Projection-based 3D/2D registration for prospective motion correction.

Purpose: To develop a registration method that is capable of estimating the full range of rigid body motion from three orthogonal collapsed images of the head. These images can be obtained using the collapsed FatNav, a previously introduced navigator for prospective motion correction. It combines a short duration with wide compatibility with different main sequences due to its robustness against spin history effects. Theory and Methods: A projection-based 3D/2D registration method is presented and then modified to take into account the peculiarities of the collapsed FatNav. Water/fat separated volumes were used in simulations to assess the accuracy of the proposed method at different resolutions by comparison with high-resolution 3D registration. The sensitivity with respect to masking strategies and starting motion parameters was investigated. Finally, prospective experiments with a healthy volunteer were performed with different types of motion patterns. A PROPELLER main sequence was chosen to compare the prospective correction with PROPELLER's own retrospective correction. Results: In the simulations the proposed method has shown comparable performance to 3D registration. Furthermore, evidence of its robustness with respect to masking strategies and starting motion parameters was presented. The combination with collapsed FatNav has performed well in correcting most of the motion artifacts prospectively with improved image quality compared to only using PROPELLER's retrospective motion correction. Conclusions: The proposed 3D/2D registration together with collapsed FatNav is characterized by a good balance between navigator duration and estimate accuracy. Further work is needed to validate the method across a wider variety of subject anatomies.