PURPOSE: To evaluate the efficacy of magnetic resonance (MR) imaging for the diagnosis of osteomyelitis in the diabetic foot by using anatomic and histologic studies of the resected tissue as a standard of reference. MATERIALS AND METHODS: Thirteen diabetic patients with high clinical suspicion of osteomyelitis underwent a total of 15 MR examinations before surgery. Correlation was made between MR findings and the histologic features of the resected tissue, which included 57 bones. RESULTS: Maximum signal intensity on the T2-weighted or short inversion time inversion-recovery images of the bones was due to osteomyelitis (prospective sensitivity, 90%; specificity, 71%). Eighteen bones with increased signal intensity showed only edema of the marrow. The range of signal intensity in edema overlapped that in osteomyelitis but was lower. The use of gadopentetate dimeglumine improved delineation of soft-tissue inflammatory masses, but this contrast material was not useful in distinguishing osteomyelitis from edema. CONCLUSION: Marrow edema cannot be reliably distinguished from osteomyelitis with MR imaging. Gadopentetate dimeglumine is of limited use. Some previously published false-positive reports of osteomyelitis were most likely due to edema of the marrow. MR imaging is useful in planning surgery of the infected diabetic foot, as it enables reliable distinction between normal and abnormal areas.