Image-quality effects of system-matrix-formation statistics in SPECT iterative reconstruction

Abstract

In SPECT imaging with iterative reconstruction, one potential source of image noise is the limited (data or MC) statistics used for the determination of the system matrix. In this paper we present a simulations study of the effect of the system-matrix statistics on the background and signal noise. We simulate a parallel-beam, hexagonal-hole collimator and we study hot cylindrical lesions with contrasts of 1, 2.3, and 9. We investigate 5 levels of projection statistics (1K, 5K 10K, 50K, 100K counts per head position) and 7 levels of system-matrix statistics (N/2 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">k</sup> counts per head position, where N = 2M and k = 0-6). The levels and structure of reconstructed-image noise depends on both the projection and system-matrix statistics, for any given level of object contrast. We present plots that determine the minimum number of system-matrix counts required so that the background noise is not degraded more than a particular level. The better the projection statistics, the more system-matrix statistics we need to maintain the low levels of background noise. Similar effects are observed for the signal noise withing the hot lesions. Determining the appropriate system-matrix statistics (for a particular level of object contrast and projection counts) is critical for not canceling the diagnostic value of higher projection statistics (higher dose).