SPECT images suffer from poor spatial resolution, which leads to partial volume effects due to cross-talk between different anatomical regions. By utilising high-resolution structural images (CT or MRI) it is possible to compensate for these effects. Traditional partial volume correction (PVC) methods suffer from various limitations, such as correcting a single region only, returning only regional mean values, or assuming a stationary point spread function (PSF). We recently presented a novel method in which PVC was combined with the reconstruction process in order to take into account the distance dependent PSF in SPECT, which was based on filtered backprojection (FBP) reconstruction. We now present a new method based on the iterative OSEM algorithm, which has advantageous noise properties compared to FBP. We have applied this method to a series of 10 brain SPECT studies performed on healthy volunteers using the DATSCAN tracer. T1 -weighted MRI images were co-registered to the SPECT data and segmented into 33 anatomical regions. The SPECT data were reconstructed using OSEM, and PVC was applied in the projection domain at each iteration. The correction factors were calculated by forward projection of a piece-wise constant image, generated from the segmented MRI. Images were also reconstructed using FBP and standard OSEM with and without resolution recovery (RR) for comparison. The images were evaluated in terms of striatal contrast and regional variability (Coy). The mean striatal contrast obtained with OSEM, OSEM-RR and OSEM-PVC relative to FBP were 1.04, 1.42 and 1.53, respectively, and the mean striatal Coy values are 1.05, 1.53, 1.07. Both OSEM-RR and OSEM-PVC results in images with significantly higher contrast as compared to FBP or OSEM, but OSEM-PVC avoids the increased regional variability of OSEM-RR due to improved structural definition. (c) 2010 Elsevier B.V. All rights reserved.
