The purpose of this study was to assess the accuracy of fluorodeoxyglucose positron emission tomography (FDG PET) in diagnosing infection in a large population of patients and in a variety of clinical circumstances where the performance of conventional imaging modalities has been questioned. We retrospectively analysed 167 FDG PET scans obtained to evaluate 175 anatomical sites for the presence of infection. The major indications for the scans were (1) complicated orthopaedic hardware (n=97), (2) chronic osteomyelitis (n=56), and (3) other (n=14: six fever of unknown origin, three vascular grafts, and five soft tissue). We assessed the overall diagnostic accuracy of FDG PET for each of these indications. In addition, we further analysed this modality's effectiveness by grouping the scans into specific clinical situations. A final diagnosis was made on the basis of surgical pathology and clinical follow-up for a minimum of 6 months. The overall accuracy of FDG PET in evaluating orthopaedic hardware was 96.2% for hip prosthesis, 81% for knee prosthesis, and 100% in 15 patients with other orthopaedic devices. Among the patients in our sample suspected of having chronic osteomyelitis, the accuracy was 91.2%. FDG PET was inaccurate in three cases of fever of unknown origin and accurate in all vascular graft and soft tissue infections. In 49 patients with a clinically apparent soft-tissue infection, FDG PET was able to detect or exclude underlying osteomyelitis with an accuracy of 92.3%. Among the 23 patients who had recent orthopaedic procedures, FDG PET imaging was accurate in 87% of cases. It is concluded that FDG PET is a highly effective imaging modality in the assessment of patients with suspected infection.