Context-sensitive angular filtering of PET transmission data

Transmission data noise is a large contributor to reconstructed PET image noise, particularly in body scans. To improve its signal to noise ratio, transmission data has been smoothed along sinogram rows and also in the axial direction (i.e., in “r” and “z”), but generally not across sinogram rows. The reason for this is that angular filtering will introduce a spatially varying blur of the transmission data. The blur is greatest for features far from the center of the field of view. In cases where (1) the attenuating object is not far from the center of the field of view and/or (2) radial filtering has already degraded the transmission data resolution, some angular filtering may be performed without introducing substantial spatial variations in the transmission data. A metric derived from the size of the object and the amount of radial filtering selected is developed to determine the amount of angular filtering which is applied to the data. This enables an appropriate amount of angular filtering to be employed without operator intervention