The onset of Alzheimer's disease (AD) is accompanied by a complex and distributed pattern of neuroanatomic change, difficult to distinguish clinically from dynamic alterations in normal aging. Extreme variations in the sulcal patterns of the human cortex have made it difficult to identify diffuse and focal variations in cortical structure in neurodegenerative disease. We report the first comprehensive 3D statistical analysis of deep sulcal structure in vivo, in both normal aging and dementia. High-resolution 3D T1-weighted fast SPGR (spoiled GRASS) MRI volumes were acquired from 10 patients diagnosed with AD (NINCDS-ARDRA criteria; age: 71.9 +/- 10.7 years) and 10 normal subjects matched for age (72.9 +/- 5.6 years), gender, educational level and handedness. Scans were digitally transformed into Talairach stereotaxic space. To determine specific patterns of cortical variation in dementia patients, 3D average and probabilistic maps of primary deep sulci were developed for both normal and AD groups. Major sulci (including supracallosal, cingulate, marginal, parieto-occipital, anterior and posterior calcarine sulci, and Sylvian fissures) were modeled as complex systems of 3D surfaces using a multi-resolution parametric mesh approach. Variations and asymmetries in their extents, curvature, area and surface complexity were evaluated. Three-dimensional maps of anatomic variability, structural asymmetry and local atrophy indicated severe regionally selective fiber loss in AD. A midsagittal area loss of 24.5% at the corpus callosum's posterior midbody (P < 0.025) matched increases in structural variability in corresponding temporo-parietal projection areas. Confidence limits on 3D cortical variation, visualized in 3D, exhibited severe increases in AD from 2 to 4 mm at the callosum to a peak SD of 19.6 mm at the posterior left Sylvian fissure. Normal Sylvian fissure asymmetries (right higher than left; P < 0.0005), mapped for the first time in three dimensions, were accentuated in AD (P < 0.0002), and were greater in AD than in controls (P < 0.05). Severe AD-related increases in 3D variability and asymmetry may reflect disease-related disruption of the commissural system connecting bilateral temporal and parietal cortical zones, regions known to be at risk of early metabolic dysfunction, perfusion deficits and selective neuronal loss in AD.