A Graph-Based Integration of Multimodal Brain Imaging Data for the Detection of Early Mild Cognitive Impairment (E-MCI)

摘要: Alzheimer's disease (AD) is the most common cause of dementia in older adults. By time an individual has been diagnosed with AD, it may be too late for potential modifying therapy to strongly influence outcome. Therefore, critical develop better diagnostic tools that can recognize AD at early symptomatic and especially pre-symptomatic stages. Mild cognitive impairment (MCI), introduced describe a prodromal stage presently classified into stages (E-MCI, L-MCI) based on severity. Using graph-based semi-supervised learning (SSL) method integrate multimodal brain imaging data select valid imaging-based predictors optimizing prediction accuracy, we developed model differentiate E-MCI from healthy controls (HC) detection AD. Multimodal scans (MRI PET) 174 98 HC participants Disease Neuroimaging Initiative (ADNI) cohort were used this analysis. Mean targeted region-of-interest (ROI) values extracted structural MRI (voxel-based morphometry (VBM) FreeSurfer V5) PET (FDG Florbetapir) as features. Our results show SSL classifiers outperformed support vector machines task best performance was obtained 66.8% cross-validated AUC (area under ROC curve) when FDG datasets integrated. Valid phenotypes selected our approach included ROI temporal lobe, hippocampus, amygdala. Employing appears have substantial detecting warranting further investigation.