Recombining respiratory gated PET frames

Respiratory gating in PET can improve image resolution, but comes at the cost of an increase in noise. We are presenting a method based on the idea that there is useful information in the voxel time-activity curves. The overall idea of the algorithm is that we want the voxels in the motion path of high activity gradients to be gated, to preserve the resolution in these areas, and the voxels in uniform regions to be ungated, to preserve the statistics in these areas. To test our methods, we generated 4D image models simulating a spherical lesion moving in sinusoidal motion in several different conditions. Both Poisson noise and blurring were added to all the images. The combined gated images demonstrated a 34% improvement in signal-to-noise ratio relative to the gated images alone, while preserving the gated images' improved resolution: the distance between 10% and 90% activity along the profile of the simulated lesion was 8 pixels for the ungated image, 3 pixels for the gated image, and 3 pixels for the combined image. This strategy may thus be useful in improving image quality and the accuracy of quantification of activity found in PET scans.