Real-time respiratory motion correction for simultaneous PET-MR using an MR-derived motion model

Respiratory motion during PET imaging causes the resulting PET images to become corrupted by artefacts. In this paper we describe a technique for motion-correction of PET data based on MR imaging that is suitable for use in a simultaneous PET-MR imaging system. The technique is based on the formation of a subject-specific respiratory motion model from near real-time dynamic MR images and is capable of making real-time motion estimates based on a one-dimensional MR navigator, allowing additional MR imaging to take place at the same time as PET imaging. The model estimates the complex freeform deformations present in the human thorax during respiration. We validate our motion compensation approach using PET simulations based on real MR data of the thorax acquired from a healthy volunteer. Qualitative results show a clear improvement in visualisation of the myocardium and three tumours that were artificially added to the emission/attenuation map of the volunteer close to the diaphragm. Quantitative analysis was based on computing DICE coefficients between true regions of interest (myocardium and the three artificial tumours) and regions manually segmented from a `no motion' PET image, an uncorrected PET image and a motion-corrected PET image. The DICE coefficients over all 4 regions were 0.8 ± 0.03 (`no motion'), 0.24 ± 0.11 (uncorrected) and 0.6 ± 0.08 (corrected), indicating that a significant improvement in PET resolution and quantification is achievable by applying our motion-correction technique.