Optimizing robot-CT trajectories for data completeness

Abstract Industrial cone-beam computed tomography (CBCT) is often constrained by a limited field of view and is susceptible to severe artifacts, including photon starvation and metal streaking. To mitigate these issues, we propose a scan-trajectory optimization algorithm that identifies source positions with high reconstructability, as measured by the Tuy score, while excluding those likely to induce metal artifacts through projection thresholding. The effectiveness of this approach was validated using numerical simulations and experimental measurements obtained from a robotic arm-based CBCT system. Compared with the conventional circular trajectory, the proposed method demonstrated superior image quality, as evidenced by both visual assessment and quantitative metrics, including structural similarity and Shannon entropy. Additionally, we discuss potential extensions that incorporate prior information, such as object boundaries, to further refine the trajectory design.