Iterative tomographic reconstruction with applications to breast and small animal imaging

CT and SPECT are two popular noninvasive medical imaging modalities. Image reconstruction algorithms play an extremely important role in these tomographic imaging systems. In particular, iterative reconstruction algorithms have the advantages of being able to handle arbitrary geometries, reconstruct from incomplete data, and incorporate stochastic image and noise models. In this dissertation, we described novel iterative reconstruction algorithms for two CT/SPECT dual modality applications. The first application is limited angle breast imaging, where only a few projection images for both CT and SPECT are available and quantifying lesion radioactivity is difficult. We developed algorithms to reconstruct 3D images, correcting for attenuation, detector blurring, and limited angle effects. We estimated lesion volume from the X-ray CT reconstruction and the total lesion radioactivity from the SPECT reconstruction. Quantification results using experimental data showed that our technique achieved an accuracy of less than 15% error. The second application is high-resolution small animal imaging, where limited gamma detector size constrains spatial resolution and field of view. We studied a novel half-cone/dual-cone SPECT data acquisition geometry, which significantly increases resolution without reducing the field of view, and we developed iterative joint reconstruction/deblurring algorithms suitable for the half-cone geometry. Reconstruction of experimental phantom data demonstrated much better spatial resolution than conventional pinhole SPECT.; Finally, as a contribution to the theory of iterative reconstruction, we proved the convergence of our reconstruction algorithms for both breast imaging and small animal imaging applications.