Detection of Portal Perfusion Abnormalities

Abstract

To estimate the accuracy, sensitivity, and specificity of 3 ferucarbotran-enhanced magnetic resonance (MR) imaging sequences prospectively for the detection of nontumoral portal perfusion abnormalities. Thirty-nine noncirrhotic patients with liver metastases underwent computed tomography during arterial portography (CTAP) and MR imaging comprising T1-weighted gradient recalled echo (GRE), T2-weighted fast spin echo (FSE), and T2*-weighted GRE sequences with and without ferucarbotran. Magnetic resonance images were reviewed by 4 blinded observers for rating based on the confidence scale. The accuracy, sensitivity, and specificity for each sequence were measured by receiver operating characteristic analysis. Contrast-to-noise ratio (CNR) and relative signal-to-noise ratio changes were statistically compared. Thirty-nine nontumoral perfusion defects were observed in 22 patients by CTAP. Receiver operating characteristic analysis showed the accuracy was higher for T2*-weighted GRE (0.884) than for T1-weighted GRE (0.572) and T2-weighted FSE (0.597). T2*-weighted imaging achieved the highest sensitivity (81.4%) and the lowest specificity (86.6%). Postenhanced T2*-weighted imaging achieved the highest CNR (19.3 +/- 9.2). T2*-weighted imaging was the most accurate and sensitive method for detecting portal perfusion abnormalities compared with T1- or T2-weighted imaging, whereas T1- or T2-weighted imaging is superior in specificity to T2*-weighted imaging during ferucarbotran-enhanced MR imaging.