Potential of quantitative X‐ray imaging from photon‐counting CT: A novel analysis based on effective atomic number and physical density

Abstract Background Conventional CT assessment of lung lesions is based mainly on morphological features. CT values represent relative X‐ray attenuation that does not directly reflect tissue composition and/or physical density. Quantitative imaging using photon‐counting CT (PC‐CT) has the potential to provide quantitative parameters, such as effective atomic number (Z eff ) and effective physical density (ρ eff ). Purpose The purpose is to develop and demonstrate the potential of an algorithm that generates both the Z eff and ρ eff images, derived directly from virtual monoenergetic images (VMIs) acquired with a clinical PC‐CT system. Methods The ρ eff image was generated by fitting a database of mass attenuation coefficients (μ/ρ) to the measured linear attenuation coefficients (μ), which were obtained from two VMIs at 70 and 100 keV. As an effective material, Z eff information was also determined and fed back into the ρ eff calculation. An in‐house low‐density phantom was scanned for quantitative evaluation of ρ eff . In addition, to demonstrate the applicability of our procedure to actual clinical imaging, representative chest PC‐CT images from patients were analyzed. Results In experiments using low‐density phantoms, the CT values did not necessarily correlate with Z eff values, but they showed a very good correlation with the ρ eff values. The ρ eff values calculated using our procedure correlated well with the correct values of ρ. A relative root mean square error of ρ eff calculation was 5.5%. Furthermore, we found that in pulmonary diagnosis, image contrast information from ρ eff makes it easier to identify and distinguish lesions compared to that from Z eff . Conclusions The proposed procedure directly generated Z eff and ρ eff images from PC‐CT data and enabled quantitative interpretation of CT value in terms of elemental composition and physical density. In this study, the feasibility of generating Z eff and ρ eff images was demonstrated.