Simultaneous segmentation and reconstruction for dual-energy CT: Experimental results

Dualor multi-energy computed tomography systems that can provide improved material identification are of great interest in airport luggage screening. While dual-energy luggage scanning systems are commercially available, dual-energy reconstruction remains challenging because data sensitivity varies greatly depending on the material parameter being estimated, and because prior information about materials of interest is imprecise. Here we extend earlier work [1] that uses a variational framework to perform simultaneous segmentation and reconstruction focusing on materials of interest, by introducing novel regularization terms and by developing a more computationally tractable variation of the previous approach. Experimental results with realistic luggage show that material property estimates can be stabilized by estimating foreground values, which can be expected to make material identification more robust.