Development and evaluation of data-driven respiratory gating methods with simulated list-mode PET data

Data-driven respiratory gating techniques in cardiac PET extract respiratory motion signal from the PET data to guide respiratory gating and motion estimation. However, the influence of the quality of the list-mode data including the uptake ratio between the myocardium and background and the detected count level on its performance has never been studied thoroughly. In this project, list-mode PET data derived from realistic Monte Carlo simulation of different uptake ratios and count levels was employed to quantitatively evaluate the respiratory motion signal extraction and estimation techniques. Simulated projection data were generated from the 4D XCAT phantom with known respiratory motion. The list-mode data were rebinned into projection data at 200 msec intervals and the centers-of-mass of a region-of-interest over the heart were calculated. Contributions from the background were subtracted in the centroid calculation. The peak in the Fourier transform of the time sequence of the centroid locations, or the frequency spectrum, was identified as the respiratory motion signal and its SNR was measured. The high frequency components of the frequency spectrum were removed for noise smoothing before extracting the respiratory motion signal for use in respiratory gating estimation. The results show that with background correction, the amplitude of the estimated respiratory motion signal is increased and is closer to the truth. Both lower myocardium to background uptake ratio and lower count level reduce the SNR of the respiratory motion signal. When they become too low, the extraction of the respiratory motion signal becomes difficult or fails completely.