Estimation of lung respiratory motion based on 4D CT

Four-dimensional computed tomography (4D-CT) has been developed as a new imaging technology recently. By scanning the multi-breath cycle of the patient continuously, it can provide full range of lung dynamic information and plays an important role in the diagnosis, tracking and radiotherapy of lung cancer. A B-spline deformable model based registration algorithm is proposed for the accurate estimation of lung respiratory motion using 4D CT data. A new image similarity measure is defined to take advantage of temporal coherence information of 4D CT and guarantee the stability of registration. Meanwhile, the deformation field between end-inhale phase CT and other respiratory phase CT is refined by an iterative way. The registration algorithm is evaluated by comparing with pairwise 3D B-spline based registration algorithm according to simulated ground-truth deformation field. The results demonstrate that the proposed algorithm is more accurate and stable, and it can yield a registration error of 1.22 ± 0.74 mm. Finally, the algorithm is applied to 4D CT data of seven patients to estimate the lung respiratory motion between end-inhale and end-exhale phases.