Compton imaging using the SmartPET detectors

Image reconstruction from Compton camera data is a complex problem requiring investigation. Generally reconstruction is conducted using iterative reconstruction methods such as Maximum Likelihood - Expectation Maximization (MLEM). However, iterative reconstruction into volumetric grids is a computational burden. Analytic methods of image reconstruction have been proposed which relieve the computational expense. However, such methods usually have unrealistic sampling assumptions or are not directly extendable to situations where the scattering detector is extended when compared to the detector-source distance. Starting from a standard inversion technique, a generic method of filtering inversion co-efficients is developed. By dynamically assigning the strength of co-efficients on an event-by-event basis, artifacts arising from the difference between assumed and actual forward transform may be reduced. Dynamic assignment allows inversion of single cone-surfaces, so that the generic technique may be implemented in volumetric reconstruction. Results are compared to direct-back-projection in a limited-angle tomography context. The ability to apply event-by-event analytic image reconstruction provides many advantages when compared to standard iterative techniques.