‘Machine Learning-Based Dual-Energy CT Parametric Mapping" accepted

We developed a novel method for generating quantitative parametric maps of effective atomic number Zeff, electron density, mean excitation energy (Ix), and stopping power from clinical Spectral CT data using machine learning. The machine learning approaches were optimized and compared to the theoretical, physics-based dual-energy method. The machine learning methods generated more accurate and precise parametric maps for the phantom and clinical patient data than those obtained using the physics-based method, especially for the low-dose data. Consequently, the proposed method can enable the use of low-dose CT techniques and can enhance prediction accuracy for various types of medical applications, e.g. material/tissue detection and treatment planning of photon and proton therapy.