Fast single scan derivation of the PSF resolution model on the TruePoint PET/CT using a printed point source array

• Published in: IEEE Nuclear Science Symposuim & Medical Imaging Conference pp. 3558 - 3562
• Main Authors: Kotasidis, F A, Matthews, J C, Angelis, G I, Noonan, P J, Markiewicz, P J, Lionheart, W R, Reader, A J
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2010
• Subjects: Arrays; Computed tomography; Image reconstruction; Kernel; PET/CT; point spread function; Positron emission tomography; printed point sources; PSF; Spatial resolution
• ISBN: 9781424491063; 1424491061
• ISSN: 1082-3654; 2577-0829
• DOI: 10.1109/NSSMIC.2010.5874471

Incorporation of the resolution modelling component of the scanner during statistical image reconstruction often produces images of improved signal-to-noise ratio (SNR). In our previous work we measured and parameterized the Hi-Rez PET/CT scanner's point spread function (PSF) in image space using printed point source arrays at multiple positions. Here we extend the work to the TruePoint PET/CT and specifically report on i) optimizing the PSF model ii) demonstrating that a simple, single scan of a printed point source array is sufficient to accurately characterize the resolution blurring properties of the TruePoint PET/CT iii) evaluating and validating an image based resolution modelled reconstruction algorithm incorporating the measured scanner response against the scanner's projection based resolution modelled reconstruction algorithm on phantom data. A single array of 126 (14×9) point sources was scanned at different axial and azimuthal positions to measure the symmetric properties of the PSF model in image space. Results show that the blurring is more pronounced in the axial direction due to the spatial resolution losses incurred by the large span (span 11) used during binning of the raw data. As expected, the 3D resolution kernel varies in the axial, radial and tangential directions, but it is importantly shown that these variations can be captured and factorized into separate radial and axial dependencies. With this parameterization for the PSF, a single scan of the array can characterize the image-based PSF without compromising the accuracy of the measured model. This provides a fast, efficient and practical means of measuring the unique resolution modelling component of the system matrix on a scanner-by-scanner basis. For scanners also having the option of a PSF reconstruction, this fast and practical procedure can be used to rapidly check and if necessary correct any drift in the blurring properties of the scanner on a routine basis.