Evaluation of the novel 3D SPECT modelling algorithm in the STIR reconstruction framework: Simple vs. full attenuation correction

TY - CHAP

T1 - Evaluation of the novel 3D SPECT modelling algorithm in the STIR reconstruction framework

T2 - 2013 60th IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2013

AU - Marti-Fuster, Berta

AU - Erlandsson, Kjell

AU - Falcon, Carles

AU - Tsoumpas, Charalampos

AU - Livieratos, Lefteris

AU - Ros, Domenec

AU - Thielemans, Kris

PY - 2013/1/1

Y1 - 2013/1/1

N2 - A 3D SPECT system matrix modelling library has recently been incorporated into STIR - a software package for tomographic image reconstruction. The SPECT modelling accounts for the effects of attenuation and spatially variant resolution by incorporating them into the system matrix. Attenuation calculation can be done either along a single line (simple model), or along multiple lines for the various detector pixels covered by the PSF (full model). Due to practical reasons, a simple central-line approximation is often used for the attenuation modelling in SPECT. Our aim was to evaluate the effect of this approximation in the reconstructed images using STIR. A rotating SPECT system was modelled, equipped with different collimators. We generated noise-free projection data using the full model and reconstructed images using both models. We used a phantom consisting of an ellipse containing 4 spherical inserts with the same activity concentration but different attenuation coefficients. Images were reconstructed using the OS-MAP algorithm, 12 subsets and up to 100 iterations. With the simple model, the four spheres were all different in terms of intensity as well as distribution. When using the full attenuation model, all spheres appeared quite similar, independent of the attenuation. Our results show that the simple attenuation model can lead to artifacts and inaccurate quantification, while the full model adds significant accuracy and stability to the reconstructed images.

AB - A 3D SPECT system matrix modelling library has recently been incorporated into STIR - a software package for tomographic image reconstruction. The SPECT modelling accounts for the effects of attenuation and spatially variant resolution by incorporating them into the system matrix. Attenuation calculation can be done either along a single line (simple model), or along multiple lines for the various detector pixels covered by the PSF (full model). Due to practical reasons, a simple central-line approximation is often used for the attenuation modelling in SPECT. Our aim was to evaluate the effect of this approximation in the reconstructed images using STIR. A rotating SPECT system was modelled, equipped with different collimators. We generated noise-free projection data using the full model and reconstructed images using both models. We used a phantom consisting of an ellipse containing 4 spherical inserts with the same activity concentration but different attenuation coefficients. Images were reconstructed using the OS-MAP algorithm, 12 subsets and up to 100 iterations. With the simple model, the four spheres were all different in terms of intensity as well as distribution. When using the full attenuation model, all spheres appeared quite similar, independent of the attenuation. Our results show that the simple attenuation model can lead to artifacts and inaccurate quantification, while the full model adds significant accuracy and stability to the reconstructed images.

UR - http://www.scopus.com/inward/record.url?scp=84904169905&partnerID=8YFLogxK

U2 - 10.1109/NSSMIC.2013.6829258

DO - 10.1109/NSSMIC.2013.6829258

M3 - Conference paper

AN - SCOPUS:84904169905

BT - 2013 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC)

PB - IEEE

Y2 - 27 October 2013 through 2 November 2013

ER -