Estimation of random coincidences from the prompt PET data

Currently the most widely implemented method for estimating the random events present in PET prompt data is the delayed-coincidence-channel method. The delayed randoms are usually subtracted from the prompts. However, the acquired randoms suffer from poor statistics and post processing variance reduction techniques are often used prior to the subtraction. We have developed a new method of estimating the randoms from the PET prompt data acquired in list or sinogram mode (LM or SM respectively). In both cases random events were obtained by swapping the detection coordinates of prompt events. The statistics of the estimated randoms could virtually be unlimited. For LM data the process was straightforward and easily implemented, but for SM data a multi-step algorithm was developed.