Whole body PET using overlapped 3D acquisition and weighted image summation

In whole body 3D PET, acquiring at distinct bed positions results in an axially varying sensitivity profile due to the cylindrical geometry and axial sampling pattern of the tomograph. Overlapping the end slices of bed positions can minimize the axial nonuniformity of the noise. We present an algorithm for the weighted summation of overlapped images, where the summation is performed after tomographic reconstruction. The weights are chosen to minimize the variance of the summed image, and are based on the noise equivalent counts in the images and the slice-to-slice coefficients of variation of a uniform cylinder. The optimal overlap is obtained and is 3-5 slices (9-15% system axial field of view) for ramp filters at Nyquist on the GE Advance tomograph in 3D mode. This overlap is smaller than the axial sensitivity profile suggests, due to the smoothing of the end slices in the 3D reprojection algorithm. The optimal overlap is independent of axial filter (due to all slices being smoothed equally); however, the optimal overlap increases as transverse smoothing is increased, because the smoothed noise profile is more consistent with what the axial sensitivity profile suggests. As compared to continuous bed motion, overlapping using weighted image summation requires no specialized acquisition software or reconstruction corrections.