Can Dry Macular Degeneration Be Diagnosed Through Video Games?
Michael Singer, MD and Karl Csaky, MD
For this month's issue of Retina Minute, I have the honor of interviewing Dr. Karl Csaky. Karl is the T. Boone Pickens Director/Managing Director & Chief Medical Officer, Clinical Center of Innovation for AMD/Retina Foundation of the Southwest.
1. Karl, tell me about the place you work, the Retina Foundation of the Southwest.
The Retina Foundation of the Southwest was started 35 years ago. I came here 10 years ago; prior to that, I was at Duke University. Universities have become so large that the bureaucracy has become overwhelming. The ability to push ideas, generate relationships and take advantage of opportunities has become so regulated that advancement is challenging. The Retina Foundation has about 100 employees with appropriate oversight. That is a major advantage. Smaller companies in general tend to be more mobile and more efficient.
2. How do you decide what projects to work on?
We can make our own decisions about what to work on. We have Dr. David Birch, one of the world’s experts on inherited retinal diseases. He has gathered a large cohort with various inherited retinal diseases and is working on the genetics and multiple gene therapy trials. I am, on the other hand, passionate about age-related macular degeneration.
3. Tell us about some of the research with AMD.
We really divide it into two areas. We have a clinical research unit and I am fortunate enough to have great coordinators and medical students that rotate in and out. The goal is to be involved in standard trials that other groups are involved in, but also push the envelope on our understanding of dry macular degeneration. We still haven’t figured out the riddle. What makes someone go from druse to multiple drusen? No two patients act the same way. We have been using state-of-the-art functional assessments. We have various unique tools that measure automated contrast sensitivity, as well as microperimetry units and OCTA units. We try to put this into a comprehensive model to predict progression and changes in the disease.
The other area is our laboratory research efforts that focus on using patented drug delivery devices that are designed to be placed on the surface of the eye and deliver compounds into the eye. We have chemists who are synthesizing novel compounds that fit into those devices that get into the retina from outside the eye. The hope is that these will be available to treat dry macular degeneration.
4. What therapy has the most promise?
The challenge is there are conversion steps. Some patients have drusen and there is a pathobiology that is occurring at that step. Some patients develop more drusen and develop functional changes. There is the conversion of extensive drusen to geographic atrophy. Once you develop geographic atrophy, what is it that drives the expansion? What is the major pathobiology in each step? In terms of macular degeneration, when I worked with the FDA on endpoints, the expansion of geographic atrophy became one of the accepted endpoints. Most of the clinical trial efforts in the last 10 years have been on reducing the rate of expansion. The challenge is we don’t understand what the pathobiology is, so we are still somewhat grasping at straws. In some earlier forms of macular degeneration, I think there is dysfunction at the mitochondrial level. But of course, the major challenge with treating earlier forms is how do you treat patients who still have relatively good vision with a therapy? There is appetite to use a PDS or other intraocular sustained delivery approaches for geographic atrophy. With the absence of any effective therapies, going inside the eye on a sustained delivery device does not seem reasonable to me yet. That is the challenge of intraocular devices when violating the ocular cavity in patients who still have good vision. You can use systemic injections that have challenging pharmacokinetics or use a local non-invasive drug delivery approach.
5. Tell us about more your non-invasive drug delivery approach.
The hope is we can deliver, in a manner similar to injecting an IOL, a delivery device that can be placed on the surface of the eye. Using this approach, cost and patient convenience could both be addressed. Many of the drugs that we test have been reported to have antimitochondrial effects, just not as robust as we would like. Part of our challenge is to figure out what agent we need to deliver. One advantage with sustained drug delivery is the ability to achieve a therapeutic index of your drug with time, simply by the accumulation of the agent in the tissue.
6. What is the endpoint you will be measuring?
Even in patients with good Snellen acuity, they can still have significant functional deficits. We can detect these changes with automated contrast sensitivity assessments. The field is evolving, and we are starting to have the dataset that we need to convince the FDA to expand their repertoire of endpoints that will be accepted.
We are involved with an effort in machine learning on OCT images and intermediate AMD. This approach has the potential to identify anatomic changes that we as humans cannot. The hope is this approach in concert with visual function assessments will provide supportive data and give us functional and anatomic readouts. The way we perform clinical trials in 3-5 years will be very different from the way it is done today.
7. Are there any barriers to getting the AI systems up and running?
Interestingly, we are working with Dr. Corey Clark and his team at Southern Methodist University, who have the number one gaming department in the country. We are using a gaming platform to try to focus and expedite the machine learning process, so there is not a need for a large dataset going forward. We have some pilot data to support that this could work. There is lots of evidence to support that the interaction between human input and machine learning can be the most productive. From gaming input algorithms, we can apply human intuition to speed up the process and apply it to an OCT image, which rapidly identifies the appropriate feature. For example, in the below image these are drusen transformed into a tank destroying game. If players successfully drive the tank around the hills (that represent the drusen) they get more points.
You can categorize people based on their intuitive potential – it’s fascinating. In general, the vast majority of people who play these games can intuit the goal within 4 of 5 rounds. It shows the power of the human brain to analyze images, even with no training as a retina specialist. We are living in a new world that can address all of these issues. What remains to be seen is whether these attempts will be fruitful. For anyone interested in playing the game, the website is: http://join.hewmen.com/.
Karl, your breadth of work is fascinating. The research you are doing may hopefully lead to some amazing therapies for dry macular degeneration. In addition, the work of leveraging the intuitive potential of many people through gaming to better train AI is simply brilliant! I am extremely excited to see how these areas progress over the coming years.
The landscape of our understanding of macular disease is changing, and it will require us to change the way we think about diagnosis and therapeutics. It may require not only great scientists like Dr. Csaky, but also the collective wisdom of masses of people to create the technology of the future.
About Our Editor:
Dr. Michael Singer is a board-certified Vitreoretinal Specialist and Director of Clinical Research at Medical Center Ophthalmology Associates in San Antonio, TX. He is currently the Clinical Professor of Ophthalmology at the University of Texas Health Science Center in San Antonio.