Image reconstruction for a 3D PET system using a minimum norm constraint

Abstract

Describes the reconstruction procedure for the three-dimensional positron imaging system, QPET, at Queen's University. Because the detectors have 10¹⁰ coincidence projection lines, the backprojection step is performed first using a new area-overlap algorithm which reduces rebinning artefacts. The filtering step is an analytical deconvolution which selects the minimum norm solution subject to a least-squares constraint. The cut-off parameter used in the constraint is evaluated from the statistical fluctuations in the backprojected data. This method provides a fast image reconstruction procedure with the amount of smoothing determined from the noise level of the data.