Improvement with the Multi-material Decomposition Framework in Dual-energy Computed Tomography: A Phantom Study

• Published in: Journal of the Korean Physical Society Vol. 77; no. 6; pp. 515 - 523
• Main Authors: Lee, Haenghwa, Kim, Hee-Joung, Lee, Donghoon, Kim, Dohyeon, Choi, Seungyeon, Lee, Minjae
• Format: Journal Article
• Language: English
• Published: Seoul The Korean Physical Society 01.09.2020; Springer Nature B.V; 한국물리학회
• Subjects: Algorithms; Computed tomography; Decomposition; Image degradation; Image quality; Mathematical and Computational Physics; Noise; Noise reduction; Noise sensitivity; Particle and Nuclear Physics; Physics; Physics and Astronomy; Theoretical; X ray spectra; 물리학; Multi-material decomposition; Dual-energy computed tomography; Material triplet library
• ISSN: 0374-4884; 1976-8524
• DOI: 10.3938/jkps.77.515

Dual energy computed tomography (DECT) enhances tissue characterization by obtaining two or three material images from two measurements with different X-ray spectra. Recently, multi-material decomposition (MMD) in DECT has been studied to obtain decomposed material images for more than three basis materials. However, the MMD method is highly sensitive to noise fluctuation due to the direct inversion and the material triplet selection for each pixel. Although several studies have reported to reduce the noise resulting from direct inversion, no studies have researched reduction in the image quality degradation caused by material triplet selection. We proposed a MMD framework for DECT that includes pre-decomposition and post-decomposition stages to reduce image quality degradation due to material triplet selection and direct inversion. The total variation denoising method was applied to the pre-decomposition and the post-decomposition stages as a noise suppression algorithm. The digital phantom, tissue characterization phantom, and Catphan phantom were employed as test objects in this study. The volume fraction accuracy (VFA) and the standard deviation (STD) were quantitatively calculated to evaluate the quality of the decomposed images. The results of the proposed method were compared to those of the direct MMD (DMMD) and the MMD with total variation denoising (MMD-TVD) methods. Compared to the DMMD method, the proposed method improved average the VFA value by 11.40%, 17.31%, and 19.13% in the digital phantom, the tissue characterization phantom, and the Catphan phantom studies, respectively. The STD values for the proposed method are better than those of the DMMD method, and are similar to those of the MMD-TVD method. Our method successfully improved quantification accuracy and suppressed noise. In conclusion, the proposed method resulted in quantitatively better multi-material images for DECT.