Marcus Brehm, Abstract Co-Author: Nothing to Disclose

Pascal Paysan PhD, Abstract Co-Author: Employee, Varian Medical Systems, Inc

Markus Oelhafen DPhil, DSc, Abstract Co-Author: Employee, Varian Medical Systems, Inc

Patrik Kunz PhD, Abstract Co-Author: Employee, Varian Medical Systems, Inc

Marc Kachelriess PhD, Presenter: Nothing to Disclose

In respiratory-correlated cone-beam CT we suffer from streak artifacts and from reduced dose usage because only a subset of the acquired projection data contributes to each time frame. The aim of this study is to find a reconstruction algorithm that obtains high quality respiratory-correlated 4D volumes from flat detector cone-beam CT scans and that guarantees 100% dose usage.

To increase image quality and dose usage we developed a new approach that simultaneously estimates the respiratory motion vector fields and reconstructs respiratory-correlated 4D volumes. For the vector field estimation a modified deformable registration algorithm is used. To make the registration algorithm less sensitive to streak artifacts the registration method is regularized by new temporal constraints like imposing cyclic motion patterns, for example. This becomes necessary due to the fact that the estimation is performed on the conventional phase-correlated 4D volumes. On the basis of these motion vector fields a motion-compensated reconstruction is performed to obtain high quality respiratory-correlated 4D volumes. The method is verified using simulated rawdata obtained by deforming a clinical patient dataset by realistic deformation fields, and by processing patient data acquired with the TrueBeam 4D CBCT system (Varian Medical Systems).

Due to the regularization the motion vector fields are insensitive to the streak artifacts inherent in the original images. The streak artifacts in the motion-compensated reconstructions from both, simulated data and patient data, are almost completely suppressed. The noise level of our motion-compensated reconstruction is comparable to a standard reconstruction, which indicates a 100% dose usage.

Motion estimation is possible based on the conventional respiratory-correlated cone-beam CT images. The impact of artifacts on the registration process can be suppressed by additional constraints. Motion vector field estimation combined with a motion-compensated reconstruction is capable to remove streak and motion artifacts to a large degree while maintaining a good spatial and good temporal resolution. The dose usage is significantly increased compared to the conventional method because all data now contribute to each time frame.

Image quality is increased by the proposed method. Therefore a radiation treatment planning and verification is improved.

Brehm, M, Paysan, P, Oelhafen, M, Kunz, P, Kachelriess, M, Motion Vector Field Estimation and Motion-Compensated Reconstruction for Flat Detector Cone-Beam CT Scans of Breathing Patients.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12021862.html