Noise reduction for Multi-Harmonic Phase Analysis of gated SPECT myocardial perfusion imaging

Multi-Harmonic Phase Analysis (MHPA) has been developed for heart failure prognosis by reliably measuring left-ventricular (LV) dyssynchrony from conventional gated single photon emission tomography (GSPECT) myocardial perfusion imaging (MPI). Although MHPA has been evaluated in multiple clinical studies, its accuracy can be limited under highly noisy conditions. The purpose of this study is to develop a 4D postreconstruction method for suppressing noise in GSPECT MPI data. SPECT images acquired at different temporal points were first registered to the same point. By summing the registered images, the uncorrelated noise was greatly reduced and the myocardial signals were strengthened. The same process then repeated for different temporal points. A deformable registration with a simulated annealing optimization scheme was implemented in this work. The proposed method was evaluated using a simulation study on an extended cardiac torso phantom (XCAT). The projection data were first calculated analytically and Poisson noise was added with a level matching that in clinical data. An ordered subsets expectation maximization (OSEM) algorithm was used in the reconstruction. The comparison of SPECT images without versus with the noise reduction showed that the proposed noise reduction method increased the signal-to-noise ratio (SNR) by an average factor of ~2.3. Our approach was able to substantially reduce image noise without losing faithful information of myocardial activities. The next step is to use this method to reduce noise in clinical GSPECT MPI data to improve the accuracy of MHPA, especially in a region with severe myocardial scar.