Algoritmi numerici per la ricostruzione di immagini da tomografia ad angolo limitato

Although Cone-beam tomography is gaining more and more popularity in this field, its high dose is often not justified in several applications, like mammographic screening and dental implantology. In these domains, algebric reconstruction techniques and limited angle acquisition have been fully exploited. However, the acquisition geometry introduces what are called truncation and band artifacts for which innovative algorithms have been developed by our group. The first ones can be eliminated by reformulating the tomographic problem in the context of weighted least-squares theory. Band artifacts can be eliminated reformulating the problem in the context of the regularization theory. We have also recently developed a CUDA-based implementation that produces the tomographic image from 11 projections in less than 3s. Most of these results have been applied on X-ray machines recently launched on the market.