Iterative Reconstruction for Dual Energy CT Using Accelerated Barrier Optimization Compressed Sensing (ABOCS)
X Dong*, T Niu, L Zhu, Georgia Institute of Technology, Atlanta, GATH-C-103-12 Thursday 10:30AM - 12:30PM Room: 103
On an onboard cone-beam CT (CBCT) system, the poly-energetic beam generated by current commercial x-ray tubes hardens as it penetrates the object. The beam-hardening effect results in CT image artifacts of up to 100 HU, especially around dense objects (e.g. bones). With conventional dual energy method, two basis materials images are decomposed from the data with two different x-ray spectra (e.g. with low and high tube kVp energies), and then are synthesized for a given single energy, which is free of beam-hardening artifacts. However, noise in the dual energy projections are propagated through the logarithm and decomposition processes, which results in high noise level and severe streaking artifacts in the basis material images, thus severely degraded the image quality of the synthesized.
In this work, we implement a published accelerated barrier optimization compressed sensing (ABOCS) algorithm, which minimized the data fidelity and the total variance (TV) norm of the two basis material images.
In both digital and physical head phantom, our proposed approach significantly suppresses the noise signals and the streaking artifacts. In the digital phantom, the root of mean square errors (RMSEs) of the two basis material image are reduced from 0.92 (Al image), 0.14(water image) to 0.08, 0.03.
In this work, we proposed an iterative image reconstruction method for dual energy decomposition. In the both digital and physical head phantoms, the proposed method greatly suppressed the streaking artifacts and reduced the noise level.