The field of glaucoma has a vast lifetime, with assets such as Virtual Reality (VR) and Artificial Intelligence (AI) being introduced and developing more every day. Today, we are going to explore the past, present, and future of glaucoma management as it relates to AI and VR.
On June 10-11, eyecare practitioners from all over the world gathered online for Eyes On Glaucoma 2022, a two-day educational event all about glaucoma disease diagnosis, treatment, and management.
Enjoy this presentation from Patrica Fulmer, OD, FAAO, and don't forget to check out our list of future events!
Please note these videos are provided for review only.
The history of glaucoma
It’s important to know not only where the field is going, but where it came from. The history of glaucoma reaches as far back as 400 BC, derived from the Greek word “glaukos” which means shining, or gleaming. Until the 17th century, there was no definitive way to differentiate cataracts and glaucoma. Then in the 18th century, we discovered IOP elevation could cause blindness. This discovery led to the iridectomy being introduced in the 1850s. Shortly thereafter, drops were introduced as treatment options. Table 1 offers a look at the progression of those drugs over the years.
Glaucoma Drug Progression
| Year | Drug Class |
|---|---|
| 1877 | Cholinergic agonists |
| 1897 | Crystalline alkaloids |
| 1904 | Osmotic agents |
| 1948 | Andrenergic antagonists |
| 1954 | Carbonic anhydrase inhibitors |
| 1955 | Andrenergic agonists |
| 1978 | β-andrenergic inhibitors |
| 1987 | α-andrenergic agonists |
| 1995 | Carbonic anhydrase inhibitor |
| 1995 | Andrenergic agonist prodrug |
| 1996 | Prostaglandin analog |
| 2017 | Rho kinase inhibitor |
Table 1
The current numbers on glaucoma
Let’s take a look at today’s snapshot of glaucoma prevalence. Recent research shows 3 million Americans have glaucoma, half of whom are unaware of their diagnosis due to lack of education or refusing to seek medical attention. There are also calculations suggesting that there are currently 60 million undiagnosed glaucoma cases worldwide. Of those with glaucoma, blindness from the disease is 6-8 times more common in African Americans than in their Caucasian counterparts.
Technology: OCT and tonometry
Glaucoma-related technology is constantly evolving. For example, advancements have recently been developed within OCT and tonometry, two of our most commonly used diagnostic tools currently.
OCT-A allows for an angiography to be able to be performed alongside traditional OCT imaging. This way, you obtain a better interpretation of the disease and are able to diagnose more accurately.
Since tonometry needs more frequent monitoring than is often provided through office visits, home-based tests have been developed that can be managed through companies like Triggerfish and iCare Home. The latter can store up to 1,000 measurements for medical review.
Visual field testing
We are all familiar with VF (Visual Field) tests and the common displeasure they typically bring for both patient and staff. Historically, VF testing has typically been performed with the 30-2 or 24-2 platforms. Recently, there has been an added option called the 24-2c, which is based on research performed by Dr. Donald Hood that shows glaucomatous damage in the macular area at all stages of glaucoma. This newest test provides the extended ability to track these changes by adding test points within the macular bundle.
Virtual reality (VR) in visual field testing
Virtual reality became mainstream through tablets, phones, and gaming. Now, it has made its way to eyecare in the form of VR VF headsets. These units have allowed for easier access to testing for patients with disabilities and difficulties with traditional testing setups, gives clinics the ability to test in small or brightly-lit areas such as waiting rooms, and can be performed at home when needed.
By popular vote, VisuALL by OLLeyes is the most common VR option with a full suite of testing, including 30-2, 24-2, and 24-c tests, and a run time of just under 4 minutes per eye. There are 3 models of VisuALL that have different upgrades and options, such as pupillometry and EOMs, costing roughly $13,000 depending on the configuration.
M&S Technologies Smart System is another product worth considering. Costing around $11,000, it offers 30-2, 24-2, and 10-2 VF Tests. M&S does plan to add more testing in the future, so keep an eye out for those releases.
A third option is Virtual Field’s VR Headset. This version focuses more on 24-2 and Ptosis testing and offers monthly or annual payment structures. It also has added features such as progression analysis, screening tests, and threshold testing.
The benefits of virtual reality
There are so many reasons that this technology is being embraced. With their lightweight design and portability, offices can utilize clinical space and staff differently, helping your practice expand and allow for more efficiency. Many of these VR tests can also be done with the patient wearing their existing contact lenses or glasses, making the test a bit more comfortable. As an extra practice management benefit, offices can often get a tax credit from the ADA for owning these units.
While the current VR headsets do instruct patients before and during the test, there will still need to be some patient education for consistent and reliable results, especially for our older patients. To make gaze-tracking easier, patient focus is monitored by the VR unit and is no longer the staff’s responsibility. One consideration that has been reported with the use of these devices is patient claustrophobia due to not being able to see their surroundings during the test; staff should be aware of this potential complication and ready to coach patients through any associated anxiety.
The role of AI in glaucoma
Now let’s take a look at Artificial Intelligence (AI) in glaucoma. This AI is perfect for identifying at-risk patients by evaluating images as well as predicting their likelihood of blindness. It helps interpret data sets, lets us utilize any “bad” data obtained, and can be customized for each and every patient.
Where did artificial intelligence (AI) begin?
Machine-learning was first developed in 1980, but was not adopted in medicine at first as it was slower and less accurate than doctors themselves. That all changed in 2010, when a “deep learning” AI process emerged. Deep-learning AI was able to take machine-learning and adapt it to function the same as a human’s mind, specifically being modeled off of the visual cortex.
Since 2010, this AI style has been successfully integrated into many offices and has been most successful when high data outputs are available. This process has been used in other fields outside of eye care, such as predicting earthquake aftershocks, predicting strokes, or assessing cancerous lesions, and is now being implemented to help detect glaucoma and other ocular conditions such as macular degeneration.
What exactly is deep learning?
We thought you’d never ask. Based on the visual vortex model, deep learning programs discern edges and orientation of images, just as our brain does via orientation and position cells. This allows for the AI to pick up an image. In addition, there is a higher degree of spatial invariance, meaning the image can nearly be shown in a 360° view and still be recognized.
Deep learning systems then break the image into layers, similarly to the neuron-to-neuron connections in the brain. However, AI does not have to include every single layered connection that a brain would need in order to make sense of the data. Next, in a manner mirroring the cerebral cortex, deep learning algorithms perform condensation and summation to reach an “understanding” of the image.
Finally, these AI machines possess both feedforward and feedback arms, meaning output data is run back through its system, ensuring the most accurate final report.
Deep learning AI consists of three different stages:
- Training: Machine is given a sample set that the algorithm is fit to
- Validating: Dataset evaluates how well the model fits the training. Manual error correction
- Testing: Testing the dataset to determine how well the model worked
How exactly is AI used by ophthalmologists?
Surveys of ophthalmologists have shown that clinicians only moderately agree on cup-to-disc ratios and associated glaucoma risk when asked to interpret fundus photography. Today’s AI, on the other hand, is very consistent with these results, and has therefore been implemented in glaucoma management in the several ways.
AI uses in glaucoma management
- Assigning glaucoma status based on cup-to-disc ratio in fundus photos: the algorithm has a learned threshold against which each photo is compared, and ratios above that threshold are labeled as glaucomatous.
- Predicting glaucomatous vision loss associated with disc cupping in fundus photos: this method also incorporates VF tests as the program is able to “learn” patterns of vision loss and their associated disc appearances based upon a dataset of fundus photos and corresponding VF tests.
- More accurately interpreting OCT and VF data: AI is able to filter through data more efficiently and accurately than clinicians and discern patterns often missed. In addition, “bad” data can be used as algorithms are able to make sense of the outputs.
- Clinical forecasting: programs are able to interpret a patient’s data over time and predict that person’s risk of blindness, risk of progression, and optimal target IOP for best disease control.
To summarize
While there is still much to learn, develop, and overcome, Artificial intelligence and virtual reality have opened doors in glaucoma management that were never before possible. We are detecting earlier stages of the disease every day, due in large part to groundbreaking technologies such as these. Make sure to keep looking for more on AI and VR in glaucoma management–and eyecare as a whole– as this new frontier is expected to continue to grow and become integral in our everyday practice.
