measured on axial images and products of diameters calculated in accordance to response evaluation criteria in solid tumors v1.1 and World Health Organization response criteria respectively. Lesion-wise intraclass correlation coefficients were used to estimate measurement agreement. All intraclass correlation coefficients were greater than 0.95. No substantive differences between SLD and products of diameters metrics. Means (∼2.8 mm, SLD) and standard deviations (∼2 mm, SLD) were similar across sequences and observers. Similarly, pairwise comparison between observers grouped by experience showed statistically significant differences, but the effect size was minor (∼2 mm). Arterial and HPB-weighted images had similar mean dimensions (2.76 cm) while the smallest mean was in the transitional phase (2.62 cm). A lesion was not measured on 140 occasions (7%), mostly in ADC. There is high interobserver and intersequence reliability despite small differences between observers based on experience level. Our results suggest that accurate measurements can be made on arterial phase despite the possibility of indistinct margins. Lesions, however, are more likely to be missed on diffusion-related sequences.