Objective: To compare the accuracy of four different reconstruction techniques of Computed Tomography [CT] images for localization of moving targets.

Materials and Methods: A respiratory motion phantom was simulated for tumor motion with the respiratory rate of 15 breaths per minute, breathing amplitude 1.0 cm in craniocaudal [CC]
direction. In treatment room, the four different reconstruction of CT image: free breath [FB], maximum intensity projection [MIP], mid-position [MidP] and average intensity projection [AIP], were compared with the FB kV-Cone-Beam CT obtained from the on-board imaging system of TrueBeam linear accelerator. The accuracy of target localization was calculated in terms of mean deviation and matching index [MI]. Additionally, image quality and tumor shape were evaluated using visual grading analysis of 10 observers.

Results: The mean deviation of target position was significant in CC direction with the maximum value of 3.11+0.65 mm in the MidP images. The high MI values were 0.86 and 0.87 in AIP and FB images, respectively. Moreover, AIP and FB had high grading scores for both image quality and tumor shape.

Conclusion: Both FB and AIP image datasets can be used as the reference image for localizing the lung tumor with less than 2 cm movement in CC direction.

Keywords: 4DCT image acquisition, kV-CBCT, Moving target, Target localization