Author: Zijia Guo, Viktor Haase, Michael F. McNitt-Gray, Frederic Noo 👨🔬
Affiliation: Siemens Healthineers, University of Utah, David Geffen School of Medicine at UCLA 🌍
Purpose: The attenuation values in CT hold strong potential for disease diagnosis. However, they lack reliability, which has limited their use to clinical trials where variability can be controlled. Spectral data acquisition and related improvements in software and hardware are essential to achieve accurate and reproducible attenuation values. Recently, we developed a novel approach for data domain material decomposition applied to CT with energy-integrated detectors (EID), which uses an analytical energy response along with an efficient object-specific scatter subtraction. Commercial photon-counting CT has also emerged as a major commercial advance for spectral data acquisition. We assess the level of accuracy reached with these two new technologies in comparison with image-based material decomposition.
Methods: The ACR phantom was scanned at 80 and 120kV on EID CT. Image-based material decomposition and our novel data-based approach were both applied. Another instance of the ACR phantom was scanned on commercial photon-counting CT using a routine abdomen protocol at 120kV. Monochromatic images were created every 5keV, from 40keV to 140kEv. Mean attenuation values were computed within the inserts of module A.
Results: Our data domain decomposition and photon counting CT provided similarly high accuracy. The error for the “bone” material varied between -3% and 3% for the former, and between -4% and 1% for the latter. For acrylic, and polyethylene, it was [-0.37%, 0.88%] versus [-0.43%, 0.58%], and [1.73%, 3.15%] versus [2.1%, 3.15%]. Image-based decomposition showed large errors at low keV and beam hardening artifacts not seen with the other methods.
Conclusion: Advances in hardware and software approaches for spectral CT show that in a well-controlled experiment high accuracy can now be achieved for the CT attenuation values using both energy-integrated and photon counting detectors. Specifically, our (EID) data domain decomposition and photon counting showed values with +/-4% accuracy across a wide range of keV.