Donald Louis
Collins1, Vladimir Fonov1, Simon Duchesne2,3
1McConnell Brain Imaging Center, Montreal
Neurological Institute, Montreal, QC, Canada; 2Centre de Recherche
Universit Laval - Robert Giffard, Quebec, Canada; 3Dpt. de
Radiologie, Facult de Mdecine, Universit Laval, Quebec, Canada
A
method is presented to predict conversion from mild cognitive impairment to
Alzheimers disease using shape analysis of baseline T1w MRI data. Using 100 MCI subjects from the ADNI
database, PCA analysis of deformation fields required to register to a
minimum deformation template is used to build a shape model of the aging
brain. LDA of the eigenvalues is used to build a classifier to identify
converters and non-converters. Testing
with 100 additional MCI subjects demonstrates accuracies of 65% at 12 months
and 64% at 24 months. Adding baseline HC volume increases accuracy to 73% and
69%, respectively.