High Accuracy Multiple Order Scatter Model for 3D Whole Body PET

Abstract

A technique for modelling multiple Compton scatter as the absolute probabilities which relate the image space to the projection space in 3D whole body PET system is presented. Such modelling is advantageous for large attenuating media where scatter is a dominant component of the measured data, and where multiple scatter may constitute up to about half of the total scatter depending on the energy settings and attenuating object. The model developed in this work goes beyond the limitations of previous methods in three distinct ways: (1) specification of the scatter distribution for every voxel based on calculation of actual probabilities, using the transmission data, the physics of Compton scattering and the specification of a given PET system (independence from a true activity estimate); (2) in principle no scaling or iterative process is required to find the scatter distribution; (3) explicit multiple scatter modelling; (4) easily adaptable to time-of-flight PET scatter estimation. The model can be included in a system matrix for statistical image reconstruction methods (avoiding the additive or subtractive approaches to scatter compensation), and of course the method can be adapted to a whole host of differing scatter compensation methods of varying speed and complexity. The proposed, high accuracy model has been validated using Monte Carlo simulation (SimSET).