Efficiency of MR imaging in the detection of malignant liver lesions

To evaluate the efficacy of preoperative magnetic resonance (MR) imaging in the detection of malignant liver neoplasms.

MR images of 23 patients who had undergone hepatic resection or liver transplantation in the last two years were evaluated retrospectively. All MR imaging studies were performed with a 1.5 T magnet using a phased-array multi-coil. The MR imaging protocol was comprised of fat-suppressed T2-weighted TSE imaging, GRE with and without fat-suppressed T1-weighted imaging, and gadolinium-enhanced multiphasic dynamic GRE imaging. Images were reviewed on a PACS workstation by two independent abdominal radiologists. The image review was conducted on a lesion-by-lesion as well as segment-by-segment basis. MR imaging findings were compared with the results of pathology studies and intraoperative ultrasound examinations. Sensitivity, specificity, and positive predictive value (PPV) of MR imaging and interobserver variation were evaluated.

A total of 59 malignant liver lesions in 23 patients were identified by pathology studies and intraoperative sonographic examinations. Sensitivity and PPV of MR imaging on a lesion-by-lesion analysis were 68-86% and 85-89%, respectively; kappa=0.175 and agreement was 65.8% in these analyses. Sensitivity of MR images for small ( < 1cm) lesions was 13-67% and for large ( > 3 cm) lesions it was 100%. In segment-by-segment analysis, sensitivity and specificity of MR images were 87-95% and 97-98%, respectively; kappa=0.207 and agreement was 76.1%. Sensitivity and PPV of MR imaging in the detection of hepatocellular carcinoma were 46-85% and 55-73%, respectively.

Although MR imaging is generally a highly accurate method for the diagnosis of malignant liver tumors, it has some difficulty in detecting small lesions and hepatocellular carcinoma in cirrhotic livers.