Program Information
Material Decomposition with Prior Information Compressive Sensing
A Haga1*, D Sakata2 , J Kotoku3 , T Magome4 , T Imae1, K Nawa1, K Nakagawa1, (1) University of Tokyo Hospital, Tokyo ,(2) University of Bordeaux, Bordeaux, Gradignan Cedex, (3) Teikyo University, Tokyo ,(4) Komazawa University, Tokyo
Presentations
WE-DE-605-10 (Wednesday, August 2, 2017) 10:15 AM - 12:15 PM Room: 605
Purpose: Investigation of material information in human body gives a new prospective in medical imaging. The purpose of this study is to develop a novel material decomposition algorithm based on the maximum a posteriori (MAP) reconstruction and to reveal the requirement of the material density estimation.
Methods: From the analogy of the conventional MAP reconstruction algorithm, we set a cost function composed of two terms; a log likelihood probability and a prior information probability term. In this study, the former was approximated as the penalized least square form, whereas in the latter, the compressive sensing was selected. Attenuation coefficients in each voxel were expanded by the regression formula of photoelectric and Compton cross sections with the fact that the human body is mainly made from hydrogen, carbon, nitrogen, oxygen, phosphorus, and calcium. A well-known relationship between electron density and material weight in human body was assumed, and the fitting functions were evaluated to be used as a prior information. The proposed material decomposition algorithm was assessed with virtual phantom made from above six elements.
Results: The formulation suggests that the energy spectrum of used x-ray and its variation during passing through the objects are required in the material decomposition. Clearly, multi-energy x-ray source works to decompose the material more precisely. In addition, a prior information of material density is necessary to compensate insufficient information from sinogram only. Although in the virtual phantom analysis, the uncertainty of prior information gave a corresponding noise, however, 5%-deviation of prior information yielded only a few percent error in the material density evaluation.
Conclusion: We have developed the material decomposition algorithm based on the prior information compressive sensing. The prior information for material density in human body, as well as a variation of x-ray energy spectrum during the object, is essential for the successful decomposition.
Funding Support, Disclosures, and Conflict of Interest: Funding: This work was partially supported by a Grant-in-Aid from JSPS (Japan Society for the Promotion of Science) KAKENHI JP Scientific Research (C) Grant Number 15K08691.
Contact Email: