Image Reconstructions from Limit Views and Angle Coverage Data for a Stationary Multi-Pinhole SPECT System

The goal of this paper is to investigate different reconstruction methods for solving the limited angle problem in reconstructing the projection data from a stationary multi-pinhole system based on a three-head clinical single photon emission computed tomography (SPECT) camera. Three iterative reconstruction approaches were studied including maximum likelihood expectation maximization (MLEM), maximum a posteriori expectation maximization (MAPEM) with a smoothing prior, and an alternating optimization schemes from MLEM and total variation (TV) optimization. A three-headed multipinhole SPECT was simulated with apertures of nine 1-mm pinholes each, and covered scanning volume of 6-mm diameter. The reconstructions were optimized for various iterations based on visual inspections, and finally 20 iterations were applied for each method. For both MLEM-TV and MAPEM-TV, various initial reconstructions before TV optimization were studied. The smoothing parameter for MAPEM and the gradient descent constant for TV were also investigated through visual comparison. The preliminary results showed the 3 reconstruction methods generated compatible images, and can restore the images from projection data suffering limited angular sampling. However, MLEM was noisy for low-count and highly limited angle data, and thus suitable smoothing in MAPEM alleviated this problem. Initial reconstructions were necessary for better edge enhancement in TV. The conclusion is that TV might be potential in producing more edge-enhanced images if all parameters were optimized.