Feasibility study of block-matching and 3D filtering denoising algorithm in multi-material decomposition technique for dual-energy computed tomography

Abstract

AS a medical imaging technology, dual-energy computed tomography (DECT) has attracted attention owing to its higher accuracy in terms of material separation compared with conventional single-energy CT imaging. A multi-material decomposition (MMD) technique that can separate two or more materials from DECT images has recently been developed. Although a highly accurate material separation can be obtained when MMD technology is applied to CT images, the inevitable addition of noise to the image is a disadvantage. Thus, block-matching and 3D filtering (BM3D) denoising algorithm was modeled to evaluate its applicability to CT images of materials separated using MMD technology. The simulation results confirmed that when the BM3D denoising algorithm was applied to CT images separated from the material using MMD technology, the root mean square (RMS), structural similarity index, and coefficient of variation (COV) were improved by 92.44%, 16.44%, and 92.82%, respectively, compared to when only MMD was applied. In addition, the experimental results showed the same tendencies as the simulations, and volume fraction accuracy (VFA) along with the RMS and COV evaluation parameters showed the best results when BM3D was applied to CT images. Improved results were obtained using the BM3D denoising algorithm when applying the MMD technique to DECT images.