Selection of point source configuration for multi-pinhole SPECT calibration

For pinhole SPECT, geometric calibration is needed to acquire an accurate description of the system matrix. Normally it is done by measuring several point sources in the field of view, and estimating all the parameters by comparing the estimated projections and the measured projections using a least squares fitting procedure. In a previous study, it was shown that the accuracy of rotating pinhole SPECT calibration is not only affected by data noise, but also influenced by the systematic deviations between the expected camera motion and the actual camera motion. The degradation of the image quality is dominated by these systematic deviations rather than by data noise. In this study, we aim to determine a point configuration which yields an accurate calibration solution for our multi-pinhole SPECT in the presence of systematic deviations. An analytical method was used for point configuration selection. Two types of systematic deviations were applied. The first one were the typical systematic deviations of our camera, estimated from previous calibration scans. The second one corresponds to the systematic deviations that produce the worst possible reconstruction error (WPRE) with constrained amplitudes. To facilitate simultaneous animal/calibration scans, the point source candidates were sampled on the surface of a cylindrical tube, which was assumed to be the animal bed positioned in the central field of view. The best 2-point configuration was applied in a real measurement of a Jaszczak-type resolution phantom. The experimental results were in agreement with the theoretical results.