Retrospective Respiratory Motion Compensation Under Deep Breathing in Spiral Transmission Scanning of 3D PET

3D continuous emission and spiral transmission (CEST) scanning can minimize respiratory motion artifacts inherent in normal free breathing, due to the averaging of emission and transmission images over multiple breathing cycles during the stay time of axial detector rings. However, under deep breathing, which is synchronized approximately with the rotation speed of the transmission source, mushroom-shaped artifacts are produced occasionally where the diaphragm and lung intersect. In this work, we have developed a retrospective respiratory motion-compensation method applied to transmission data which was acquired using six detector rings under deep breathing. Our scheme uses a hybrid compensation method that is based on the filtering and summation of different time-stamp images (FDT) and the filtering of sharp edges (FSE) in the axial plane. FDT calculates the deviations from the average image produced using six time-frame images. FDT first identifies pixels in six time frames whose deviations are over the predetermined threshold, then replaces those pixel values with the minimum-deviation pixel value. Since the bed moves continuously through these detector rings, the six time-frame images can be obtained by reconstructing the six transmission sinograms acquired by each detector ring. After FDT compensation is applied, FSE detects residual sharp edges in the axial direction and applies median filtering, using widths calculated based on edge sharpness. We evaluated this method using patient data that displayed artifacts at the intersection of the diaphragm and lung. These results show that this proposed method provides a practical means to significantly reduce artifacts without using a respiratory gating device.