A practical method for randoms subtraction in volume imaging PET from detector singles countrate measurements

There are several methods to estimate random coincidences in PET scans including delayed coincidence, profile distribution, and singles detector countrate measurements. The delayed coincidence method collects delayed coincidence randoms along with the prompt coincidence randoms rate, so increasing both the noise and deadtime due to randoms. The profile distribution background method estimates randoms (and scatter) from the measured distribution outside the object boundaries. This method is necessarily a compromise between scatter and randoms subtractions which have dissimilar distributions. The authors propose a practical method to perform randoms subtraction in PET based on the singles countrates. The singles distribution is measured across the detectors and is used to construct a randoms distribution sinogram. This distribution is scaled to the appropriate randoms scan countrate by means of a lookup table of randoms countrate vs detector singles countrate, generated from phantom calibrations. The advantages of performing randoms subtraction by this method are: (1) there is no increase in camera deadtime, (2) the method compensates for non-uniformities in randoms distributions due to both the activity distribution and non-uniform geometric response of the camera for on and off bankpairs, and (3) it deals with randoms subtraction independently of scatter so that different scatter correction routines may then by applied to the data. Phantom tests at high countrates with /sup 18/F, and also with a single-photon emitter, /sup 137/Cs, where all measured events are randoms, have shown the method to accurately represent the randoms distribution. Comparisons have been made with the profile distribution and delayed coincidence methods and the new method has been successfully applied to body PET studies.