Abstract:
The goal of this paper is to compare strategies for reducing partial volume effects by either minimizing the cause (i.e. improving resolution) or correcting the effect. Correction for resolution loss can be achieved either by modelling the resolution for use in iterative
reconstruction or by imposing constraints based on knowledge of the underlying anatomy. Approaches to partial volume correction
largely rely on knowledge of the underlying anatomy, based on well-registered high-resolution anatomical imaging modalities (CT or MRI). Corrections can be applied by considering the signal loss that results by smoothing the high-resolution modality to the same resolution as obtained in emission tomography. A physical phantom representing the central brain structures was used to evaluate the quantitative accuracy of the various strategies for either improving resolution or correcting for partial volume effects. Inclusion of
resolution in the reconstruction model improved the measured contrast for the central brain structures but still underestimated the true
object contrast ( 0.70). Use of information on the boundaries of the structures in conjunction with a smoothing prior using maximum
entropy reconstruction achieved some degree of contrast enhancement and improved the noise properties of the resulting images. Partial volume correction based on segmentation of registered anatomical images and knowledge of the reconstructed resolution permitted more
accurate quantification of the target to background ratio for individual brain structures.
