Liver lesions discovered incidentally on ultrasound: evaluation of reader ability to characterize lesions on MRI without intravenous contrast

Rationale and objectives: Liver lesions incidentally discovered on ultrasound (US) are often further evaluated with magnetic resonance imaging (MRI). The purpose of this study is to evaluate the added effectiveness of contrast-enhanced MRI, compared to noncontrast MRI, to correctly guide management of liver lesions incidentally identified on ultrasound in patients with low pretest probability of malignancy. We conducted the evaluation using a multireader study.

Materials and methods: Liver MRI studies ordered to evaluate incidental liver US lesions were selected for analysis. Patients with no prior history of cancer or chronic liver disease who had 2 years of clinical follow-up (72 patients) were selected to ensure low pretest probability of malignancy and adequate follow-up to establish proof of diagnosis. Fifty of these studies were randomly selected and analyzed by two abdominal radiologists. In the initial interpretation session, only nonenhanced images were reviewed; after 6 weeks, the complete exam including the contrast enhanced images were viewed. Differences in interpretation between the noncontrast and contrast enhanced reading sessions were assessed.

Results: Sixty-nine patients (95.8%) had no abnormalities, benign masses, or hepatic steatosis accounting for the ultrasound findings. One patient was newly diagnosed with hereditary hemochromatosis, and 2 patients were newly diagnosed with metastatic disease. The most likely diagnosis changed for 15 of the lesions described by reader A, and 16 by reader B. The majority of these changes (26/31 lesions, 84%) were from one benign entity to another. Five of 31 (16%) of the changes involved that of a benign entity to a malignant one. Reader A recommended additional contrast enhanced imaging in 13/50 (26%) patients; reader B, 9/50 (18%) patients. After viewing the nonenhanced images, both readers recommended that all patients with eventual malignancies return for contrast administration and/or biopsy. After viewing the contrast-enhanced imaging, both readers accurately diagnosed all malignancies. Patient-based analysis receiver operating characteristic curves demonstrated no significant difference between the enhanced and noncontrast limbs of the study, and no significant difference between readers.

Conclusion: Most of the changes in lesion detection and characterization made after contrast administration were clinically insignificant. In the lesions in which clinically important differences were made, both radiologists recommended additional imaging or biopsy that would have eventually led to the correct diagnosis. None of the patients with malignancy was categorized as completely benign either on noncontrast or contrast-enhanced imaging. This supports the utility of initially performing a noncontrast examination in a population with a low pretest probability of malignancy, especially those in whom gadolinium administration is contraindicated.