|
KMID : 1144120140040040362
|
|
Biomedical Engineering Letters
2014 Volume.4 No. 4 p.362 ~ p.369
|
|
|
A strategy to reduce blocky pattern and contrast loss in emission tomography reconstruction with reduced angular sampling and total variation minimization
|
|
Son Jung-Ah
Kim Soo-Mee
Lee Jae-Sung
|
|
|
Abstract
|
|
|
Purpose: Reducing radiation dose or scanning time is important for patient safety when using nuclear medicine technique. The aim of this study is to develop a reconstruction method to suppress deterioration of image quality with only a small sampling number of projection data in single photon emission computed tomography (SPECT).
Methods: We used total variation (TV) image reconstruction. However, the images reconstructed using expectation maximization (EM)-TV show systematic loss of contrast and blocky artifacts when the measurement data is noisy like SPECT data. Therefore, we first reduced projection data noise using a combination of diffusion filters and then reconstructed images from noise-reduced sinogram. Perona-Malik (PM) anisotropic diffusion filter and the nonlinear geometric diffusion filter were combined.
Results: For both the 3D Torso phantom and the NEMA IEC phantom, systematic contrast loss was seen when the images were reconstructed using EM-TV. In the Torso phantom, the combined filter showed similar normalized mean square error, streak indicator, and beta values to PM filter, but the signal-to-noise ratio gain of the image was the highest using the combined filter. In the NEMA phantom, background variability was considerably reduced when the combined filter was applied as pre-filters. Compared to geometric-EMTV, combined-EM-TV preserved edges well and produced a high-contrast image.
Conclusions: EM-TV with the combination of PM and geometric nonlinear diffusion filters was found to improve uniformity while maintaining contrast-to-noise ratio when angular sampling is low. The TV image reconstruction combined with the proposed diffusion filter can be beneficial for clinical SPECT imaging.
|
|
|
KEYWORD
|
|
|
Single photon emission computed tomography (SPECT), Anisotropic diffusion, Geometric nonlinear diffusion, Total variation (TV) image reconstruction
|
|
|
FullTexts / Linksout information
|
|
|
|
|
|
Listed journal information
|
|
|
|