Predicting Cognitive Decline in Older Adults Through Multi-Voxel Pattern Analysis

PREDICTING COGNITIVE DECLINE IN OLDER ADULTS THROUGH MULTI-VOXEL PATTERN ANALYSIS Nathan Hantke, M.S. Marquette University, 2014 Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that is associated with cognitive and structural decline beyond what is seen in normal, healthy aging. Functional magnetic resonance imaging (fMRI) research indicates that prior to the onset of measureable cognitive impairment, individuals at-risk for AD demonstrate different patterns of neural activation than individuals at lower risk. Thus, differences in task-activated fMRI may be beneficial in predicting cognitive decline at a “presymptomatic” stage. The present study utilizes multi-voxel pattern analysis (MVPA) of baseline fMRI task-related activation to predict cognitive decline, with the hypothesis that famous and non-famous name task activation will discriminate older adults who go on to experience cognitive decline from those who do not. Ninety-nine cognitively intact older adults underwent neuropsychological testing and a semantic memory fMRI task (famous name discrimination). After follow-up neuropsychological testing 18-months later, participants were grouped as “Stable” (n = 65) or “Declining” (n = 34) based on > 1.0 SD decline in performance on cognitive measures. MVPA classification accuracy was 90% for stimulus type (famous and non-famous names), thereby supporting the general approach. Mean MVPA classification accuracy for famous and non-famous names was 83% for both the Stable and Declining groups. Finally, MVPA produced greater than chance classification accuracy of participant groups for both famous name activation (56%) and non-famous name activation (55%) as determined via binomial distribution. The results of the current study suggest that MVPA possesses potential in predicting cognitive decline in older adults.