Featured Technology: Retinal Image Classification
Pictured: Normal Eye Anatomy - "Normal
Fundus"
Credit: National Eye Institute, National
Institutes of Health
Overview
This classification technology is an inexpensive and easy way to detect early retinal disorders. GMU research associate Dr. Domenico Napoletani, GMU professor Dr. Timothy Sauer, and Chapman University Provost Dr. Daniele Struppa invented a technique that allows for retinal image data to be extracted from a patient and compared against healthy retinas from a database. These extracted images can be classified according to their health status, such as the degree of macular degeneration.
Technological Description
The algorithm generates a variety of masking functions and extraction features with several generalized matching pursuit iterations. In each iteration, a recursive process modifies several coefficients of the transformed signal with the largest absolute values according to the specific masking function. In this way, a greedy pursuit is turned into a slow, controlled, dissipation of the structure of the signal, that for some masking functions, enhances separation among classes.
Top Picture: Early age-related
macular degeneration.
Bottom: Neovascular age-related
macular degeneration.
Credit: National Eye Institute, NIH
Market Significance
Hundreds of millions of people worldwide have low vision. The National Eye Institute defines low vision as a visual impairment, not corrected by standard eyeglasses, contact lenses, medication or surgery, that interferes with the ability to perform everyday activities.
Low vision can result from a variety of diseases, disorders and injuries affecting the eye. These include age-related macular degeneration (AMD), cataract, glaucoma, diabetic retinopathy. Approximately 45% of all low vision cases are based on AMD.
Using this classification technology can aid care providers in the early detection of an eye health problems. Yet, it is not limited to the field of retinal imaging. Rather, this technology can be applied to computed tomography (CT) scans, magnetic resonance imaging (MRI), and time series of biological data, such as electroencephalogram (EEG) data. Furthermore, it can be used in non-biological applications, such as, computing data from remote sensing, such as false color images of landscape usage, and high frequency financial time series, such as currency exchange data.
If you’re interested in commercial opportunities for this technology, please contact David Yee, Patent Agent & Licensing Assistant, of George Mason University’s Office of Technology Transfer at (703) 993-3949. Mr. Yee's e-mail address is dyee@gmu.edu.