Improved axial resolution in 2D PET with 3D reconstruction

Abstract

A 3D reconstruction method originally developed by the authors for 3D scanners has now been implemented for handling data from a 2D scanner. The scanner consists of 4 detector rings and gives 7 simultaneous planes including cross-planes. The axial sampling is improved by moving the patient couch. This will halve the plane separation, but there will be cross-talk between the planes. The axial sensitivity profile for each plane is, for the direct planes, fairly independent of radial position, but varies significantly for the cross-planes. The authors' 3D reconstruction technique is based on 2D reconstruction of transaxial planes and deconvolution of a position dependent axial spread function. The authors have tested two different methods for the deconvolution step; the iterative Landweber method and singular value decomposition. Evaluations were made with Monte Carlo simulations and phantom studies. The result of 3D reconstruction an improved axial resolution at the cost of a higher noise level. Iterative deconvolution gives the best compromise between resolution and noise.