We aimed to optimize internal margin (IM) determination for respiratory-gated radiotherapy using end-expiratory phase assessments using a motion phantom. Four-dimensional computed tomography (4D CT) data were acquired using a GE LightSpeed RT CT scanner, a respiratory-gating system, and a motion phantom designed to move sinusoidally. To analyze the accuracy of 4D CT temporal resolution, a 25.4 mm diameter sphere was inserted into the motion phantom, and we measured the differences in sphere diameters between static and end-exhalation phase images. In addition, the IM obtained from the maximum intensity projection within the gating window (MIP(GW)) image was compared to theoretical value. Cranial-caudal motion displacement ranged from 5.0 to 30.0 mm, and the respiratory period ranged from 2.0 to 6.0 sec. Differences in sphere diameters between static and end-exhalation phase images ranged from 0.37 to 4.6 mm, with 5.0-mm and 30 mm target displacements, respectively. Differences between the IM obtained from the MIP(GW) and the theoretical values ranged from 1.12 to 6.23 mm with 5.0mm and 30 mm target displacements, respectively. These differences increased in proportion to the target velocity due to a motion artifact generated during tube rotation. In this study, the IMs obtained using the MIPGW image were overestimated in all cases. We therefore propose that the internal target volume (ITV) for respiratory-gated radiotherapy should be determined by adding the calculated value to the end-exhalation phase image. We also demonstrate a methodology for subtracting motion artifacts from the ITV using a motion phantom.