CORRECTION OF DISTORTIONS IN A DISCONTINUOUS IMAGE

Large area position-sensitive NaI detectors have been successfully app lied to positron emission tomography (PET). Typical PET studies involv e detector singles rates in excess of 500 kcps, which can lead to pile -up and image degradation as a function of countrate. Better high coun trate performance can be achieved with a local centroid algorithm, in which the position of each event is calculated from a small group of p hotomultipliers (PMTs) immediately surrounding the PMT with the highes t signal (the peak PMT). The local centroid contains most of the light from the scintillation. If a local centroid of only seven PMTs is use d, the position resolution becomes quite stable at high countrates, ho wever, discontinuities appear in the detector flood image as events cl uster around each PMT. We therefore developed a method for distortion correction of a discontinuous flood image. For each PMT on the detecto r, a peak PMT domain is defined. The peak PMT domain is the area on th e detector where that PMT could have the highest signal. The peak PMT domains overlap slightly, so that all combinations of peak PMT and pos ition are represented. A collimated source is moved through a regular grid of points - a template - over each peak PMT region. A short colle ction is performed at each point, and the real and measured position d ata are recorded. For each point in the spatial range of a given PMT, distortion correction offsets are computed by interpolating between th e data points which correspond to that PMT. This new method has been i mplemented and evaluated. System resolution has been measured at low a nd high countrates. The high countrate resolution is better with the n ew method, with no degradation in low countrate resolution. The axial sensitivity profile is also more stable at high countrates, compared t o the previously developed method. 3-D brain phantom images show a cle ar improvement in image quality at high countrates.