Posted June 30, 2020

Duygu Tosun-Turgut, Ph.D., Northern California Institute for Research and Education; U.S. Department of Veterans Affairs (VA), San Francisco VA Health Care System; and University of California San Francisco

  Dr. Maiken Nedergaard, University of Rochester
Dr. Duygu Tosun-Turgut

The long-term risks of head injuries are poorly understood, with some studies suggesting severe consequences and others no consequence at all. Two major challenges exist when trying to understand the long-term nature of traumatic brain injuries (TBIs). First is the lack of large and well-characterized datasets. Large datasets reveal how TBIs exert long-term effects across a population that is truly representative of individuals who experience a TBI. Injuries also need to be well characterized in terms of severity, age of injury, and the overall health of the individual at the time of injury. The second major challenge is finding high quality, objective data like medical images or blood tests that can tell us more about the biological nature of an individual TBI. When biology and data science are combined, a precise prognosis can be made within a given population.

These two challenges each separately represent daunting tasks, and combining them into a single study would seem to be impossible. Information technologies, such as artificial intelligence, now make bringing together and analyzing large, disparate types of data feasible. Dr. Duygu Tosun-Turgut is working with a large dataset that includes magnetic resonance imaging (MRI) data of more than 1.6 million Veterans with and without a TBI. As the MRI data from these Veterans is verified and processed, it will be used to test an algorithm that can predict 5+ year risk of developing post-TBI dementia. Dr. Tosun-Turgut anticipates building a database of 200,000 patient records with as many as 12,000 structural MRIs used to inform the algorithm. The algorithm may predict a dementia or another outcome based on MRI data years before signs of cognitive decline. Understanding TBI-dementia at its earliest stages may lead to improved study designs for future clinical trials.

Subject stratification to network-based neurodegeneration sub-types of post TBI-dementia by quantifying topographic similarity (i.e., Jensen-Shannon divergence) between dementia atrophy signature and networks of interest implied in related neurodegenerative diseases including Alzheimer's disease (AD), behavioral variant frontotemporal dementia (bvFTD), semantic dementia (SD), Progressive nonfluent aphasia (PNFA), Corticobasal syndrome (CBS), Progressive supranuclear palsy (PSP).



Pubic and Technical Abstracts: Neuroimaging Endophenotypes and Predictors of Post-Traumatic Brain Injury Dementia in a Nationwide Cohort of Veterans

Top of Page