ARRS 2022 Abstracts

RETURN TO ABSTRACT LISTING


1784. Optimization of CT Reconstruction Algorithms for Precise Cardiac CT Quantifications Through Virtual Imaging Trials
Authors * Denotes Presenting Author
  1. Nick Felice *; Duke University
  2. Ehsan Abadi; Duke University
  3. Darin Clark; Duke University
  4. Cristian Badea; Duke University
  5. W Segars; Duke University
  6. Ehsan Samei; Duke University
Objective:
Cardiac CT can quantify calcification and stenosis in coronary arteries. These quantifications, however, can be impacted by the parameters of image formation and by motion artifacts. The purpose of this study was to ascertain the impact of choices pertaining to the reconstruction algorithm on the accuracy and precision of calcium scoring and stenosis quantifications and optimize those choices in view of cardiac motion.

Materials and Methods:
A virtual imaging trial provided a validated and realistic model of cardiac pathologies and clinical CT scanners. A four-dimensional anthropomorphic, computational phantom (XCAT) with varying heart rates (60, 90, and 120 beats per minute) with plaques of varying calcium and lipid compositions was imaged using DukeSim, a validated CT simulator that accounts for scanner-specific geometry, source spectrum, bowtie filter, anti-scatter grid, and energy-dependent detector response. The scans were obtained at a constant clinical technique of 100 effective mAs, 0.14 pitch, and 280 msec rotation time. The images were acquired with and without iodinated contrast agent per targeted task. Each sinogram was reconstructed using weighted filtered back-projection across different kernels to assess the optimum reconstruction choice.

Results:
Results demonstrate that reconstruction algorithms significantly affect calcium scoring and stenosis measurements, with motion influencing the results (up to18% deviation from ground truth for stenosis measurements). As expected, cardiac motion affected the visualization of the plaques; plaques were better distinguished in phases with less underlying motion. Sharper kernels were more accurate for stenosis measurements while softer kernels were more favorable for calcium scoring.

Conclusion:
We performed a realistic virtual imaging trial to comprehensively assess and optimize CT reconstruction for cardiac CT quantifications of coronary artery calcification and stenosis. The results define the conditions under which such quantifications can be trusted for follow-up care.