Artifact model-based respiratory motion compensation (MoCo) for simultaneous PET/MR based on strongly undersampled radial MR data

Abstract

We propose a new method for PET/MR respiratory motion compensation, which is based on strongly undersampled MR data. In our simulation study, we applied a 3D encoded radial stack-of-stars sampling scheme with 160 radial spokes per slice and an acquisition time of 38 s for MR data acquisition. Based on gated but strongly undersampled and thus streak artifact-contaminated 4D MR images, high-fidelity motion vector fields were estimated applying our newly-developed artifact model-based registration framework. Subsequently, MoCo 4D PET images of a simulated breathing thorax were reconstructed. Evaluation of eight artificial hot lesions in the lungs and upper abdomen showed a significant visual as well as a quantitative improvement in terms of SUVmean values, lesion size and localization for MoCo 4D PET images compared to 3D and 4D gated reconstructions especially for small lesion sizes.

Nuts and bolts of 4D-MRI for radiotherapy

“…This can be a single k-space line, k-space segments or an entire k-space, given that the respiratory amplitude/phase for that part of k-space is known. Most k-space-based sorting methods use 3D non-Cartesian MR sampling (Buerger et al 2012, Stemkens et al 2015, Yue et al 2015a, Deng et al 2016, 2017, Rank et al 2016a, 2017, Freedman et al 2017, Mickevicius and Paulson 2017, while some used Cartesian sampling in 2D (Liu et al 2015b) or 3D (Stemkens et al 2015. An advantage of k-space sorting, and non-Cartesian sampling in particular, is that missing data does not lead to black lines in the reconstructed volumes, but to an increase of undersampling artifacts, which manifest as aliasing in Cartesian and streaking in radial sampling.…”