Diagnostic performance of a minimized protocol of non-contrast MRI for hepatocellular carcinoma surveillance

Abstract

Purpose To evaluate the per-patient diagnostic performance of a minimized non-contrast MRI protocol for hepatocellular carcinoma (HCC) surveillance in cirrhotic liver, as well as factors affecting diagnostic sensitivity. Methods A total of 226 patients who underwent MRI for HCC surveillance over an 8 year period were included in this retrospective study. Set1 consisted of diffusion-weighted imaging and respiratory-triggered, fast-spin echo T2-weighted imaging with fat suppression. Set2 included T1-weighted in/opposed-phase images added to the images from Set1. Image sets were scored as positive or negative for HCC according to predetermined criteria by two readers independently. The diagnostic performance of the two sets in conjunction with alpha-fetoprotein (AFP) was assessed and compared using the McNemar test. Logistic regression was used to determine factors that affected sensitivity. Results The sensitivity, specificity, and accuracy of Set1 of readers 1 and 2 were 84.4%/87.3%, 86.8%/86.8%, and 85.0%/87.2%, respectively; and those for Set2 were 87.3%/89.6%, 81.1%/79.2%, and 85.8/87.2%, respectively. The sensitivities of the sets were not significantly different (p = 0.063). Sensitivities of both sets in conjunction with AFP were higher than those of MRI alone without statistical significance (87.3%/89.6%, p = 0.063/> 0.99; 89.6%/89.6%, p = 0.125/> 0.99). In very early-stage HCC, the sensitivities of Sets1 and 2 were 73.1%/76.9% and 76.9%/82.7%, respectively. Perihepatic ascites and size less than 2 cm were associated with sensitivity (p < 0.05). Conclusions A minimized non-contrast MRI protocol consisting of Fat-sat T2WI and DWI is highly sensitive and may be a viable method for HCC surveillance of the cirrhotic liver. The inclusion of T1-weighted in/opposed-phase and AFP may increase this sensitivity.