Least-square pinhole SPECT calibration using a forward projector modeling misalignment

• Published in: 2007 IEEE Nuclear Science Symposium Conference Record Vol. 6; pp. 4543 - 4545
• Main Authors: Wietholt, C., Chin-Tu Chen, Geng Fu, Ling-Jian Meng
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2007
• Subjects: Calibration; Computational modeling; Detectors; Failure analysis; High-resolution imaging; Image resolution; Laboratories; Optimization methods; Shape; Spatial resolution
• ISBN: 1424409225; 9781424409228
• ISSN: 1082-3654; 2577-0829
• DOI: 10.1109/NSSMIC.2007.4437121

Recent improvements in detector technology allow pinhole SPECT to approach spatial resolutions as low as 50 mum. Such high resolutions require a high accuracy of the imaging gantry for preclinical SPECT systems, and is not easily achieved in a laboratory setting. Analytical calibration methods are deemed to fail due to practical implications such as size and activity concentrations of the required point sources. Thus, we have developed a numerical calibration method based on least-square optimization that incorporates the modeling attributes of a forward projector frequently used in iterative reconstruction. The forward projector is currently capable of modeling detector misalignment such as detector rotation and pinhole translation, and will incorporate the pinhole point spread function as well as the size and shape of the point source. Currently, our calibration phantom consists of a single point source, positioned off-center in the object space. The optimization method involves a sequentially fitting of each individual parameter to the data, followed by fitting multiple parameters simultaneous in consecutive steps including all five parameters for the final fitting. The results of the optimization show an error of expected magnitude considering the relatively coarse sampling grid of our simulations. The relatively slow convergence of the y- and z-tilt require simulation of additional point sources. Further improvements of computational and memory efficiency need to be made, which enable the method to fit high resolution data as well as incorporation of the pinhole point spread function and the shape and size of the point sources.