The inherent noise in positron emission tomography (PET) leads to the instability of quantitative indicators, which may affect the diagnostic accuracy for differentiating malignant and benign lesions in the management of lung cancer. In this paper, the reliability of retention index (RI) is systematically investigated by using computer simulation for the dual-time-point imaging protocol. The area under the receiver operating characteristic (ROC) curve is used to evaluate the optimal protocol. Results demonstrate that the reliability of RI is affected by several factors including noise level, lesion type, and imaging schedule. The RIs with small absolute values suffer from worse reliability than those larger ones. The results of ROC curves show that over delayed second scan cannot help to improve the diagnostic performance further, while an early first scan is expected. The method of optimization based on ROC analysis can be easily extended to comprise as many lesions as possible.