This project aims to apply novel machine learning techniques to recently developed optical imaging measurement to improve the accurate prediction and detection of glaucomatous progression. Complex functional and structural tests in daily use by eye care providers contain hidden information that is not fully used in current analyses, and advanced pattern recognition/machine learning-based analysis techniques can find and use that hidden information. We will use mathematically rigorous techniques to discover patterns of defects and to track their changes in longitudinal series of perimetric and optical imaging data from up to 1,800 patient and healthy eyes, available as the result of long-term NIH funding. We also will investigate deep learning and novel statistical techniques for this purpose. The required longitudinal measurements from several newly developed optical imaging techniques were not available to our previously funded NEI- supported work. The proposed work potentially can enhance significantly the medical and surgical treatment of glaucoma and reduce the cost of glaucoma care by informing clinical decision-making based on mathematically based, externally validated methods. Moreover, improved techniques for predicting and detecting glaucomatous progression can be used for refined subject recruitment and to define endpoints for clinical trials of intraocular pressure-lowering and neuroprotective drugs.

Public Health Relevance

The proposed project will improve machine learning techniques for predicting and detecting glaucomatous change in patient eyes tested longitudinally by visual field and optical imaging instruments and will make use of a very large amount of data, obtained using previously awarded NIH funds, to do so. This proposal addresses the current NEI Glaucoma and Optic Neuropathies Program objectives of developing improved diagnostic measures to characterize and detect optic nerve disease onset and characterize glaucomatous neuro- degeneration within the visual pathways at structural and functional levels. The development of a clinically useful novel, empirical system for predicting and detecting glaucomatous progression can have a significant impact on the future of clinical care and on the future of clinical trials designed to investigate IOP lowering and neuroprotective drugs.

Agency
National Institute of Health (NIH)
Institute
National Eye Institute (NEI)
Type
Exploratory/Developmental Grants (R21)
Project #
1R21EY027945-01
Application #
9298423
Study Section
Special Emphasis Panel (ZRG1)
Program Officer
Liberman, Ellen S
Project Start
2017-07-01
Project End
2019-06-30
Budget Start
2017-07-01
Budget End
2018-06-30
Support Year
1
Fiscal Year
2017
Total Cost
Indirect Cost
Name
University of California, San Diego
Department
Ophthalmology
Type
Schools of Medicine
DUNS #
804355790
City
La Jolla
State
CA
Country
United States
Zip Code
92093
Christopher, Mark; Belghith, Akram; Bowd, Christopher et al. (2018) Performance of Deep Learning Architectures and Transfer Learning for Detecting Glaucomatous Optic Neuropathy in Fundus Photographs. Sci Rep 8:16685
Bowd, Christopher; Zangwill, Linda M; Weinreb, Robert N et al. (2018) Racial Differences in Rate of Change of Spectral-Domain Optical Coherence Tomography-Measured Minimum Rim Width and Retinal Nerve Fiber Layer Thickness. Am J Ophthalmol 196:154-164
Ghahari, Elham; Bowd, Christopher; Zangwill, Linda M et al. (2018) Macular Vessel Density in Glaucomatous Eyes With Focal Lamina Cribrosa Defects. J Glaucoma 27:342-349