A fast, energy-dependent scatter reduction method for 3D PET imaging

We present a new method of scatter reduction for PET with list-mode acquisition. Our method is based on a predetermined true fraction table for preferentially reducing scatter events based on the two detected photon energies in a coincidence event. For this method to work, the true fraction table needs to be insensitive to the anatomy of the phantom. In this paper, by use of Monte-Carlo technique we calculate the true fraction tables for various activity and attenuation distributions and demonstrate that these tables are indeed robust to wide variations of activity and attenuation distributions. Hence, by employing a single true fraction table, very effective scatter reduction can be obtained for PET data that are derived from substantially different activity distributions. The resulting scatter-reduced PET data are also shown to have improved noise equivalent count and to produce images of better contrast. This scatter reduction method is computationally efficient; the operations can be performed in real time for list-mode acquisition and is attractive for practical use.