Diagnostic per-lesion performance of a simulated gadoxetate disodium-enhanced abbreviated MRI protocol for hepatocellular carcinoma screening

Abstract

To evaluate the diagnostic per-lesion performance of a simulated gadoxetate disodium-enhanced abbreviated MRI (AMRI) in cirrhotic and chronic hepatitis B (CHB) patients for hepatocellular carcinoma (HCC) screening. Seventy-nine consecutive patients at risk for HCC due to cirrhosis and/or CHB were included in this retrospective study. For each patient, the first gadoxetate disodium-enhanced MRI between 2008 through 2014 was analysed. Two independent readers read an anonymised abbreviated image set comprising axial T1-weighted (W) images with fat saturation in the hepatobiliary phase, 20 minutes or more after gadoxetate injection, and axial T2W single-shot fast spin echo images. Each observation >10mm was scored as negative or suspicious for HCC. Inter-reader agreement was assessed. A composite reference standard was used to determine the per-lesion diagnostic performance for each reader. Inter-reader agreement was substantial (κ=0.75). The final reference standard showed 27 HCCs in 13 patients (median 21mm, range 11-100mm). The two readers each correctly scored 23 as suspicious for HCC (sensitivity=85.2%), scored a total of 27 and 32 observations as suspicious for HCC (positive predictive value [PPV]=85.2% and 71.9%), and scored 83 and 78 observations or complete examinations as negative for HCC (negative predictive value [NPV]=95.2% and 94.9%). The AMRI protocol provides higher per-lesion sensitivity and NPV than reported values for ultrasound, the current recommended technique for screening, and similar per-lesion sensitivity and PPV to reported values for complete dynamic contrast-enhanced MRI.