IC–P–025: Regional Network Pattern of MRI Gray Matter Volume in Alzheimer's Dementia

Gene E. Alexander,Kewei Chen,Melaney Aschenbrenner,Eric M. Reiman,Richard J. Caselli,Tricia L. Merkley,Diana J. Lewis,Pietro Pietrini,Stefan J. Teipel,Harald Hampel,Stanley I. Rapoport,James R. Moeller

DOI: https://doi.org/10.1016/j.jalz.2006.05.2230

2006-01-01

Abstract:Alzheimer's disease (AD) is characterized by reductions in brain volume measured by magnetic resonance imaging (MRI), with temporal lobe structures preferentially affected early. To investigate the regional pattern of gray matter atrophy throughout the brain in AD, we used a multivariate model of regional covariance with voxel–based morphometry (VBM) to identify networks of gray matter volume that distinguish AD from healthy aging. Ten otherwise healthy patients with probable AD [NINDS–ADRDA criteria; mean±s.d. age=65±9 years; 6M/4F; Mini–Mental State Exam (MMSE)=16.1±6.3] were compared to 15 age–matched, healthy controls [age=63±8 years; 10M/5F; MMSE=29.5±0.7] with T1–weighted volumetric MRI scans. Statistical parametric mapping (SPM2) was used with optimized VBM to produce smoothed maps of gray matter volume. We used a form of principal component analysis, the Scaled Subprofile Model (SSM) on a voxel–basis, to identify patterns of regional interactions in gray matter that distinguished the AD patients from healthy controls. SSM analysis for the two groups combined identified a linear combination of five component patterns whose subject scores were highly predictive of group membership (R2adj=0.85, p<0.000001). The combined pattern indicated that the AD patients differed from controls with less gray matter in bilateral hippocampal, lateral temporal, and parietal/precuneus regions with relative preservation in small areas that included bilateral prefrontal cortex and right cerebellum. In the AD group, memory performance was best characterized by a linear combination of four of these five component patterns (R2adj=0.66, p<0.05), with poorer memory associated with less gray matter in bilateral hippocampal and parietal/precuneus areas with relative preservation in small areas that included bilateral prefrontal and left basal ganglia regions. AD patients differ from healthy elderly in a regionally distributed network pattern of gray matter atrophy that includes brain regions known to be preferentially affected in AD. Further, the expression of such gray matter networks is associated with individual differences in cognitive dysfunction. Multivariate network analysis of MRI VBM with SSM may aid early detection, assist in tracking disease progression, and potentially help in the evaluation of treatments for AD.