Publication:
Impact of BPL penalty functions and PSF models in PET/CT radial spatial resolution recovery

Loading...
Thumbnail Image
Date
2024-07
Authors
Si, Wang Jie
Journal Title
Journal ISSN
Volume Title
Publisher
Research Projects
Organizational Units
Journal Issue
Abstract
Introduction: A conventional three-point source method adhering to NEMA NU 2-2018 has been introduced where three-point sources are placed at 1 cm, 10 cm, and 20 cm within the PET/CT FOV to evaluate impact of radial distance on spatial resolution. Advanced reconstruction algorithms like TOF, BPL, and PSF models, along with OSEM have shown spatial resolution enhancements based on the FWHM measured. Combining OSEM with TOF improves event localization, and PSF modelling mitigates parallax effects. However, OSEM often struggles with full convergence, necessitating BPL with smooth penalty functions (β values) for regularized iterative reconstruction. Since National Cancer Institute (NCI) currently practice OSEM+TOF+PSF+BPL (β 450) as the in-house protocol, this study explores alternative BPL and PSF functions in OSEM to enhance spatial resolution and determine optimal β values for NCI's routine clinical use. Methods: In-house spatial resolution phantom acquisition was conducted using single F-18 point source method with activity > 5mCi/cc. A capillary tube was used to draw the selected drop, creating the point source with volume 1 mm3. Tubes filled were placed at 5 cm intervals from the center towards the PET/CT FOV's periphery. Scans were done with default settings of scout view, followed by CT axial slice scan and subsequently PET scan with 1 minute per bed position. PET data was then reconstructed with OSEM+TOF, OSEM+TOF+PSF, and OSEM+TOF+PSF+BPL (β 200,400,450, 600,800,1000 and 1200). Tangential and radial FWHM of the point source at different radial distances and locations were calculated by using Xeleris Functional Imaging Workstation Version 4.1. Radial distance effects on spatial resolution under the NCI protocol (OSEM+TOF+PSF+BPL with β450) and varying BPL β values (β 200-1200) were evaluated using fitted function slopes. Additionally, different reconstruction algorithms' impact on spatial resolution within PET/CT FOV was assessed using standard deviation calculations and plotting error bars to show spreading of the data around the mean as well as to determine the significant difference between the FWHM measured using different types of reconstruction. Result: The tangential FWHM at Location D was the only measurement that aligned with the theoretical prediction, increasing linearly with the increment of the radial distance from 1.10 mm to 1.76 mm (R2 = 0.8602). Additionally, OSEM+TOF+PSF+BPL provided greatest spatial resolution compared to OSEM+TOF and OSEM+TOF+PSF, where BPL with β200 reduced FWHM measurements closed to actual size of point source. Conclusion: OSEM+TOF+PSF+BPL (β200) reconstruction algorithms is suggested to be the optimum reconstruction algorithm that can be practiced by NCI in clinical setting due to best radial spatial resolution recovery even with increased of radial distances.
Description
Keywords
Citation