Impact of PSF Modeling On PET/CT Reconstructed Phantom Images
Presentations
PO-GePV-I-80 (Sunday, 7/10/2022) [Eastern Time (GMT-4)]
ePoster Forums
Purpose: To assess the impact of PSF modeling on the quantitative accuracy and noise in reconstructed PET/CT phantom images.
Methods: An ACR Jaszack phantom (ACR. Phantom Testing: PET (Rev 11-30-2020).) was scanned following the recommendation of the hot cylinder and background activity assuming 15 mCi of whole body dose. The 4 reconstruction methods evaluated were: Iterative (no PSF) 2i, Iterative 8i, TrueX (with PSF) 2i, and TrueX 8i, with various subset and filter combinations. The SUV accuracy was estimated by comparing the measured SUV, obtained by drawing regions of interest on hot cylinder of all sizes, with the theoretical SUV calculated by the ratio of the decay-corrected activity concentration of the hot cylinder and the activity concentration of the whole body. The standard deviation of the phantom background was used as a metric for noise. The average SUV percent error measurements (ΔSUV) in all cylinders and the standard deviation (SD) were compared between 2i and 8i.
Results: For 2i, quantitative accuracy was superior for TrueX with ΔSUV=23.31% versus ΔSUV=27.42% for Iterative. The average noise in the TrueX reconstructed images was less (SD=0.0583) than in the Iterative images (SD=0.0613). At 8i, the quantitative accuracy improved for both algorithms: TrueX ΔSUV=18.04% and Iterative ΔSUV=16.42%. The average noise increased from 2i to 8i for both reconstruction algorithms, but more significantly for Iterative (156% increase vs. 88% increase).
Conclusion: This study indicates that an algorithm including PSF is optimal for the reconstruction of PET/CT phantom images. The TrueX algorithm demonstrated better quantitative accuracy and less noise for smaller iterations compared to Iterative. The marginally superior average quantitative accuracy offered by Iterative reconstruction at higher iterations comes with the cost of a substantial increase in noise; therefore, in consideration of quantitative accuracy and noise together, reconstruction with PSF is optimal for higher iterations.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email