Multi-modal machine learning detection and tracking of traumatic brain injury neurodegeneration and its differentiation from Alzheimer's disease

Adm Diagnostics, Inc., Northbrook IL

Investigators

Abstract

SUMMARY The proposed research focuses on the use of neuroimaging and machine learning to detect and understand the progression of neurodegeneration caused by repetitive brain trauma, and to differentiate this from that caused by Alzheimer?s disease (AD) and other dementias. The deliverables defined in this Phase I stage will support the development of a commercial software product for use in pharmaceutical clinical trials and in clinical diagnosis. Traumatic Brain Injury (TBI) has been shown to cause cognitive deterioration and other symptoms that overlap those arising with AD. TBI can also lead to the development of Chronic Traumatic Encephalopathy (CTE), which shares similarities in brain atrophy and tau accumulation with AD. There is mixed evidence as to whether TBI increases the risk or rate of developing AD, but regardless, the likelihoods of co-morbidity and misdiagnosis become high as individuals age. The relevant population includes professional sports athletes, individuals who played head contacting sports during high school or college, military veterans, and persons experiencing falls. However, the process by which TBI causes progressive damage, and the ways in which it can best be discriminated from AD, have not been determined. The aims of this grant focus on characterizing the progressive structural and pathology effects of TBI and discriminating these from AD using imaging of structure, function, and pathology, and machine learning. Classifiers will be developed using these information types alone and in combination. Innovations of this work include the use of a unique, comprehensive data set of imaging, cognitive, and other data acquired on more than 600 fighters by the Cleveland Clinic, a database of more than 10,000 well-characterized scans from persons across the spectrum of pre-symptomatic and symptomatic AD and other dementias, and a sophisticated machine learning software platform that addresses issues such as data overfitting and validation. Specific Aim 1 focuses on characterizing neurodegenerative changes that occur in fighters using volumetric (T1 weighted MRI) and white matter (Diffusion Tensor, DTI) imaging. Aim 2 will characterize tau accumulation in fighters and its relationship to and structural changes, comparing the two tracers 18F-AV1451 and 18F-FDDNP. Specific Aim 3 will determine methods to differentiate TBI related neurodegeneration from AD using structural and tau imaging. Follow on work in Phase II will include model refinement and additional validation with additional independent data, further prediction of cognitive impairment, extension to functional imaging modalities ASL and fMRI BOLD, dissociation of co-existing TBI and AD, comparisons to other forms of trauma, and development of software tools for commercialization. This work can have significant societal benefit through improved detection of TBI effects and precursors to greater damage and impairment, and accurate differentiation of AD and other dementias versus TBI effects to support optimal patient care.