An implementation of dual energy CT scanning.

We have described a prereconstruction method for dual energy (PREDECT) analysis of CT scans. In theory, this method can (a) eliminate beam hardening and produce an accuracy comparable with monoenergetic scans and (b) provide the effective atomic number and electron density of any voxel scanned. Our implementation proves these statements and eliminates some of the objectionable noise. We constructed a phantom with a cylindrical sleeve-like compartment containing known amounts of high atomic number material simulating a removable skull. Conventional scans, with and without this beam hardener, were done of a water bath containing tubes of high electron and high atomic number material. Dual energy scans were then done for PREDECT. To increase the effective separation of the low and high energy beams by using more appropriate tube filtration, we fabricated a beam filter changer containing erbium, tungsten, aluminum, and steel. We used erbium, tungsten, and steel at high energy and aluminum, steel, and erbium at low energy for data acquisition. The reconstructions were compared visually and numerically for noise levels with the original steel only filtration. We found a decrease in noise down to approximately one-half the prior level when erbium/aluminum or tungsten/aluminum replaced the steel/steel filter. Erbium and tungsten were equally effective. Steel/erbium and steel/aluminum also significantly reduced image noise. The noise in the photoelectric (P) and Compton (C) images is negatively correlated. At any pixel, if the noise is positive in the P image, it is most probably negative in the C. Using this fact, the noise was reduced by postreconstruction processing.