Attenuation correction in quantitative SPECT of cerebral blood flow: a Monte Carlo study

A Arlig, A Gustafsson, L Jacobsson, Michael Ljungberg, C Wikkelso

Research output: Contribution to journalArticlepeer-review


Monte Carlo simulation has been used to produce projections from a voxel-based brain phantom, simulating a 99mTc-HMPAO single photon emission computed tomography (SPECT) brain investigation. For comparison, projections free from the effects of attenuation and scattering were also simulated, giving ideal transaxial images after reconstruction. Three methods of attenuation correction were studied: (a) a pre-processing method, (b) a post-processing uniform method and (c) a post-processing non-uniform method using a density map. The accuracy of these methods was estimated by comparison of the reconstructed images with the ideal images using the normalized mean square error, NMSE, and quantitative values of the regional cerebral blood flow, rCBF. A minimum NMSE was achieved for the effective linear attenuation coefficient mu(eff) = 0.07 (0.09) cm(-1) for the uniform(pre) method, the effective mass attenuation coefficient mu(eff)/rho = 0.08 (0.10) cm2 g(-1) for the uniform(post) method and mu(eff)/rho = 0.12 (0.13) cm2 g(-1) for the non-uniform(post) method. Values in parentheses represent the case of dual-window scatter correction. The non-uniform(post) method performed better, as measured by the NMSE, both with and without scatter correction. Furthermore, the non-uniform(post) method gave, on average, more accurate rCBF values. Although the difference in rCBF accuracy was small between the various methods, the same method should be used for patient studies as for the reference material.
Original languageEnglish
Pages (from-to)3847-3859
JournalPhysics in Medicine and Biology
Issue number12
Publication statusPublished - 2000

Subject classification (UKÄ)

  • Radiology, Nuclear Medicine and Medical Imaging


Dive into the research topics of 'Attenuation correction in quantitative SPECT of cerebral blood flow: a Monte Carlo study'. Together they form a unique fingerprint.

Cite this