Combined Motion Estimation and Reconstruction in Tomography

Publication Type:

Conference Paper


12th European Conference on Computer Vision, Lecture Notes on Computer Science, Volume 7583, Firenze, Italy, p.12-21 (2012)


If objects or patients move during a CT scan, reconstructions suffer from severe motion artifacts. Time dependent computed tomography (4DCT) tries to minimize these artifacts by estimating motion and/or reconstruction simultaneously. Most current methods assume a known deformation or a reconstruction without artifacts at a certain time point. This work explores the possibilities of estimating the motion model and reconstruction simultaneously. It does so by modifying the simultaneous iterative reconstruction technique (SIRT) to incorporate motion (trans-SIRT) and uses this method in an optimization routine that computes motion and reconstruction at the same time. Results show that the optimization routine is able to estimate motion accurately, assuming only the type of parametrization for the motion model. Our approach can potentially be extended to more complex motion models.

Research area: