Reconstruction of a high‐quality volumetric image and a respiratory motion model from patient CBCT projections

• DOI: 10.1002/mp.13595
• Publication Date: 2019-05-14T05:04:31Z
• Authors: Minghao Guo; Geraldine Chee; Dylan O’Connell; Salam Dhou; Jie Fu; Kamal Singhrao; Dan Ionascu; Dan Ruan; Percy Lee; Daniel A. Low; Jun Zhao; John H. Lewis
• Keywords: Image registration; Ground truth; Motion field

Purpose To develop and evaluate a method of reconstructing patient‐ treatment day‐ specific volumetric image motion model from free‐breathing cone‐beam projections respiratory surrogate measurements. This Motion‐Compensated Simultaneous Algebraic Reconstruction Technique (MC‐SART) generates uses derived directly the projections, without requiring prior measurements 4DCT, can compensate for both inter‐ intrabin deformations. The be used to generate images at arbitrary breathing points, which estimating during delivery. Methods MC‐SART was formulated using simultaneous reconstruction estimation. For reconstruction, were first binned according external Projections in each bin reconstruct set MC‐SART. estimated based on deformable registration between reconstructed bins, least squares fitting parameters. projection backprojection operations subsequent iterations. These updated then update model, two steps alternated between. final output is reference that any other time point Results A retrospective patient dataset consisting eight lung cancer patients method. absolute intensity differences regions compared ground truth 50.8 ± 43.9 HU peak exhale phases (reference) 80.8 74.0 inhale (generated). 50th percentile voxel error all voxels with >5 mm amplitude 1.3 mm. also applied measured acquired linac‐mounted CBCT system. this data demonstrate feasibility showed qualitative quality improvements state‐of‐the‐art algorithms. Conclusion We have developed estimation Cone‐beam computed tomography (CBCT) high‐quality position. provided superior performance accuracy positional existing methods. resultant combined representing anatomy points.