Toward a Multimodal Characterization of Chronic Traumatic Encephalopathy

• Activity Number: 461
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Mental Health Statistics Section
• Author(s): Laura Mariano*+ and John Irvine and Alexander Lin and Yorghos Tripodis and Robert Stern
• Companies: Charles Stark Draper Laboratory and Charles Stark Draper Laboratory and Brigham & Women's Hospital and Boston University School of Public Health and Boston University School of Medicine
• Keywords: chronic traumatic encephalopathy ; biomarkers ; magnetic resonance spectroscopy ; multiple kernel learning ; multimodal data ; machine learning

Abstract

Chronic Traumatic Encephalopathy (CTE) is a neurodegenerative disease believed to be the result of repeated concussive and sub-concussive head traumas. Originally linked to the long-term neurological deterioration seen in former boxers, confirmed cases of CTE have more recently been observed in former professional football players, wrestlers, and soccer players. The diagnosis of CTE can only be verified upon autopsy, but efforts to quantify the manifestation of the disease in living individuals are underway. Using a dataset collected from a cohort of former professional football players, we are investigating statistical and machine-learning based approaches to identification of diagnostic and predictive biomarkers for CTE. The dataset consists of a variety of heterogeneous sources, including demographics, cognitive and behavioral assessments, genotype, and neuronal metabolite levels. We propose a novel multiple-kernel learning approach for analysis of this diverse dataset, and demonstrate the utility of this method for the identification of robust CTE biomarkers.