(Sunday, 7/12/2020) [Eastern Time (GMT-4)]
Room: AAPM ePoster Library
Purpose: This study aims to quantify and compare the noise between a statistical-based hybrid iterative reconstruction algorithm, AIDR 3D, and a new model-based iterative reconstruction algorithm, FIRST, for dose levels and processing kernels appropriate for three clinical exams: lung screening, abdomen, and brain imaging.
Methods: Scans of the ACR phantom have been acquired on the Canon Aquilion Genesis CT scanner with CTDIVol of 2.8 mGy, 19.7 mGy, and 29.4 mGy (reported in the 32 cm phantom). These values are clinically appropriate for lung screening, abdomen, and head studies respectively (with the 29.4 mGy scan equivalent to ~70 mGy when reported with the 16 cm phantom). Ten scans were performed at each dose level. Each scan was reconstructed using six reconstruction
methods: AIDR 3D with kernels FC18, FC56, and FC64 and FIRST with Body, Lung, and Brain settings. Eighteen NPS curves were calculated using ten slices from each scan for a total of 100 images per dose level and reconstruction method. Area under the curve (AUC) and the peak of the NPS were reported for each curve. The peak frequency was also recorded to demonstrate change in noise texture. A one-tailed paired T-test was performed to test significance of differences in AUC and peak frequency between reconstruction algorithms.
Results: Overall, the AUC was significantly lower for images reconstructed with FIRST algorithms than with AIDR algorithms, indicating lower overall noise. The peak frequency was also significantly lower for FIRST images demonstrating a change in noise texture between the two algorithms.
Conclusion: FIRST offers a significant reduction in noise over AIDR 3D. FIRST also changes the peak frequency and hence the noise texture of the image.
Funding Support, Disclosures, and Conflict of Interest: Funded by Canon Medical Systems
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
