Dual Energy CT Reconstruction From Incomplete Data

Facing an increasing amount of terrorist attacks, all countries highly concern onpublic security issues and demand inspection equipments for dangerous articles. Dualenergy CT (DECT) system with high material discrimination capability is regarded as apowerful candidate. However, the typical pseudo DECT system uses two layers ofdetectors and the high cost becomes a bottle neck for wide commercial deployment.We propose a novel low-cost pseudo DECT system design which uses a full dataset for lower energy region and an incomplete data set on the higher energy region. Thecost of the system can be significantly reduced. However, to keep the reconstructionquality not degraded, a suitable reconstruction algorithm is critical. We develop twonew algorithms for the system in this thesis.The first algorithm is DECT reconstruction algorithm from incomplete data basedon image segmentation. It is used for detecting the objects which can be easilysegmented. Experiment results indicate that with the sample interval as1cm, usingonly20%high energy detectors, this algorithm keeps the relative error between thecomplete and incomplete cases as low as5%.The other one is DECT reconstruction algorithm from incomplete data based onCompressed Sensing. Compared to the former one, it is more robust to detect complexobjects. It restores incomplete data with ART-TV algorithm, using the lower energyreconstruction set for the initial estimation of the iteration, and ordered angles techniqueto improve the reconstruction rate and quality. Results demonstrate a close qualityreconstruction with only half of the higher energy detectors. In the meantime, theconvergence rate is speeded up to1.5magnitudes. We also successfully shorten thereconstruction time from dozens of minutes to a few seconds by using GPU.In addition, we also extend these two algorithms and applied them to low-cost trueDECT system, DECT systems for straight-line trajectory scan and few view scan.