Once considered rare, neurodegenerative diseases are now recognized as among the most common and devastating health problems of aging. However, the clinical symptoms and brain imaging signs on which current diagnostic criteria are based still rely heavily on subjective clinical interpretation. As a result, diagnosis is often made relatively late in the clinical course, and is often inaccurate, particularly in community settings. Clinical trials for potentially disease-modifying treatments are proliferating, but these are targeted to specific neuropathologies such as a-beta, tau, and progranulin. Early and accurate diagnosis is thus essential to move appropriate patients into clinical trials at the earliest possible stage and to start specific, disease modifying treatment as early as possible.
A precision medicine approach to neurodegenerative disease requires that all levels of the patient’s clinical system be represented, maximizing clinicians’ ability to identify interactions among key features of the genetic, genomic, metabolic, brain, and cognitive profile to correctly predict and treat disease. However, before such multilevel interactions can be understood in an individual patient, clinical researchers must first identify them in large, comprehensively characterized cohorts of patients. International efforts such as the Alzheimer’s Disease Neuroimaging Initiative (ADNI) have paved the way by gathering not only structural and functional brain images, but also genes, biological specimens, and comprehensive clinical data on large cohorts of patients. However, there is still a pressing need to build such multilevel datasets from cohorts of patients with other neurodegenerative diseases. The rapid proliferation of studies using ADNI data proves that when such a resource is made openly available to the research community, there is an exponential increase in the speed of discovery, leading more quickly to precise disease characterization and diagnosis.
While open sharing of data from carefully characterized neurodegenerative disease cohorts is key to accelerating scientific discovery, a second essential element in a precision medicine approach to these diseases is to build mechanisms that bring these scientific refinements back to the community. Currently, though a clinician might obtain a structural MRI as a standard part of their diagnostic evaluation, they often do not possess the level of expertise to accurately interpret that scan. If clinicians anywhere were able to access an automated resource that incorporated the most refined diagnostic approaches available for interpreting their patients’ MRIs, it would substantially reduce the chance of misdiagnosis and immediately translate to substantial benefit for the patient and their family.
Aim 1: We propose to create a national web-resource for automated quantitative analysis of structural brain MRIs, designed to provide diagnostic differentiation across neurodegenerative diseases. The technology already exists to perform automated analysis of regional atrophy from structural MRI scans. Voxel-based morphometry combined with automated diagnostic algorithms derived from support vector machine (SVM) learning approaches have been effective in differentiating patients with Alzheimer’s disease, progressive aphasia, and frontotemporal dementia. We propose to enhance and pipeline this process by
1) refining the SVM algorithms for all of the most common neurodegenerative diseases, including cross-validating algorithm accuracy on additional samples
2) determining and testing the quality control parameters involved in evaluating MRI scans from scanners with varying acquisition protocols, and
3) programming an online system that would accept uploads of MRIs from any center, automatically inspect the images for quality issues influencing interpretation, and generate a report providing probabilities of pattern-matching to different diseases.
Our collaborators for this project will include Kaiser Permanente, a rich source for well-characterized elderly patients with MRI and genetic data, as well as imaging/visualization specialists at Lawrence Berkeley National Labs, who have agreed to collaborate on development of algorithms for both quality control and interpretation, and the UCSF Center for Imaging of Neurodegenerative Disease, the lead site for ADNI.
Aim 2: We will create and archive a multilevel precision medicine dataset carefully describing all aspects of clinical presentation and biology in this valuable cohort of patients with various neurodegenerative diseases, which will be made available to the wider scientific community for download and analysis. In the process of constructing and validating this tool for automated MRI interpretation, we estimate we will need to expand our cohort to a total of 1,000 neurodegenerative disease patients and 500 healthy older controls. This creates a unique opportunity to establish and disseminate a dataset that describes these patients at every key biological level from gene to cognition, making it possible to approach the diagnosis and treatment of these diseases from a systems perspective.
1) Though we will already have these patients’ brain scans and clinical phenotyping, we propose to also systematically perform assays of the specimens already collected from these individuals to more fully characterize their genetic, genomic, and metabolic profiles.
2) We will fully document, quality check, deidentify, and homogenize the data so it can be effectively but securely downloaded and utilized by the scientific community.
Commenting is closed.