Skip to Main Content
Clinical use of positron emission tomography (PET) for brain imaging is limited by the partial-volume effect (PVE) induced by the limited spatial resolution of most scanners. Correction for this effect is often performed using a post-reconstruction processing framework involving external information provided by an MRI acquisition. This approach has the major drawback of being very sensitive to the unavoidable MRI segmentation and PET-MRI registration mismatches. Under the assumption that PVE is better compensated when it is modeled in the reconstruction process, we developed in this work an approach based on the combined usage of a realistic system response function and of a Bayesian framework allowing the incorporation of the external information in the reconstruction process through the blurred anatomical labels method. PVE compensation performance of the proposed methodology was validated on a phantom double-isotope acquisition, in comparison with the post-reconstruction correction method of the geometric transfer matrix (GTM). A Monte Carlo simulation of a realistic brain PET study allowed us to show the performances of our method relative to the residual mismatches mentioned above.