PURPOSE: To compare prospectively the diagnostic accuracy of T2-weighted conventional spin-echo (SE) and fast SE magnetic resonance (MR) imaging for differentiation of benign (hemangiomas or cysts) from malignant (metastases or hepatocellular carcinoma) liver lesions. MATERIALS AND METHODS: Fifty-three patients with 55 confirmed liver lesions (20 hemangiomas, eight cysts, 22 metastases, four hepatocellular carcinomas, one malignant fibrous histiocytoma) underwent T2-weighted conventional SE (repetition time msec/echo time [TE] msec = 3,100/80, 160) and fast SE (5,000-5,500/99-104 [effective TE]) MR imaging at 1.5 T. The images were evaluated quantitatively (lesion-liver signal intensity ratio and lesion T2 index [SIlesionTE80/SIlesionTE160, where SI = signal intensity] and qualitatively (blinded reading) by using receiver operating characteristic analysis. RESULTS: Quantitatively, the T2 index (A(z) = .988) was more accurate than the signal intensity ratio at conventional SE (A(z) = .920) and fast SE (A(z) = .910) imaging (P < .05). Qualitatively, there was no statistically significant difference between the blinded reading of conventional SE (A(z) = .988) and fast SE (A(z) = .986) images in lesion characterization (benign vs malignant). There was a trend, however, toward superiority of conventional SE imaging for evaluation of ''hypervascular'' metastases (n = 9). CONCLUSION: With the specific parameters used, conventional SE and fast SE images provide equal accuracy for characterization of focal liver lesions. In patients with hypervascular metastases, double-echo conventional SE images may be preferable.