According to a recent study led by Anglia Ruskin University in the UK, the global blindness population is on course to triple by the year 2050. In percentage terms the numbers are improving. Global blindness declined from 0.75 percent of the population in 1990 to 0.48 percent in 2015, while the rate of moderate to severe vision impairment reduced from 3.83 percent to 2.90 percent. Unfortunately, the raw numbers tell a much different story. Researchers predict an increase in the number of blind individuals around the world from 36 million to 115 million within the next four decades—primarily due to growing senior populations worldwide. Since the majority of vision impairments are in some way age-related, as the population "grays," the raw number of people affected will similarly continue to increase globally.
Early detection and intervention are among our most effective tools to prevent blindness, particularly in the developing world where medical services and qualified professionals are limited. This article introduces a pair of recent developments: the first, EyeLogic, is a new way of testing for the early signs of diabetic retinopathy. The second is a new eyedrop that may soon take the place of injected VEGF inhibitors to treat advanced macular degeneration. Both of these breakthroughs have the added advantage of not requiring highly trained ophthalmologists to administer, which means they may be deployed in areas where these professionals are at a premium.
An EyeLogic Approach to Retinopathy
With new and improved treatments for river blindness and more available cataract surgery, diabetic retinopathy is quickly becoming one of the leading causes of preventable blindness globally. There are nearly 30 million individuals with diabetes in the US alone, and another 86 million who are prediabetic. Globally the World Health Organization estimates the number of diabetics at over 420 million, or one in every eleven adults.
At least 40 percent of people with diabetes will develop diabetic retinopathy at some point in their life—their chances increase with each passing year they have the disease. Unfortunately, here in the US only half of diagnosed diabetics will consult an eye doctor before experiencing permanent vision loss. In the developing world the numbers are even worse, due to a lack of medical resources and trained diagnosticians.
Early diagnosis can go a long way toward staving off the effects of diabetic eye disease with medication, surgery, and, in the earliest stages, diet and lifestyle changes. The exam is straightforward. The individual's retina is viewed through a slit lamp, which scans for abnormalities in the delicate capillaries that supply the tissue with blood. Certain cholesterol formations known as hard exudates may also signal the presence of the disease process, and, in later stages, retinal hemorrhages.
At some institutions, such as the VA, diabetes patients have their retinas routinely photographed, and these images are forwarded to remote specialists for screening. The images are also banked for training and research. Recently, a team of researchers led by Theodore Leng, MD, Assistant Professor of Ophthalmology at the Stanford University Medical Center, used this extensive database to develop EyeLogic: a new and less resource-intensive way to screen for and diagnose this vision stealing disease.
You may already be familiar with the way Google and others have improved language translation by developing algorithms and then feeding them massive amounts of data to sort, match, and perform the "deep learning" which makes it possible for you to obtain nearly instant translations of webpages, text and even speech. To put it very simply, Dr. Leng and his team have done the same with retinal images. "After creating our algorithms we fed the software over 80,000 retinal images, ranging from healthy to severely damaged," Leng explains. "The images were marked with the appropriate level of disease, if any was present, and we then employed a method of computer learning using a convolutional neural network until the software could detect the difference."
The next step was to test the software using a different set of retinal images from a different collection. "This time we didn't include the diagnoses. We allowed the trained machine to perform the screening," says Leng.
The results? "The algorithm had a 97 percent success rate, which matched or exceeded the average rate of disease detection by a retinal specialist," says Leng. "The false negatives, which would encourage individuals to not seek treatment, were also equally minimal."
Google has also performed some "machine learning" on retina imaging, but, Leng explains, "in addition to identifying diseased eyes, our algorithm marked the pathologic regions on the images to aid in identifying the areas of concern. They also indicate a severity level from zero to four." Level zero indicates no presence of retinopathy. Levels one through four indicate mild, moderate, severe, or proliferative levels, respectively.
Since the images were taken using differing resolutions, light levels, and pixel sizes, considerable pre-processing is required to standardize the images for screening. Nonetheless Leng and his team have been able to load the entire EyeLogic package onto a desktop computer, and even an iPhone, using a $300 attachment that can image retinae. The plan is to have the database stored on secure and HIPAA-compliant EyeLogic servers. After providing an image, the software only takes three or four seconds to process the image and highlight any possible pathologies.
Leng and his team hope to have a pilot program up and running soon. "We'll start with systems where they are already forwarding their images to specialists for screening," he explains. "We'll screen a duplicate copy, judge how well the system works, and add even more images to our database."
According to Leng, FDA approval may be years off. "The agency has not yet approved any algorithm-based diagnostic tool for eye care, so we will need to provide considerable study data," he says.
The team hopes to soon begin rolling out screening systems in developing countries where there is a lack of facilities and vision experts. "Eventually we would like to see EyeLogic in the offices of every physician who treats diabetes," he says. "If the doctor can not only tell you that you have the early stages of retinopathy, but actually show you the affected area on the scan, we feel a lot more people will seek the early treatment which is critical."
After getting the EyeLogic system up and running Leng says the technology also holds promise for early detection of age-related macular degeneration. In the meantime…
Eyedrops to Treat AMD
Age-related macular degeneration (AMD) is the leading cause of blindness in people age 65 and older. Currently, an estimated 13 million people in the US suffer from AMD, and the problem is expected to grow as the population ages. One large study found that people in middle age have about a 2 percent risk of developing AMD. By age 75 this risk increases to nearly 30 percent.
A new class of medications, known as vascular endothelial growth factor (VEGF) inhibitors, can slow or sometimes even halt the progression of the disease. VEGF is the protein that promotes blood vessel growth, also called angiogenesis. Angiogenesis is necessary for body growth and wound healing, but when it misfires in overstressed retinal cells it can lead to excessive vessel formation in the macula, causing the permanent damage known as AMD. VEGF inhibitors work by suppressing angiogenesis. Unfortunately, VEGF inhibitors are large molecules, "Much too large for an eyedrop to penetrate the cornea and other outer structures of the eye," says biochemist Felicity de Cogan, Ph.D., Research Fellow at the Institute of Microbiology and Infection, University of Birmingham, and leader of their Ocular Drug Delivery Team. "Between the tear film and the cornea and the vitreous, basically, the eye is designed to keep things out."
The active compounds in glaucoma eyedrops tend to be small molecules that are able to penetrate the eye. However currently VEGF inhibitors can only be effective if they are injected directly into the vitreous—the fluid that fills the eyeball and keeps it inflated--where they can make contact with the damaged retina. Patients often require between 8 and 12 injections in each treated eye every year for at least three years to stem the tide of damaging vessel growth. Needless to say, the procedure is rather uncomfortable at best. "Apart from being an unpleasant procedure for patients to undergo, the injections can cause tearing and infections inside the eye, and an increased risk of blindness," de Cogan notes. "There is also an issue of calculating dosages. "With an injection you're giving the retina a large dose each month only to have it decrease over the time until the next injection. With VEGF inhibitor eyedrops the eye would be able to receive the same dose every day," says de Cogan.
In a recent laboratory breakthrough, de Cogan and her Ocular Drug Delivery Team were able to do just that—successfully administer a VEGF inhibitor into animal eyes using an eyedrop. The work began with the study of certain peptides used to battle bacteria. "One particular peptide used a positive charge to create an electrostatic interaction with the negatively charged cell wall and penetrated the bacteria," de Cogan explains. "We discovered the same peptide would briefly open a tiny pore in corneas and other eye structures. We then employed the same electrostatic effect to bind the VEGF inhibitor to the end of the peptide chain, and when the peptide opened a pore and passed through the eye tissue the VEGF inhibitor was dragged along for the ride, and subsequently released into the vitreous."
In animal testing the VEGF inhibitors were still present in the vitreous up to 72 hours after administration. More testing is currently underway, but de Cogan is cautiously optimistic. "Without the need for expensive and painful injections more people will be able to undergo treatment earlier," she says. "We will also be able to more effectively raise and lower dosages, depending on how the medication works on any individual. Since the medication may eventually be administered by optometrists instead of more highly trained specialists, the treatment will be much less expensive and more widely available around the world. Patients will also be more involved in and in control of their own treatment, which can only enhance compliance."