Premium
IC‐P3‐166: Correcting for partial volume effects in perfusion MRI of Alzheimer's disease
Author(s) -
Asllani Iris,
Borogovac Ajna,
Brown Truman,
Habeck Christian,
Stern Yaakov
Publication year - 2008
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2008.05.110
Subject(s) - voxel , partial volume , magnetization transfer , weighting , magnetic resonance imaging , perfusion , nuclear medicine , linear regression , pattern recognition (psychology) , artificial intelligence , atrophy , nuclear magnetic resonance , mathematics , radiology , medicine , computer science , statistics , physics
Background: Partial volume effects (PVE) due to limited spatial resolution in brain imaging can be quite substantial, particularly in studies of Alzherimer’s Disease (AD). In arterial spin labeling (ASL) MRI, PVE are exacerbated by the nonlinear dependency of the signal on voxel tissue-heterogeneity. We have developed an algorithm that corrects for PVE in ASL and applied it to ASL data from AD and healthy controls (HC) in order to separate the true, disease-related, hypoperfusion from that due to brain atrophy. Methods: 1) The algorithm is based on linear regression and estimates the pure tissue signals by modeling the voxel magnetization as a weighted sum of MGM, MWM, and MCSF contributions and the ASL difference signal as a weighted sum of MGM and MWM contributions. The weighting coefficients in both cases are the tissue’s fractional volume, which are obtained from high-resolution structural MRIs. The algorithm is based on the assumption that local magnetization and CBF are constant over a small region surrounding the voxel. 2) Subjects & Imaging ASL perfusion images were acquired on 10 patients with moderate-to-severe probable AD and 19 age-matched HC as described in a previous study (Asllani et al, JCBFM Oct24(2007)). For each subject, linear regression was performed on control and difference ASL images in subject’s native space using a 11x11x1 regression-kernel. Images were then transformed into Talairach standard space. Results: PVE correction algorithm enables the extraction of pure GM CBF that is, theoretically, uncontaminated by contribution from WM. Global GM CBF was 57 18 and 96 21 (ml/100g*min) for AD and controls, respectively. This corresponded to 63% and 55% increases in estimation of GM CBF for AD and HC, respectively, as compared to the uncorrected method (Fig.1) Voxelwise t-statistic for between-group comparisons showed 30% more voxels surviving the statistical threshold (cluster-level corrected, P 0.001) when pure GM CBF images were used as compared to the PVEuncorrected data. These additional areas were mainly in the cingulate, inferior frontal, and superior occipital gyri (Fig.2). Conclusions: Extraction of pure tissue CBF by correcting for PVE shows great promise for providing a sensitive biomarker of early detection of AD. IC-P3-167 THE DEVELOPMENT OF SMART CONTRAST AGENTS FOR MRI IMAGING OF ALZHEIMER’S DISEASE