작성자 M.I.D
[논문] L1 minimization based EM algorithm for PET Reconstruction
본문
L1 minimization based EM algorithm for PET Reconstruction
(PET 영상 재구성방법 개선 연구)
IEICE technical report
IEICE technical report 108(385), 145-149, 2009-01-12
The Institute of Electronics, Information and Communication Engineers
Choi Seokyoon / Dept. of Medical Engineering, Korea University, Oh Jangseok / Dept. of Electronics&Information Engineering, Korea University, Kim Chang-Soo / Dept. of Radiological Science, Catholic University of Pusan, Kim Mingi/ Dept. of Electronics&Information Engineering, Korea University
In PET reconstruction, iterative (EM, MAP) methods known as more accurate system modelling than analytic reconstruction methods such as backprojection filtering and filtered backprojection. These methods are considered to reduce noise effects in reconstructed images. In this study, we suggest L1-EM minimization algorithm to iterative reconstruction and L1-EM algorithm are implemented and compared with other results. the observation vectors (sinogram data) were constructed by forward projection, then scaled up to total counts by multiplying a constant, corrupted with Poisson noise, and finally scaled back. To quantify the quality of the reconstructed images, we calculated the profile information and root mean square error (RMSE) for the reconstruction. Finally L1 minimization base EM algoritnm's result shows better performance than other methods.
(PET 영상 재구성방법 개선 연구)
IEICE technical report
IEICE technical report 108(385), 145-149, 2009-01-12
The Institute of Electronics, Information and Communication Engineers
Choi Seokyoon / Dept. of Medical Engineering, Korea University, Oh Jangseok / Dept. of Electronics&Information Engineering, Korea University, Kim Chang-Soo / Dept. of Radiological Science, Catholic University of Pusan, Kim Mingi/ Dept. of Electronics&Information Engineering, Korea University
In PET reconstruction, iterative (EM, MAP) methods known as more accurate system modelling than analytic reconstruction methods such as backprojection filtering and filtered backprojection. These methods are considered to reduce noise effects in reconstructed images. In this study, we suggest L1-EM minimization algorithm to iterative reconstruction and L1-EM algorithm are implemented and compared with other results. the observation vectors (sinogram data) were constructed by forward projection, then scaled up to total counts by multiplying a constant, corrupted with Poisson noise, and finally scaled back. To quantify the quality of the reconstructed images, we calculated the profile information and root mean square error (RMSE) for the reconstruction. Finally L1 minimization base EM algoritnm's result shows better performance than other methods.