Purpose: To examine maximum intensity projections (MIPs) as a method for segmenting lung tumor internal target volumes (ITVs) in positron emission tomography (PET) images using two threshold‐based methods, and to assess various parameters that potentially affect the accuracy of tumor segmentation by this technique. Methods and Materials: An acrylic phantom setup, consisting of a 1.5cm diameter sphere filled with 11 ‐C solution inside a cylindrical bath of 18‐fluorodeoxyglucose solution, was used for a PET list‐mode acquisition. The phantom was attached to a robotic arm that moved according to four patient breathing trajectories via an abdominal breathing bellows device. List‐mode data was gated based on the breathing trajectories, and PET‐MIPs were created from the reconstructed gated images. ITVs were calculated and compared for both un‐gated PET images and PET‐MIPs by using varying thresholds of absolute standardized uptake value (SUV) and percentage of maximum SUV. In addition, a simulation of PET‐MIPs was performed to examine the effects of tumor size, image smoothing, and tumor trajectory on the accuracy of ITV measurement. Results: For any given threshold value, PET‐MIPs tended to produce larger ITVs than un‐gated PET. For optimal percentage thresholds across all trajectories, PET‐MIPs measured an ITV to within 5% accuracy in a larger SUV range with lower initial values than un‐gated PET for the same accuracy. For both PET‐MIP and un‐gated PET, a percentage threshold method produced a smaller deviation from true ITV in similar SUV ranges when compared to an absolute threshold method. The PET‐MIP simulation demonstrated greater accuracy of ITV measurement with increased image smoothing and sphere size. Breathing trajectories exhibiting baseline drift negatively impacted accurate ITV determination. Conclusion: Threshold‐based ITV segmentation is sensitive to lung tumor volume, trajectory, and SUV. PET‐MIPs appear to be a promising tool in the accurate delineation of lung tumor ITV.