Investigation of angular and axial smoothing of PET data

Radial filtering of emission and transmission data is routinely performed in PET during reconstruction in order to reduce image noise. Angular smoothing is not typically done, due to the introduction of a non-uniform resolution loss; axial filtering is also not usually performed on data acquired in 2D mode. The goal of this paper was to assess the effects of angular and axial smoothing on noise and resolution. Angular and axial smoothing was incorporated into the reconstruction process on the Scanditronix PC2048-15B brain PET scanner. In-plane spatial resolution and noise reduction were measured for different amounts of radial and angular smoothing. For radial positions away from the center of the scanner, noise reduction and degraded tangential resolution with no loss of radial resolution were seen. Near the center, no resolution loss was observed, but there was also no reduction in noise for angular filters up to a 7/spl deg/ FWHM. These results can be understood by considering the combined effects of smoothing projections across rows (angles) and then summing (back-projecting). Thus, angular smoothing is not optimal due to its anisotropic noise reduction and resolution degradation properties. However, uniform noise reduction comparable to that seen with radial filtering can be achieved with axial smoothing of transmission data. The axial results suggest that combined radial and axial transmission smoothing could lead to improved noise characteristics with more isotropic resolution degradation.