learning brain connectivity of alzheimer's disease from neuroimaging data
DESCRIPTION
Literature suggests functional brain connectivity difference between AD and normal aging. Existing functional connectivity studies have limitations: -Correlation-based methods capture only pair-wise info. -Most approaches are confirmative, not exploratory. - PowerPoint PPT PresentationTRANSCRIPT
Learning Brain Connectivity of Alzheimer's Disease from Neuroimaging Data Shui Huang1, Jing Li1, Liang Sun1, Jun Liu1, Teresa Wu1, Kewei Chen2, Adam Fleisher2, Eric Reiman2, Jieping Ye1
1: Arizona State University, 2: Banner Alzheimer’s Institute
This work was sponsored by the
NSF.
• Literature suggests functional brain connectivity difference between AD and normal aging.• Existing functional connectivity studies have limitations: -Correlation-based methods capture only pair-wise info. -Most approaches are confirmative, not exploratory. -A small number of brain regions are focused on. -Thorough comparison between AD, MCI, and normal aging with statistical significant assessment is lacking. -fMRI data are mostly used, not PET
Introduction
Sparse Inverse Covariance Estimation (SICE)
10θ
||)θ(||)θ())θlog(det(maxargθ̂ vecStr
)( 1kC )( 2kC
kX 1 2 21 )()( 21 kk CC
Objective
• Build functional brain connectivity models of AD, MCI, and normal controls using a machine learning technique, called inverse covariance, based on ADNI-PET data.• Assess statistical significance of the connectivity difference and summarize the results.
Approach & Monotone Property
Monotone PropertyLet and be the sets of all the connectivity components of with and respectively. If , then .Intuitively, if two regions are connected (either directly or indirectly) at one level of sparseness, they will be connected at all lower levels of sparseness.
Results
Small λLarge λ λ3 λ2 λ1Small λLarge λ
AD MCI NC
• AD: between-lobe connectivity weaker than within-lobe con.• AD: left-right same region connectivity much weaker.• MCI: patterns not as distinct from normal controls as AD.
Observations:
AD MCI NC AD MCI NC AD MCI NC
Strong Connectivity Mild Connectivity Weak Connectivity
•Temporal: decreased connectivity in AD, decrease not significant in MCI.• Frontal: increased connectivity in AD (compensation), increase not significant in MCI.• Parietal, occipital: no significant difference.• Parietal-occipital: increased weak/mild con. in AD.• Frontal-occipital: decreased weak/mild con. in MCI.• Left-right: decreased strong con. in AD, not MCI.
Observations: