CT Model-Based Iterative Reconstruction: How Much Dose Can We Save?
C Dodge*, J Rong, MD Anderson Cancer Center, Houston, TXTU-C-103-9 Tuesday 10:30AM - 12:30PM Room: 103
Purpose: To determine the potential for dose reduction with adaptive statistical iterative reconstruction (ASiR) and model-based iterative reconstruction (MBIR) as compared to conventional filtered back-projection (FBP) reconstruction methods.
Methods: The low-contrast module of a Catphan 600 phantom was scanned using a GE Discovery HD750, at 120 kVp, 0.8 second rotation time, and pitch factors of 0.516, 0.984 and 1.375. The mA was selected for each of the three pitch factors to achieve CTDIvol values of 24, 18, 12, 6, 3, 2 and 1 mGy. Images were reconstructed at 2.5mm thickness with FBP; 20%, 40% and 70% ASiR; and MBIR. The 1% contrast, 15mm diameter supra-slice object and adjacent background were selected to determine the noise and CNR. To minimize statistical variations in noise, for each scan condition, images from 10 independent acquisitions were analyzed. The results were confirmed in the anthropomorphic Kyoto Kagaku abdomen phantom at 6, 3, 2 and 1 mGy.
Results: Compared to FBP, MBIR considerably reduced noise and improved CNR at all dose levels. At low doses in particular, a five-fold decrease in noise and four-fold increase in CNR were observed. ASiR also improved image quality compared to FBP across the range of dose values investigated, but to a lesser degree than MBIR. At 70% ASiR, up to a two-fold increase in CNR and 50% noise reduction were achieved. The quantitative results over 3 different pitch factors were similar. By maintaining image quality at the level of 1 mGy, dose reduction was estimated for MBIR and ASiR compared to FBP. Up to 97% dose saving was achieved with MBIR and up to 15% for ASiR (70%).
Conclusion: Both MBIR and ASiR maintained image quality at substantially reduced dose levels as compared to FBP. MBIR demonstrated the feasibility for low-dose patient imaging at sub-mGy dose levels.
Add this talk to vcal | ical | Contact Email: