SU‐C‐BRA‐02: Evaluation of 2D DIR from CBCT to 4DCT Projections as a Tool for IGART

Abstract

Purpose: To demonstrate 2D parallelized DIR as a method to correct for anatomical changes of patient anatomy and respiratory motion. CBCT to simulated 4D‐CT projections using DIR can be used in real time for respiratory phase determination, patient setup or as a retrospective analysis to track inter‐ and intra‐fractional deformations. Methods: Ten phases of simulated 4DCT respiratory projections were individually registered to pulmonary CBCT projections using a modified 2D‐DIR demons algorithm implemented on GPUs. An automated global image‐preprocessing algorithm of the simulated projection was implemented to account for the relative local inter‐modality intensity mismatch. The deformation vector field intermodality congruence was inspected using bony structures, diaphragm position, and tumor position. A phase prediction matrix was constructed by choosing the lowest sum absolute value of the ten deformation vector fields generated by DIR. Results: For AP projections SI and tangential respective deformation maximum displacements were 4.8mm and 4.78mm with a mean displacement of 1.71mm +/− 1.17mm and 1.07mm +/− 0.94mm. For Lateral projections SI and tangential respective deformation maximum displacements were of 3.67mm and 3.07mm with a mean displacement of 0.6mm +/− 0.73mm and 0.72mm +/− 0.64mm. The phase prediction model performs better for AP projections than for lateral projections as lateral patient thickness shrouds respiratory motions. Conclusion: Respiratory motion can still be accurately reconstructed even in the presents of large bony structure deformations. The phase determination matrix nullifies respiratory deformation in the AP projections as SI and tangential maximum and mean deformations are nearly equal and respiration produces large SI motion. Larger non‐respiratory deformations typically occur in the AP projections as the couch helps maintain rigidity in lateral projections. Implementation of parallelized DIR increases spatial accuracy of the integration between multimodality imaging and as an automated tool to assess geometrical changes intra‐ and inter‐fractionally.