Bridge to real data: Empirical multiple material calibration for learning-based material decomposition

In this study, we proposed an empirical multi-material calibration pipeline for learning-based material decomposition. We used realistic short scan CT data from a general metric phantom using a Siemens C-arm system, and built the corresponding numeric phantom data in a software framework. After that we applied registration approaches for matching the simulated data to the acquired data, which generates prior knowledge for the following material decomposition process, as well as the ground truth for quantitative evaluations. According to the preliminary decomposition results, we successfully decomposed the inserted phantom plugs of different materials using learning-based material decomposition process, which indicates that the proposed approach is valid for learning-based material decomposition.