(Monday, 7/15/2019) 1:15 PM - 1:45 PM
Room: Exhibit Hall | Forum 9
Purpose: Low contrast detectability (LCD) is an essential performance metric of a CT scanner, but existing mechanisms of LCD measurement are inappropriate for advanced reconstruction (e.g., CNR) or time consuming (e.g., 2AFC). The purpose of this work was to develop an automated mechanism to measure LCD with a single CT scan.
Methods: We fabricated a phantom consisting of 150 small polycarbonate spheres (3mm in diameter) interspersed in acrylic pellets placed in a 9.5% saline solution, which approximates the attenuation of the acrylic pellets. The plastic container holding the saline solution was placed within a larger glass jar filled with water to increase the diameter of the phantom to 15 cm. The polycarbonate spheres function as low-contrast markers within an acrylic-saline matrix (less than 50 HU contrast). Detectability was assessed using a non-prewhitening (NPW) model observer. The model observer performance was characterized using exponentially-transformed free-response receiver operating characteristic (EFROC) curves. We used this framework to compare the ADMIRE iterative reconstruction (IR) and filtered backpropagation (FBP).
Results: We scanned at three dose levels: 10, 15, and 20 mAs. The area under the EFROC curve for FBP reconstruction was 0.36, 0.56, and 0.70 for 10, 15, and 20 mAs, respectively. With ADMIRE reconstruction the AUC increased to 0.40, 0.62, and 0.77, respectively. From this, we calculate a first order estimate of the dose efficiency improvement using iterative reconstruction as 12%. We rescanned at 15 mAs in an attempt to estimate uncertainty and found that the AUC differed by only 0.3%. We separately used numerical simulations to estimate the uncertainty as less than 3.5%.
Conclusion: Our work demonstrates the feasibility of the rapid and repeatable assessment of LCD with an easily constructed phantom for linear and newer iterative reconstruction algorithms, enabling comparisons between different scanner models, or measurement of LCD in regular quality assurance.
Keywords
Low Contrast Detectability, Phantoms, Quality Assurance
Taxonomy
Contact Email
