AI-Powered Breakthrough in CT Scans: Faster, Smarter Material Imaging

Dual-energy computed tomography (DECT) uses the energy dependence of X-ray attenuation to allow material decomposition, however traditional image-domain or model-based approaches have drawbacks such as beam-hardening effects, large processing costs, or a need for manually segmented training data. By embedding the DECT spectral model into the training loss and integrating learnt priors in the material image domain, we present End-to-End Material Decomposition (E2E-DEcomp), a deep learning system that directly maps CT projection data to quantitative material images.