2019年10月
Effect of Total Variation Regularization in Bone SPECT Reconstruction from a Small Number of Projections
2019 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2019
- ,
- ,
- ,
- ,
- ,
- ,
- 記述言語
- 掲載種別
- 研究論文(国際会議プロシーディングス)
- DOI
- 10.1109/NSS/MIC42101.2019.9059901
Bone scintigraphy is difficult to understand the anatomical structure and quantitatively evaluate functions due to two-dimensional image, especially in the region such as sternum and pelvis, while bone SPECT providing three-dimensional image is useful for them. However, the imaging time of SPECT using many projection data is long. Shortening of the SPECT imaging time is desired. The aim of this study is to apply the image reconstruction method using total variation (TV) regularization to bone SPECT, and to examine the feasibility of bone SPECT from a small number of projections. In the image reconstruction, we used the expectation maximization-TV (EM-TV) algorithm consisting of the L1 norm regularization called TV, one of the methods of compressed sensing, and the maximum likelihood-expectation maximization (ML-EM) method, which is a statistical iterative image reconstruction method. First, it was validated by numerical phantom simulation that EM-TV algorithm could reconstruct a small number of projection data successfully. Next, bone SPECT imaging with Tc-MDP was performed using clinical SPECT-CT scanner, and image reconstruction was performed with equally spaced 12 out of 72 directions as projection data of a small number, and comparison with the conventional method, ML-EM, was performed. From results of bone SPECT study, the artifact which appears on the image reconstructed by ML-EM was dramatically improved by EM-TV reconstruction. In addition, EM-TV reconstruction significantly improved the quantitative accuracy in the region such as the pelvis. In conclusion, this study suggested the feasibility of bone SPECT with a small number of projections by EM-TV image reconstruction method. 99m
- リンク情報
- ID情報
-
- DOI : 10.1109/NSS/MIC42101.2019.9059901
- SCOPUS ID : 85083554144