Rationale and Objectives The aim of this study was to evaluate the accuracy of dual-echo (DE) magnetic resonance imaging (MRI) with and without fat and water separation for the quantification of liver fat content (LFC) in vitro and in patients undergoing liver surgery, with comparison to histopathologic analysis. Materials and Methods MRI was performed on a 1.5-T scanner using a three-dimensional DE MRI sequence with automated reconstruction of in-phase (IP) and out-of-phase (OP) and fat-signal-only and water-signal-only images. LFC was estimated by fat fractions from IP and OP images (MRIIP/OP) and from Dixon-based fat-only and water-only images (MRIDIxON). Seven phantoms containing a titrated mixture of liver and fat from 0% to 50% were examined. Forty-three biopsies in 22 patients undergoing liver surgery were prospectively evaluated by a pathologist by traditional determination of the cell-count fraction and by a computer-based algorithm, the latter serving as the reference standard. Results In vitro, both MRIIP/OP and MRIDIxON were significantly correlated with titrated LFC (r = 0.993, P < .001), with a smaller measurement bias for MRIIP/OP (+2.6%) than for MRIDIxON (+4.5%). In vivo, both MRIIP/OP and MRIDIxON from DE MRI were correlated significantly better with computer-based histologic results (P < .001) and showed significantly smaller measurement bias (4.8% vs 21.1%) compared to histologic cell-count fraction (P < .001). Measurement bias was significantly smaller for MRIIP/OP than for MRIDIxON (P < .001). Conclusions DE MRI allows the accurate quantification of LFC in a surgical population, outperforming traditional histopathologic analysis. DE MRI without fat and water separation shows the highest accuracy and smallest measurement bias for the quantification of LFC.