TOF scatter estimation through TOF true distribution generation from non-TOF image reconstruction

An alternate methodology to estimate TOF scatter distribution is proposed. The scatter estimation is obtained by subtracting attenuated TOF true coincidence event estimations from measured TOF prompt coincidence events, corrected for random events and normalization. TOF true coincidence distribution is estimated by TOF forward projection of non-TOF reconstruction images, fully corrected for normalization, attenuation, random events, and non-TOF scatter. Therefore TOF model of scattered events was replaced by a TOF model of true coincidence events. The estimated noisy TOF scatter distribution needs additional smoothing based on an a priori assumption of the low frequency content of scatter distribution. Initial investigations of torso phantom experimental data showed that proposed TOF scatter correction resulted in similar TOF image reconstruction quality, when compared to clinically used TOF scatter correction.