1: \begin{abstract}

2: 

3: When dealing with material classification in baggage at

4: airports, Dual-Energy Computed Tomography (DECT) allows

5: characterization of any given material with coefficients

6: based on two attenuative effects: Compton scattering and

7: photoelectric absorption.  However, straightforward

8: projection-domain decomposition methods for this

9: characterization often yield poor reconstructions due to the

10: high dynamic range of material properties encountered in an

11: actual luggage scan.  Hence, for better reconstruction

12: quality under a timing constraint, we propose a

13: splitting-based, GPU-accelerated, statistical DECT

14: reconstruction algorithm.  Compared to prior art, our main

15: contribution lies in the significant acceleration made

16: possible by separating reconstruction and decomposition

17: within an ADMM framework.  Experimental results, on both

18: synthetic and real-world baggage phantoms, demonstrate a

19: significant reduction in time required for convergence.

20: 

21: \end{abstract}

22: