A Simulation Framework to Establish Ground Truth for Motion Correction in Total-Body PET: Initial Evaluation for Complex Respiratory Motion

The enhanced sensitivity of total-body PET/CT scanners substantially decreases noise, however the full potential for spatio-temporal resolution is still hampered by artefacts and quantification degradation due to patient motion. To compensate for this, a variety of different motion correction methods were developed, but lack a motion free reference. To this end, our group developed a framework enabling Monte Carlo simulations of patient-like phantoms with motion modelling that accurately resembles event processing and image reconstruction of the Biograph Vision Quadra total-body PET/CT scanner. Simulations were performed with the XCAT phantom to mimic a clinical 5 min $\left[{ }^{18} \mathrm{~F}-\mathrm{FDG}\right]$ scan, with modelling of breathing motion with various breathing frequency/amplitude. To evaluate the impact of motion correction, an elastic motion deblurring approach, OncoFreeze AI, was used. Image quality, quantification of 5-20 mm lesions in lung and liver, and lesion line profiles were compared for a static (no motion) reference and for modelled motion with and without correction. Motion induced blurring in the images and could be compensated by the motion correction. The respiratory curve could be recovered by OncoFreeze AI with high precision (difference $\mathrm{Hz} \pm 0.5 \%$ over all motion segments). The motion correction helped to increase the recovery of the static activity concentration, e.g. for a 10 mm lung lesion: $9.4 \mathrm{kBq} / \mathrm{ml}$ (static no motion), $7 \mathrm{kBq} / \mathrm{ml}$ (motion) and $8.3 \mathrm{kBq} / \mathrm{ml}$ (motion corrected) and recovery of the lesion size 8.0 mm FWHM (static no motion) to 13.4 mm (motion) and 8.6 mm (motion corrected). This work underlines the need for motion correction and confirms the suitability of our framework to provide ground truth information on patient motion for total-body PET scans. Future work will extend this initial evaluation to more motion correction methods and a larger, more diverse data set with different scan durations, activities, and lesions.