A Machine Learning Analysis of the Progression of Alzheimer's Disease Using the Alzheimer's Disease Neuroimaging Initiative Data Set

• Activity Number: 441
• Type: Topic Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Imaging
• Author(s): Bruno Jedynak*+
• Companies: Johns Hopkins University
• Keywords: clinical data ; Alzheimer ; disease progression ; dimension reduction ; longitudinal data

Abstract

Alzheimer's disease (AD) is a dreadful neurodegenerative disease. The Alzheimer's disease neuroimaging initiative (ADNI-I-II-GO) is a publicly available clinical dataset including subjects diagnosed with AD dementia, mild cognitive impairment (MCI), and normal controls. In ADNI, hundreds of measurements, including clinical, cognitive, biochemical, genetic and imaging are available at baseline and longitudinally (M~10 visits per subject) for more than N= 900 subjects. The AD trajectory of a subject is then a curve in the Euclidian space of dimension K (number of measurements) out of which M points are partially observed. The ADNI dataset provides N such curves. Since the subjects in the ADNI dataset are developing the same disease, we hypothesize that a few continuous latent variables explain the collection of measurements observed at successive visits, modulo certain individual characteristics. We will present experiments in modeling which address this hypothesis and reveal low dimensional representations of the progression of AD