Premium
IC‐P‐078: Resting‐state fMRI activity profile in prodromal Alzheimer's disease and older adults with cognitive complaints
Author(s) -
Wang Yang,
West John,
Magee Tamiko,
McDonald Brenna,
Risacher Shan,
Farlow Martin,
O'Neill Darren,
Saykin Andrew
Publication year - 2011
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.2011.05.144
Subject(s) - resting state fmri , posterior cingulate , atrophy , neuropsychology , voxel , medicine , audiology , hippocampus , cognitive impairment , cognitive decline , dementia , psychology , cognition , neuroscience , psychiatry , disease , radiology
AD (CDR 0.5) and 141 age-matched, cognitively normal individuals. FreeSurfer was applied to all scans to generate structural information including cortex thickness, average convexity ("sulc") and cerebral surface area distortion (Jacobian) representing morphological change on a normalized surface atlas. Principal components analysis (PCA) was applied to each individual structural measure to generate principal components (PCs) that accounted for 80% variance across all participants for each measure and hemisphere. Stepwise logistical regression was used to select significant PCs to discriminate the groups using both individual measure and combined-measure approaches. Receiver Operating Characteristics (ROC) curves were generated for these classifiers to compare discrimination power. Results: Surface map analysis revealed non-overlapping patterns between cortical thickness, convexity and area distortion maps (see figure). MANOVA showed a significant group difference for each structural measure on each hemisphere (p < 0.05), except for area distortion in the right hemisphere (p 1⁄4 0.11). 1) cortical thickness in left hemisphere (LH): AUC 1⁄4 0.83; 2) cortical thickness in right hemisphere (RH): AUC 1⁄4 0.87; 3) convexity in LH: AUC 1⁄4 0.85; 4) convexity in RH: AUC 1⁄4 0.86; 5) area distortion in LH: AUC 1⁄4 0.75; 6) cortical thickness and convexity combined in RH: AUC 1⁄4 0.98; 7) cortical thickness, convexity and area distortion combined in LH: AUC 1⁄4 0.99. The discriminating power of the combined-measure models (6, 7) was significantly higher than the other five individual measure based models (p < 0.001). See table. Conclusions: PCA-based high-dimensional markers integrating across cortical thickness, convexity and area distortion measures resulted in improved discrimination power. The models with best discrimination of very mild AD from cognitively normal individuals involved all 3 types of measures in the left hemisphere and thickness and convexity for the right hemisphere. This also suggests that the morphological manifestations of early symptomatic AD may be more prominent in the left hemisphere.