Towards model-based randoms correction for 3-D dual head coincidence imaging

Random coincidences ("randoms") can be a major source of image degradation in 3-D dual head coincidence imaging (DHCI) systems which do not use measured randoms correction. Smooth background subtraction may be used, but may not be accurate for complex activity and attenuation distributions such as found in the thorax. We are developing a novel, calculation model-based randoms correction for DHCI. It requires emission and transmission data, but no additional scans. First, the distribution of single events ("singles") to the projections from a set of randomly distributed emission points within the reconstructed objects are calculated. The probability of detecting single events from an emission point is calculated for every pixel in each of the opposing detectors at multiple detector positions. The randoms distribution for each projection is then calculated from the product of the singles distributions of opposing detectors. Singles from activity outside the field of view may be included. The code is written in MATLAB for portability and is optimized by pre-calculation and vectorization. Initial results show good agreement between calculated singles distributions and Monte Carlo simulations for 20 cm diameter, 20 cm long and 73 cm long cylindrical phantoms. Further code development and validation are underway.