Click here to

Session: New Technologies and Image Reconstruction in CT and CBCT [Return to Session]

Simultaneous Analytical Image Reconstruction and Material Decomposition Denoising in Single-Scan Dual-Energy Cone-Beam CT Using Dual Energy Vectorization and Joint Bilateral Filtering

X Jiang*, C Fang, H Cui, L Zhu, Y Yang, University of Science and Technology of China, Hefei, Anhui, P.R.China


TU-F-TRACK 3-7 (Tuesday, 7/27/2021) 4:30 PM - 5:30 PM [Eastern Time (GMT-4)]

Purpose: Dual-energy cone-beam CT (DE-CBCT) with a flat-panel detector (FPD) has promising applications in various clinical tasks. Rotating-filter based DE-CBCT (RF-DECT) technique could acquire dual-energy projections within single circular rotation, enabling shorter scan time and reduced patient dose compared with two-scan method. Here we propose a new method for RF-DECT imaging which achieves analytical image reconstruction and material decomposition denoising simultaneously.

Methods: RF-CBCT employs a rotatable filter to produce low and high energy projections alternately. CT images were first separately reconstructed using filtered back-projection (FBP) method on the down-sampled projection set of each energy, then we compactly represented the reconstructed low and high energy images in a vectorized form, which was further characterized as an amplitude image and an argument image. Based on the analysis of noise dependence and image complementarity, a joint bilateral filter using amplitude image as a guide was then designed for edge-preserved smoothing of argument image. Finally, reconstructed images with reduced steaks were recovered from the amplitude image and the smoothed argument image, decomposed noise could be effectively suppressed simultaneously.

Results: On Catphan®600 phantom, the proposed method reduced streaking artifacts while preserving spatial resolution in the reconstructed images. Compared with typical two-scan results, proposed method reduced noise by a factor of 21.2, 38.0, 2.5 for soft tissue, bone tissue and electron density image, respectively, and achieved electron density accuracy within 1.5% for each material. On anthropomorphic head phantom, the proposed method successfully recovered the intricate brain structures in both high energy and electron density images.

Conclusion: Using down-sampled projections in single-scan DE-CBCT, proposed method achieves simultaneous analytical image reconstruction and material decomposition denoising. This method demonstrated superior performance on spatial resolution enhancement, decomposition noise reduction and electron density accuracy, allowing better material differentiation and dose calculation, while using only half radiation dose.



    Not Applicable / None Entered.


    Not Applicable / None Entered.

    Contact Email