Room: AAPM ePoster Library
Purpose: Recently, multi-material decomposition (MMD) in dual energy computed tomography (DECT) has been studied to obtain material images for more than three basis materials. However, the MMD method is highly sensitive to noise fluctuation due to direct inversion and material triplet selection step for each pixel. In this study, we proposed an MMD framework that used for total variation denoising (TVD) with both before and after the decomposition.
Methods: The proposed MMD framework consisted of three steps: pre-decomposition, decomposition, and post-decomposition stages. In pre-decomposition stage, noise was suppressed for selecting CT pixel to the appropriate material triplet in decomposition stage. Decomposition stage was calculated using a material triplet library, selected the proper material triplet for each pixel, and then obtained volume fraction of selected basis material. The post-decomposition stage reduced the noise caused by the direct inversion. We used TVD method as noise suppression algorithm for pre- and post-decomposition stages. The proposed method was evaluated using digital phantom, Catphan700 and Tissue characterization model 467 phantoms.
Results: Compared to the MMD without TVD, the proposed method improved average volume fraction (VF) accuracies by 7.5 %, 19.0 %, 17.0 % in the digital phantom tissue characterization phantom, and Catphan phantom study, respectively.
Conclusion: We proposed MMD framework that reduced magnified noise resulted from direct inversion and material triplet selection for each pixel. This study demonstrated that proposed method can improve the decomposition accuracy by suppressing noise before and after decomposition.
Quantitative Imaging, Cone-beam CT