2D Iterative Image Reconstruction for a Dual Planar Detector for Positron Emission Mammography

Tese de doutoramento em Engenharia Biomédica e Biofísica, apresentada à Universidade de Lisboa através da Faculdade de Ciências, 2008The Clear-PEM system is a prototype scanner for Positron Emission Mammography (PEM), currently under development. In the scope of this project, we present the implementation of three iterative algorithms for 2D image reconstruction, using linograms: the Algebraic Reconstruction Technique (ART), the Maximum Likelihood Expectation Maximization (ML-EM) and the Ordered Subsets Expectation Maximization (OS-EM) algorithms. To calculate the system matrix required by all algorithms, we have developed three different methods (the pixel-driven, the ray-driven and the tube-driven method). Additionally, we have implemented a method to correct the sensitivity differences between different regions of the field-of-view and to compensate for gaps between the detector elements. The comparison of the algorithms' performance was done with the help of Monte Carlo simulated data, which allowed evaluating image spatial resolution, convergence speed, image uniformity, contrast between lesions and background, signalto- noise ratio and computational speed. The results showed that the OS-EM algorithm produced the best results regarding all evaluated parameters, except the computation time by iteration, in which ML-EM and ART presented better results. Compared to OS-EM, ML-EM presented similar results but at the expense of additional iterations, while ART converged after a similar number of iterations, but produced worse results. Regarding the system matrix, the best results were obtained using the tube-driven method. The results pointed to the need of a precise rebinning method and the importance of the system sensitivity correction. The results indicate that, with these algorithms, we can achieve a spatial resolution of 1,5 mm and detect lesions having 3 mm in diameter. The results also lead to think that 2 mm lesions may be detected with acquisition times higher than the ones simulated for this work. These results compare well with results from other PEM scanners, being better in some of the evaluated parameters. However, these results must be validated with real data during the clinical trials that will be carried on in a near future.Fundação para a Ciência e a Tecnologia (FCT), (SFRH/BD/6187/2001); Fundação da Faculdade de Ciência