Keywords:
Artificial Intelligence, Radiation physics, CT, CT-Quantitative, Neural networks, Perception image, Physics, Screening, Cancer, Image verification, Quality assurance
Authors:
K. Boedeker, D. Shin, N. Akino, T. goto, K. Haioun, Y. Hamada
DOI:
10.26044/ecr2023/C-15391
Purpose
Lung cancer is associated with both high incidence and high mortality, resulting in an estimated 387,913 deaths throughout Europe in 20181. Earlier detection of lung cancer is key to improving survival rates. Low dose Lung Cancer Screening (LCS) with Computed Tomography (CT) has shown success in reducing mortality2-4. A new Silver filter that shifts the energy of the x-ray beam to the higher energies has been developed on a wide volume CT scanner. The Silver filter’s harder beam increases the penetrating power of the beam in larger patients and through the shoulder region. Traditionally, such beam-shaping filters reduce anatomical contrast and increase noise, resulting in lower contrast-to-noise ratios (CNR) than standard beam filters with conventional Hybrid Iterative Reconstruction (HIR). Recently, Deep Learning Reconstruction (DLR) has been introduced to CT. DLR is optimized to minimize noise and preserve low contrast. The combination of a beam-shaping Silver filter and DLR has the potential to deliver both increased penetrating power while maintaining or improving image quality for low dose Lung Cancer Screening.
The purpose of this study is to use objective physics metrics to evaluate the image quality of low dose Lung Cancer Screening CT protocols using a beam-shaping Silver filter and Deep Learning Reconstruction relative to a baseline standard Lung Cancer Screening protocol using a standard filter and conventional Hybrid Iterative Reconstruction.