Hybrid Head motion correction in PETMR Brian Imaging

Hybrid Head motion correction in PET/MR Brian Imaging Alaleh Rashidnasab 1, Benjamin A. Thomas 1, Richard Manber 1, Marilena Rega 1, Ilaria Boscolo Galazzo 2, Francesco Fraioli 1, Anna Barnes 1, Brian F. Hutton 1, 3 and Kris Thielemans 1 1 Institute of Nuclear Medicine, University College London, UK of Neuroradiology, University Hospital Verona, Italy 3 Centre for Medical Radiation Physics, University of Wollongong, Australia 2 Department

Aim A simple hybrid method to correct dynamic PET data for head motion during simultaneous PET/MR using routinely available MR and the PET data. Fast volumetric time frames

Overall Framework • • MR-based motion estimation in the beginning of the scan Motion detection using PCA on list-mode PET-based motion estimation in later frames Motion correction ASL PET MR motion estimates* PCA ** Realign Data-driven time frames NAC recon. + * Rigid registration using Niftyreg ** Syncing MR and PET + Reconstruction using STIR with attenuation, scatter and randoms correction MRAC μ-map PET motion estimates* AC recon. + Realign recon. frames

Overall Framework • • MR-based motion estimation in the beginning of the scan Motion detection using PCA on list-mode PET-based motion estimation in later frames Motion correction ASL PET MR motion estimates* PCA ** Realign Data-driven time frames NAC recon. + * Rigid registration using Niftyreg MRAC μ-map PET motion estimates* AC recon. + Realign recon. frames

Overall Framework • • MR-based motion estimation in the beginning of the scan Motion detection using PCA on list-mode PET-based motion estimation in later frames Motion correction ASL PET MR motion estimates* PCA ** Realign Data-driven time frames NAC recon. + * Rigid registration using Niftyreg ** Syncing MR and PET MRAC μ-map PET motion estimates* AC recon. + Realign recon. frames

Overall Framework • • MR-based motion estimation in the beginning of the scan Motion detection using PCA on list-mode PET-based motion estimation in later frames Motion correction ASL PET MR motion estimates* PCA ** Realign Data-driven time frames NAC recon. + * Rigid registration using Niftyreg ** Syncing MR and PET + Reconstruction using STIR with scatter and randoms correction MRAC μ-map PET motion estimates* AC recon. + Realign recon. frames

Overall Framework • • MR-based motion estimation in the beginning of the scan Motion detection using PCA on list-mode PET-based motion estimation in later frames Motion correction ASL PET MR motion estimates* PCA ** Realign Data-driven time frames NAC recon. + * Rigid registration using Niftyreg ** Syncing MR and PET + Reconstruction using STIR with attenuation, scatter and randoms correction MRAC μ-map PET motion estimates* AC recon. + Realign recon. frames

Software issues • Niftyreg reg-aladin rig. Only: registration not good for small changes • Syncing MR and PET start time based on scanner time stamps • Offset for first time stamp for first timing event derived from list-mode (list_lm_events --num-events-to-list) • Offset for injection time delay, derived from list-mode (list_lm_countrates), threshold difference between prompt and delayed events • Patient orientation/rotation from Siemens= HFS -> added into STIR Interfile • Subject mu-map using e 7 tool: convert to 1/cm divide mu-map values by 10000. • Remove offset keywords in mumap. hv as STIR recon don’t understand patient position • Values output from STIR (proportional to counts) need scaling for KBq/ml related to global calibration factor

Patient data Epilepsy study using dynamic PET and perfusion-ASL Acquired on a Siemens Biograph m. MR scanner. 4 ASL sequences, 100 brain volumes each with a temporal resolution of 2. 8 s

Results NAC reconstructed frame with motion blur Reconstructed frames before and after motion occurrence Aligned motion frame using MR motion estimates Registered motion frame using NAC PET motion estimation Difference image of two motion corrected frames