Effects of subject motion and acquisition time on cone‐beam CT‐guided adaptive therapy for novel 6 second fast scanning protocols

Abstract Background Recent advancements in cone‐beam CT (CBCT) technology have significantly improved image quality leading to increased adoption of CBCT‐guided online adaptive radiotherapy (ART). However, the impact of subject motion on these novel fast acquisition approaches on ART is not fully understood. Purpose This study aims to assess the impact of subject motion on tasks critical to CBCT‐guided ART, comparing novel fast CBCT acquisition protocols with different spatiotemporal sampling patterns to diagnostic fan beam CT (FBCT). Methods Two phantoms were imaged to characterize motion, an anthropomorphic abdomen and a cylindrical phantom with material inserts in the presence of translational periodic and non‐periodic motion using two fast CBCT systems (Halcyon, Varian Medical Systems, and HyperSight, Varian Medical Systems) with acquisition times of 16.6 and 6 s (denoted CB‐16.6 and CB‐6), as well as a fan‐beam CT (FB‐0.5) system (Brilliance BigBore, Philips). For the anthropomorphic phantom, geometric distortion was assessed by deformably registering each image to a FB‐0.5 scan of the stationary phantom and analyzing the resultant motion vector field (MVF). Auto‐contouring was performed on all images with a validated deep‐learning method and evaluated using Dice similarity coefficient (DSC). Dose delivery accuracy of a synthetic CT‐based ART workflow was evaluated through a simulated adaptive planning study using a CBCT‐based synthetic CT workflow, comparing Paddick conformity indices (CI) of simulated delivered dose distributions. For the cylindrical phantom, 3D motion patterns more representative of complex patient motion were assessed by measuring and calculating the edge spread function (ESF) and circularity error of the inserts. Results CB‐16.6 and CB‐6 exhibited reduced geometric distortion in the presence of periodic and non‐periodic motion compared to FB‐0.5. Under periodic motion, auto‐contouring using FB‐0.5 provided the highest mean(range) DSC of 0.78(0.39–0.99), followed by CB‐6 with 0.67(0.28–0.86), and CB‐16.6 with 0.62(0.19–0.86). Under non‐periodic motion, auto‐contouring using FB‐0.5 also provided the highest mean(range) DSC of 0.72(0.46–0.80), followed by CB‐6 with 0.67(0.43–0.80), and CB‐16.6 with 0.64(0.31–0.84). For periodic motion, the FB‐0.5 system shows the largest differences in CI, followed by the CB‐6 and CB‐16.6 with mean values of ‐0.249, ‐0.007, and ‐0.002, respectively. For non‐periodic motion, again, the FB‐0.5 exhibited the greatest change, followed by the CB‐16.6 and CB‐6 with mean differences in CI values of ‐0.122, ‐0.040, and ‐0.002, respectively. For the cylindrical phantom, the CB‐6 demonstrated reduced blur and streak artifacts compared to CB‐16.6 with consistently improved edge sharpness for both low and high contrast edges. Conclusions CB‐16.6 and CB‐6 provided high‐quality images for ART. Compared to FB‐0.5, fast CBCT exhibited increased blurring and streaking but significantly less geometric distortion with increasing motion, leading to reduced dose delivery error in a simulated ART workflow. The complex 3D motion phantom study supported trends observed in the 1D translational phantom study. Faster CBCT (6 vs. 16.6 s) reduced blur, improving auto‐contouring performance for both periodic and non‐periodic motions, but the interplay between the motion period and 6 s image acquisition must be carefully managed.