(Sunday, 7/12/2020) [Eastern Time (GMT-4)]
Room: AAPM ePoster Library
Purpose: Scatter is a key source of image quality degradation in Cone-Beam Computed Tomography (CBCT). Monte Carlo (MC) models incorporating the planning CT can be used to predict the scatter signal for subtraction from CBCT images before reconstruction. However, MC simulations of sufficient quality to be used directly are too computationally expensive for clinical utility. We present a Bayesian method for extrapolating scatter patterns from relatively cheap MC simulations and demonstrate effectiveness on Respiratory-Motion-Guided (RMG) sparse angle 4DCBCT clinical trial data.
Methods: One phase of a sparsely acquired RMG-4DCBCT scan comprising just twenty projections was used to benchmark our mixed Bayesian-Monte Carlo (MBMC) scatter correction model. A Uniform Scatter Correction (USC) model was separately implemented for comparison. The USC model uniformly subtracted a scatter fraction - assumed proportional to the mean of projection pixels - from each projection. The MBMC model first used MC simulations with relatively few histories to model radiation transport through the planning CT and predict the scatter signal at the detector. We developed a Bayesian inference framework for estimating true scatter patterns from the MC predictions, which were then subtracted from projections. The uncorrected, MBMC and USC data were reconstructed with the Feldkamp-Davis-Kress (FDK) algorithm, and reconstruction quality assessed based on Contrast-to-Noise Ratio (CNR).
Results: CNR for the uncorrected, USC and MBMC reconstructions was 0.94, 1.03 and 2.33, respectively. The MBMC reconstruction therefore exhibited relative increases in CNR of 2.47 and 2.27 with respect to the uncorrected and USC reconstructions, respectively.
Conclusions: Conventional MC-based scatter corrections improve CBCT image quality at high computational cost. Because RMG-4DCBCT scans acquire few projections at predictable locations in response to patient respiration, Bayesian extrapolation of noisy MC scatter patterns simulated prior to treatment may be used to correct projections for scatter and improve reconstruction image quality.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by grant #1123068 awarded through the Priority-driven Collaborative Cancer Research Scheme and funded by Cancer Australia. RO would like to acknowledge the support of a Cancer Institute of NSW Career Development Fellowship.
Keywords
Cone-beam CT, Bayesian Statistics, Monte Carlo
Taxonomy
Contact Email
