Toward a framework for high resolution parametric respiratory motion modelling

A framework to facilitate the realization of dynamic MRI data with simultaneously increased temporal and spatial resolution is proposed. Deformable registration to a reference frame of spatially sparse, high temporally resolved two dimensional sagittal slices acquired sequentially across a volunteers lateral dimension serve as a subset of incomplete observation of full three dimensional vector fields. Registration of an averaged, re-binned, single respiratory cycle with full spatial sampling serves as a basis for estimation of full three dimensional vector fields derived from the sparse subset. The inverse of the estimated full 3D vector fields from sparse measurements allows propagation of a high resolution, re binned static volume with breathing modes derived from the sparse dynamic data. Proof of concept experiments are undertaken with the anthropomorphic XCAT phantom. A quantitative evaluation of full vector fields derived from sparse samples in comparison to their ground truth results in a mean error of the order of 1mm. A qualitative assessment of the motion of a propagated high resolution static MRI volume is presented.