Image enhancement in positron emission tomography using expectation maximization

Positron Emission Tomography (PET) tomography is one of the imaging modality. PET Tomography scanners collect measurements of a patient's in vivo radiotracer distribution. These measurements are reconstructed into cross-sectional images. Tomographic image reconstruction forms images of functional information in nuclear medicine applications and the same principles can be applied to modalities such as X-ray computed tomography and Single Photon Computed Tomography (SPECT). Reconstruction in PET can be done in two ways, direct and algebraic methods. Iterative reconstruction is an algebraic reconstruction method. The great advantage of iterative methods is that correction to attenuation and depth-dependent detector response can be incorporated to the reconstruction process. One of the drawbacks of the iterative reconstruction methods is the huge computation, due to large system matrices. This system matrix is very sparse. In Matlab 7, matrices having elements more than 100 million can not be executed or stored due to its size restriction. To overcome this problem we have implemented a new storage technique. By this technique, large system matrices can be manipulated in Matlab7. Reconstructed images are compared with the images which are obtained by using direct reconstruction algorithms, namely, Filtered Backprojection.