OBJECTIVE This study prospectively evaluates the serial change of FDG uptake and its diagnostic value in malignant versus benign lung lesions in patients with suspected lung cancer. PATIENTS AND METHODS Patients with suspected lung malignancy underwent whole-body FDG PET/CT at 1, 2, and 3 hours after an IV injection of F-FDG. The SUVs of FDG in lung nodules and hilar/mediastinal nodes at each time point were correlated with biopsy/surgical pathologic findings. RESULTS There were a total of 45 malignant lesions and 80 benign lesions from 43 patients with pathologic diagnosis that were included for analysis. The SUVmax had an average of 25.5% increase in all tumor-positive lesions from 1 to 2 hours (vs 1.6% decrease in all tumor-negative lesions, P < 0.0001) and an average of 39.1% increase from 1 to 3 hours (vs 4.5% increase in all tumor-negative lesions, P < 0.0001). The receiver operating characteristic analysis showed that the 2-hour and 3-hour SUVmax had similar area under the curve and outperformed the SUVmax on the 1-hour initial imaging or retention index (RI). The optimal cutoff values to differentiate malignancy from benign lesions were 3.24 for 1-hour SUVmax, 3.67 for 2-hour SUVmax, and 4.21 for 3-hour SUVmax, with 11.6% for 1- to 2-hour RI and 23.9% for 1- to 3-hour RI. The 3-hour delayed SUVmax of 4.21 provided the best overall performance (accuracy of 88.8%). The analysis of the lesion-to-background ratio revealed that delayed imaging improved the image quality significantly, leading to much easier detection of either malignant or benign lesions. CONCLUSIONS Multiple time point FDG PET/CT imaging moderately improves the diagnostic accuracy of lung cancer and significantly improves the image quality.