Non-Rigid Motion Compensation for Breast CT

The image quality in dynamic contrast-enhanced breast CT is expected to suffer from motion artifacts due to the extended acquisition time involved. We propose an iterative method for compensation of motion artifacts due to non-rigid movement of the breast. The motion vector field is approximated using b-splines on a sparse grid and its values are found by minimizing errors in the projection domain. We evaluated the method on an anthropomorphic phantom with realistic motion and visual assessment yielded a clear reduction in motion artifacts. Quantitatively, we observed an increase of the structural similarity from 0.9988 to 0.9995 and a decrease of the normalized root mean squared error from 0.1448 to 0.0932.