DETECTOR RESPONSE RESTORATION IN IMAGE-RECONSTRUCTION OF HIGH-RESOLUTION POSITRON EMISSION TOMOGRAPHY

A mathematical method was studied to model the detector response of hi gh spatial-resolution positron emission tomography systems consisting of close-packed small crystals, and to restore the resolution deterior ated due to crystal penetration and/or nonuniform sampling across the field-of-view (FOV). The simulated detector system had 600 bismuth ger manate crystals of 3.14 mm width and 30 mm length packed on a single r ing of 60 cm diameter. The space between crystals was filled up with l ead (i.e., septa). Each crystal was in coincidence with 200 opposite c rystals so that the FOV had a radius of 30 cm. The detector response w as modeled based on the attenuating properties of the crystals and the septa, as well as the geometry of the detector system. The modeled de tector-response function was used to restore the projections from the sinogram of the ring-detector system. The restored projections had a u niform sampling of 1.57 mm across the FOV. The crystal penetration and /or the nonuniform sampling were compensated in the projections. A pen alized maximum-likelihood algorithm was employed to accomplish the res toration. The restored projections were then filtered and backprojecte d to reconstruct the image. A chest phantom with a few small circular ''cold'' objects (almost-equal-to 4 mm diameter) located at the center and near the periphery of FOV was computer generated and used to test the restoration. The reconstructed images from the restored projectio ns demonstrated resolution improvement off the FOV center, while prese rving the resolution near the center.