Improved image reconstruction for dedicated cardiac SPECT with truncated projections

105 Objectives The small field-of-views (FOV) of dedicated cardiac SPECT can cause projections truncated by radiotracer in organs outside the FOV. This may lead to artifacts and overestimation in quantifications. In this study, we investigated the impact of different image matrix size on reconstruction and propose using body contour obtained from CT to improve reconstruction on a multiple pinhole small FOV SPECT/CT system. Methods We simulated the system with 27 pinholes arranged on a cylindrical surface that covers 180 degree of the body. The FOV of the system was 21cm in diameter. We compared MLEM reconstruction with image matrix of 60x60x128 that covers just the FOV (IM60), 80x80x128 that covers the majority portion of the body (IM80), large image matrix 128x128x128 with body contour (LIMBC) obtained from CT images that constrains the reconstruction, and the same matrix without constraints (LIM). Noiseless data and 20 noise realizations with ~230K counts per projection of a NCAT phantom with defects were simulated. Quantitative accuracy was evaluated in terms of bias vs noise, and contrast recovery. Results The reconstructions with IM80, LIMBC and LIM reduced the artifacts in the myocardium and blood pool caused by projection truncations. For quantitative evaluations, IM60 yielded accurate results in myocardium, but ~20% and ~115% over-estimation in defect and blood pool, respectively, which led to reduced contrast recovery. LIMBC yielded the most accurate quantification, where LIM yielded up to 10% under-estimation of regional counts in the myocardium. Conclusions Our results indicate that application of an image matrix that matches with the body contour provides the most accurate reconstruction for small FOV multiple pinhole SPECT, while smaller image matrix can cause artifacts and over-estimation, and arbitrary large image matrix can lead to an under-estimation of counts