Meeting Banner
Abstract #2555

A New High-Dimensional Machine Learning Approach for Identifying Alzheimer Disease from MRI Structural Images

Ramon Casanova1, Benjamin Wagner2, Christopher T. Whitlow2, Jeff D. Williamson3, Sally A. Shumaker4, Joseph A. Maldjian2, Mark A. Espeland1

1Biostatistical Sciences, Wake Forest University Baptist Medical Center, Winston-Salem, NC, United States; 2Radiology, Wake Forest University Baptist Medical Center, Winston-Salem, NC, United States; 3Geriatrics & Gerontology, Wake Forest University Baptist Medical Center, Winston-Salem, NC, United States; 4PHS, Wake Forest University Baptist Medical Center, Wisnton-Salem, NC, United States


Many classification methodologies for structural MRI (sMRI) images are based on a severe reduction of the feature space. Here we introduce a new classification method sMRI images, based on penalized logistic regression combined with a high dimensional image warping technique called ANTS that uses voxels as input features. We illustrate its performance when classifying images from a set of Alzheimer Disease Neuroimaging Initiative (ADNI) cognitive normal (CN) and Alzheimer Disease participants. Our methodology shows high levels of accuracy, sensitivity and specificity when automatically classifying sMRI images of CN subjects and AD patients.