作者: David Sarrut , Vlad Boldea , Serge Miguet , Chantal Ginestet
DOI: 10.1118/1.2161409
关键词:
摘要: Purpose: We propose to simulate an artificial four-dimensional (4-D) CT image of the thorax during breathing. It is performed by deformable registration two scans acquired at inhale and exhale breath-hold. Materials methods: Breath-hold images were with ABC (Active Breathing Coordinator) system. Dense registrations performed. The method was a minimization sum squared differences (SSD) using approximated second-order gradient. Gaussian linear-elastic vector field regularizations compared. A new preprocessing step, called priori lung density modification (APLDM), proposed take into account changes due inspiration. consisted modulating densities in one according other, order make them comparable. Simulated 4-D then built interpolation resampling initial images. variation taken generate intermediate Jacobian deformation used compute voxel values Hounsfield units. accuracy assessed spatial correspondence anatomic landmarks located experts. Results: APLDM produced statistically significantly better results than reference methodmore » (registration without preprocessing). mean (and standard deviation) distances between automatically found landmark positions set experts 2.7(1.1) mm APLDM, 6.3(3.8) without. Interexpert variability 2.3(1.2) mm. linear elastic not significant. In second experiment images, difference automatic manual for 2.6(2.0) Conclusion: generation feasible. This can lower dose needed acquisitions or help correct acquisition artifacts. model be propagate contours, map, irregular breathing signal. could serve as basis radiation therapy planning. Further work simulation more realistic taking hysteresis complex trajectories.« less