Iterative deconvolution of simultaneous dual radionuclide projections for CdZnTe based cardiac SPECT

We propose a method of separating primary ^99mTc and ²⁰¹Tl photons in projections acquired on a SPECT camera based on CdZnTe detectors during a simultaneous dual radionuclide rest-stress myocardial perfusion scan. The method is based on a model of spectral and spatial distribution of counts in the measured projections. We use the Maximum Likelihood Expectation Maximisation (ML-EM) iterative algorithm to solve the model equations for the pure primary counts. The deconvolved primary count projections are then reconstructed using the standard OSEM algorithm. Considerable improvement of defect contrast can be observed in anthropomorphic phantom scans.