EVALUATION OF ACCURACY AND INTRINSIC UNCERTAINTY OF AUTOMATED IMAGE REGISTRATION FOR A 6D KV-CBCT IMAGE GUIDANCE SYSTEM: A CONCURRENT ANALYSIS WITH A MACHINE PERFORMANCE CHECK

Abstract

Abstract The accuracy and uncertainty of the automated image registration (AIR) algorithm in a six-dimensional (6D) kilovoltage cone-beam computed tomography (kV-CBCT) image-guided radiation therapy (IGRT) system were evaluated with a concurrent analysis of machine performance check (MPC). The MPC was performed before (MPCpre) and after (MPCpost) each accuracy and intrinsic uncertainty measurement. The accuracy was evaluated for 25 sets of the known shifts applied to the Catphan-504 phantom through a 6D robotic couch in the head, thorax, and pelvis CBCT acquisition modes. The uncertainty was evaluated for the intensity range, soft tissue, and bone matching filters in the head, thorax and pelvis CBCT acquisition modes. The mean ΔMPC (MPCpost–MPCpre) for all test parameters was within 0.02 ± 0.08 mm and 0.00 ± 0.02°. The overall average accuracy in AIR of 6D kV-CBCT IGRT in all translational and rotational axes was within 0.05 ± 0.76 mm and 0.02 ± 0.07°, respectively, for all CBCT modes. The overall mean population (Mpop), systematic (Σ) and random (σ) errors were within 0.47, 0.53 and 0.24 mm and within 0.03, 0.08 and 0.07° in translational and rotational axes, respectively, for all matching filters in all CBCT modes. The accuracy and intrinsic uncertainty in the AIR of the 6D kV-CBCT IGRT were within acceptable limits for clinical use.