Iterative region-of-interest reconstruction of truncated projections with slit-slat collimation

Slit-slat collimation is a geometry for SPECT imaging that combines a slit along the axis of rotation of a gamma camera with a set of septa perpendicular to the slit. Herein we investigate through simulation and experiment the potential to accurately reconstruct a region of interest using truncated slit-slat projection data. Projections of digital phantoms (10-cm diameter) were simulated for circular orbits with 5 different RORs, which produced various degrees of truncated projection data. The projections were reconstructed slice by slice using an MLEM algorithm (10 iterations). Experimental data were obtained with a prototype slit-slat collimator attached to a gamma camera. The phantoms had 3- mm alternating hot and cold regions. Three phantom configurations were acquired at RORs of 9.1 and 21.1 cm. In the 21.1-cm ROR case, the phantom was offset 0, 3, 6, 9 and 12 cm from the AOR to search for sampling deficiencies. The experimental projections were reconstructed with the same iterative algorithm. Analytical calculations were also carried out to determine the expected regions of the reconstruction that utilized only untruncated data, as well as regions that were well sampled using cutoffs of 80, 95 and 99%. Both simulated and experimental data show that central ROIs could be accurately reconstructed from truncated projection data, even though blurring occurred outside this ROI. The sizes of the accurately reconstructed regions were consistent with the analytical calculations.