Software tools for 3-D PET imaging : design and development

This thesis describes the design and development of PET image reconstruction tools within the framework of an easy to use graphical user interface (GUI). The components of this software include image conversion codes, data correction, and codes for image reconstruction and registration of image volumes. The data correction codes compensate for phenomena such as Compton scatter, random coincidence events, attenuation, and detector efficiency in PET sinograms. Normally, scatter and randoms correction techniques do not improve image quality much for two-dimensional images. However, since nearly 50% of three-dimensional PET sinogram data is noise from these sources, data correction to compensate for these phenomena must be done to achieve reasonable image quality for 3D PET image reconstruction. The image reconstruction algorithms implemented include Fourier rebinning (FORE) and filtered backprojection (FBP). Filtered backprojection is implemented as a 2D reconstruction method and FORE as a 3D reconstruction application which rebins 3D data into 2D data sets and then uses the filtered back projection method to reconstruct the images from sinograms. Image registration is the process by which images of different modalities or subjects can be conformed to a common coordinate system so that they can be compared to provide complementary information as in PET and MRI or CT studies. The graphical user interface (GUI) which wraps around all these modules, presents a cohesive, and easy-to-use framework for any application environment. It has been specifically designed in C++ using the AMULET libraries, for a clinical setting. The interface is user-friendly and provides control over all useful parameters for image conversion, correction, reconstruction, and visualization.