Improved I-131 SPECT resolution through modeling individual medium-energy collimator holes

In previous work, a thick-septa collimator that rotates about an axis perpendicular to the collimator surface has been proposed. This collimator has thick septa which nearly eliminate septal penetration but which also block much of the image plane at any one rotation stop. The axis of rotation and the hexagonal configuration of holes are such that by summing data acquired at three rotation stops equally spaced over 360 degrees, approximately uniform sampling of the image plane is obtained. These summed data can then be reconstructed by general purpose analytic or iterative methods. However, summing the data may erase information inherent in the three, initially separate projection data sets. The present paper has two objectives: (1) to develop a method for realistically modeling the response of thick-septa collimators within iterative reconstruction and (2) to use this method to investigate whether it is beneficial to reconstruct rotating collimator projection data with the three data sets maintained as separate, rather than summed. Reconstructing with the three data sets maintained as separate has at least two advantages. First, it accelerates reconstruction significantly. In the study considered here, computation time to reach a particular bias was reduced by a factor of 3.3. Second, it results in less noise for a given bias. Both of these effects, acceleration and reduced noise, were increasingly pronounced with decreasing bias