Toward a dedicated system for quantitative SPECT mammotomography.

This work extends our previous quantitative brain SPECT research to breast imaging-SPECT mammotomography. A cost-effective dedicated SPECT mammotomography system is presented, which aims to acquire sufficient information for efficient reconstruction of large volumetric images. A very short focal-length fan-beam collimation is designed to maximize the information collected from a relitively small vital organ that has a uniform attenuation property. Data noise of a Poisson nature is accurately modeled and effectively treated in sinogram space, followed by efficient compensation for Compton scattering, uniform attenuation, and collimater response. This sinogram-space statistical approach has the potential to reconstruct high-resolution images at a very fast speed. Yet it provides the same image quality as iterative reconstruction algorithms did, in terms of abnormality detectability by observer studies.