Purpose: To assess prospectively the accuracy of 2-[18F]-2-deoxy-D- glucose positron emission tomography (FDG-PET) far predicting regional node involvement in cutaneous malignant melanoma (CMM). Patients and Methods: Twenty-three patients with CMM (primary lesions > 1.5 mm thick) scheduled far lymph node dissection (LND) were preoperatively studied with FDG-PET. Thirteen patients underwent therapeutic LND of 14 node basins, while nine patients had elective LND of 10 node basins. Medical problems precluded surgery in one patient. Two observers unaware of the clinical node status - apart from whether a recent surgical scar was present - read attenuation- corrected reconstructed transverse images acquired between 50 and 60 minutes after injection. Intensity of FDG uptake was scored as 0 to 3 + on a semiquantitative four-point scale: 0, no uptake; 1 +, faint; 2 +, moderate; and 3 +, intense uptake. A node group was considered positive on FDG-PET if it contained at least one focus of FDG uptake of ≤ 2+ intensity. Histopathologic examination of the 24 dissected node groups served as a reference. Results: Considering regional node basins, PET imaging demonstrated 11 true-positive (TP), 10 true-negative (TN), two false-negative (FN), and one false-positive (FP) result, for an overall accuracy of 88%. Histopathologic from one FN case showed seven malignant cells in a marginal node sinus. The FP was due to reactive changes postbiopsy. In one patient, clinically involved lymph nodes were correctly categorized TN by PET. At least four additional 2 + foci seen outside the dissected regions on PET may represent metastases and are being monitored. Conclusion: FDG-PET accurately predicted regional node status in 88% of CMM cases. The failure to detect micromatastatic disease may be due to the limitations of the imaging equipment and technique used here.