This study was performed to investigate the detection performance of trapped air in dynamic chest radiography using 4D extended cardiac-torso (XCAT) phantom with a user-defined ground truth. An XCAT phantom of an adult male (50th percentile in height and weight) with a normal heart rate, slow-forced breathing, and diaphragm motion was generated. An air sphere was inserted into the right lung to simulate emphysema. An X-ray simulator was used to create sequential chest radiographs of the XCAT phantom over a whole respiratory cycle covering a period of 10 seconds. Respiratory changes in pixel value were measured in each grid-like region translating during respiration, and differences from a fully exhaled image were then depicted as color-mapping images, representing higher X-ray translucency (increased air) as higher color intensities. The detection performance was investigated using various sizes of air spheres, for each lung field and behind the diaphragm. In the results, respiratory changes in pixel value were decreased as the size of air sphere increased, depending on the lung fields. In color-mapping images, air spheres were depicted as color defects, however, those behind the diaphragm were not detectable. Smaller size sampling depicted the air spheres as island color defects, while larger ones yielded a limited signal. We confirmed that dynamic chest radiography was able to detect trapped air as regionally-reduced changes in pixel value during respiration. The reduction rate could be defined as a function of residual normal tissue in front and behind air spheres.