Position-dependent noise in multi-pinhole SPECT

Objectives Dedicated cardiac SPECT cameras employing multi-pinhole designs boast improved energy and spatial resolutions, and count sensitivity over traditional parallel-hole cameras. Variable sensitivity across the field-of-view (FOV) inherent to the pinhole-collimator design can however lead to position-sensitive variations in image noise. Attenuation effects also influence this position-dependent sensitivity. Changes in noise magnitude over the FOV will introduce variation in the detectability of defects in myocardial perfusion imaging. Our objective was to evaluate the magnitude of noise variation in multi-pinhole cardiac SPECT.

Methods Theoretical sensitivity maps including attenuation effects were estimated analytically for the Discovery NM530c camera (GE Healthcare) by tracing the distance and angle between a given voxel and each pinhole. Following Poisson statistics, the standard deviation (σ) in counts for each voxel position was inferred from the square root of the sensitivity map. The impact of the reconstruction on noise was not included. To validate the analytical method, repeat images of a uniform spherical Tc-99m-water phantom (SP; n=16) and a cardiac-torso phantom (TP; n=10) were acquired. The TP acquired counts were set to clinical levels. Images were reconstructed using vendor software with attenuation correction and a voxel-by-voxel σ was calculated. Ten 4-min images were created from clinical listmode data (CC; 350MBq injection, 8min acquisition) by bootstrapping and similarly analyzed.

Results Correlations between the analytical model and experiment had R² values of 0.60 (SP), 0.53 (TP), and 0.47 (CC) showing that >47% of the reconstructed image noise is due to scanner geometry. For TP and CC, σ increased from 0.3% and 0.7% (apex) to 1.1% and 1.2% (base). The experimental σ increased 2.1× across a 12 cm sphere, and 2.6× (TP) and 3× (CC) across the myocardium. The impact on clinical cardiac evaluation will require further study.

Conclusions The variable sensitivity of multi-pinhole cameras caused up to a 3× variation in the noise over the cardiac volume.