Program Information
Quantifying Differences in Image Quality Between Filtered Backprojection, An Adaptive Iterative Reconstruction Algorithm, and a Model-Based Iterative Reconstruction Algorithm for CT
H Whitson*, T Griglock , L DeWeese , Oregon Health & Science Univ, Portland, OR
Presentations
TU-C2-GePD-I-5 (Tuesday, August 1, 2017) 10:00 AM - 10:30 AM Room: Imaging ePoster Lounge
Purpose: In order to inform and advocate for clinical implementation of model-based iterative reconstruction (MBIR) techniques for computed tomography (CT), the complex relationships between spatial resolution, noise, and dose level must be explored and quantified. This study aimed to provide data illustrating which clinical applications would gain from implementing MBIR by comparing metrics with commonly used reconstruction algorithms (adaptive IR and filtered back-projection).
Methods: Using scanned images of the American College of Radiology CT accreditation phantom, the noise-power spectrum and task-based modulation transfer function were calculated for five dose levels on Module 3 and Module 1, respectively. Because it has been shown that spatial resolution is dependent on contrast for MBIR, we also investigated whether the use of a fat ring around the phantom changes the limiting spatial resolution. In addition, by subtracting MBIR and adaptive IR images, a qualitative description of detail loss between the algorithms was evaluated.
Results: Preliminary studies suggest both a reduction in noise and improved or maintained spatial resolution at reduced dose levels for IR techniques. Our results will help to gain further insight into the clinical trade-offs between spatial resolution and noise when using MBIR at specific dose levels. Due to the inherent non-stationary nature of MBIR, we expect these trade-offs to be non-linear and likely complex. Our results will include full evaluations of the noise-frequency properties and values for the limiting spatial resolution as well as qualitative comparisons between algorithms.
Conclusion: Although promising in theory, MBIR has experienced difficulties transitioning into clinical use. The inherent complexities of MBIR necessitate more studies in order to fully understand the clinical potential of these algorithms and to provide robust reasoning for their application. The findings of studies such as this one can provide information on balancing various image characteristics for optimal clinical applications for MBIR CT.
Funding Support, Disclosures, and Conflict of Interest: This project is currently under grant review with Toshiba America Medical Systems.
Contact Email: