Anisotropic diffusion filtering of PET attenuation data to improve emission images

Attenuation correction in positron emission tomography (PET) is an essential part of clinical and research studies. However, correction using noisy transmission data acquired over short scan durations has been a problem as the noise is introduced into emission images. This study investigates the effect of smoothing the two-dimensional projections of the attenuation maps (mu-map) using the nonlinear anisotropic diffusion filtering method. Experiments are presented on a whole-body study to qualitatively evaluate the efficacy of the method in reducing the random noise and streak artefacts. The results show that image quality is significantly improved with minimal resolution loss. A reduction in statistical noise was quantitatively demonstrated when the same approach was applied to a cylindrical phantom dataset.