Abstract
319
Objectives Partial Volume Correction (PVC) addresses the loss in signal in small structures due to the limited resolution of the scanner. Methods for PVC can be divided into region-based methods like that of Rousset et al. (1998) or voxel-based ones like Lucy Richardson iterative deconvolution. Currently region based methods require a complete and accurate segmentation for all voxels within a subvolume of interest, which is very difficult to achieve within the medial temporal lobe. Iterative deconvolution methods have slow convergence and can result in Gibb’s artefacts at the borders of homogeneous regions. Consequently both methods are suboptimal for resolving reductions in glucose metabolism from atrophy in the hippocampus. We have developed PARSLR (PARtially Segmentable Lucy Richardson) : a novel method that combines region-based and voxel-based methods for PVC (Segobin et al. 2010) and have applied PARSLR to correct glucose metabolism for atrophy effects in hippocampal tissue of Alzheimer’s Disease (AD) patients.
Methods We performed volumetric T1-weighted MRI and HRRT (FWHM 2.5mm) FDG-PET in 6 elderly healthy controls (age 67.1±5.3 years; MMSE>28) and 12 early AD patients (age 65.6±6.7 years; MMSE 18.1±3.2). We segmented the left and right hippocampus using Freesurfer and applied PARSLR to motion-corrected FDG-PET images with the segmented and co-registered hippocampal volume kept as a homogeneous region.
Results Glucose metabolism in the atrophied hippocampi was further reduced by PVC with PARSLR in patients (uncorrected 18.1%±4.2, PARSLR 24.1%±5.0). Thus, metabolism in remaining tissue is impaired severely.
Conclusions PARSLR is able to cope with cases where a complete and accurate segmentation of the image volume is difficult. It demonstrates that hippocampal dysfunction is a leading event in AD and not an artefact due to partial volume effect in atrophic hippocampus