American Statistical Association
This talk will discuss our recent work on imaging pattern analysis, emphasizing the construction of imaging-derived indices that relate to normal and abnormal brain structure and function. Earlier work using support vector machines driven by regional image elements is first presented, and its application to classification of AD and schizophrenia from sMRI is summarized. It is shown that indices derived from this approach, and in particular the SPARE-AD index, are early predictors of conversion from normal aging to MCI and from MCI to AD. Similarly, the Brain Development Index (BDI) derived from sMRI and DTI is shown to capture the very consistent trajectory of brain development, and deviations from normal BDI trajectories indicate subtle delays in cognitive performance. Current methodological challenges are then discussed, including optimal selection of imaging features, deriving statistical significance maps from SVM classifiers, and extracting networks of functional connectivity using sparse image decompositions.
|Date:||Tuesday, April 8, 2014|
“Central Park” meeting area at the Child Study Center
One Park Avenue (between 32nd and 33rd Streets)
New York, New York