Yi-Zhong Wang, Ph.D., Scientific Director, Macular Function Laboratory, is working to achieve automatic and accurate identification of defects in the structure of patients’ retinas by applying deep machine learning method. Over the past year, using a cutting-edge computer system, Dr. Wang has been training a deep convolutional neural network model to learn how to detect defects more efficient and cost-effective than the world’s experts in eye disease. Currently, there is tremendous demand for technology capable of automatically segmenting and analyzing optical coherence tomography (OCT) images, which are vital to understanding each individual’s anatomy and the progression of their eye disease over time.
The OCT is an imaging technique used to obtain detailed images from within the retina. It is effectively a noninvasive ‘optical ultrasound’, providing three-dimensional, cross-sectional images of the retina. The laser output from the OCT is low – eye-safe near-infrared light is used. The Retina Foundation of the Southwest has three machines that perform OCT tests. Every patient who comes to the Retina Foundation with an inherited eye disease or with age-related macular degeneration gets an OCT done. Currently accurate measurements of retinal defects require our researchers to assess OCT images manually, which can be very time consuming. Each image set for a single patient can take a trained researcher 30 minutes or more per eye to manually segment the layers of the retina that are captured by the OCT.
Because collecting OCT images is a routine practice at the Retina Foundation, Dr. Wang is able to use decades’ worth of images to help train his deep machine learning model. This type of artificial intelligence is highly complex and will take months, even years to complete for each eye disease. Thanks to the recently awarded three-year $300,000 Foundation Fighting Blindness grant, Dr. Wang will begin with developing the method for patients with retinitis pigmentosa (RP).
This deep machine learning method, once proven successful for RP, can be applied to help with other eye diseases such as age-related macular degeneration and Stargardt disease. Being able to use artificial intelligence to properly read and analyze OCT images will save researchers countless hours of time.