Art Epstein on how consumer tech will change eye care

AdminWolf

Site Administrator & Tech Lead
Staff member
Feb 24, 2001
17,083
3,541
113
School/Org
University of Michigan Medical School
City
Lake Oswego
State
OR
What does it mean when $30 aberrometers become commonplace?

From the 2018 AOA conference in Denver, we spoke with Dr. Art Epstein about the Consumer Electronics Show, and how consumer tech will change eye care going forward.

 
Thanks again to Art for sitting down with me, this is one of my favorite topics. I'm hoping to make it to CES this year (considering even skipping SECO to do it...)
 
Adam,
More impactful than abeerrometers will be OCT's in the kiosk. They are already in FDA application phase.
wait, what are those about? OCTs in a kiosk?! Seriously?
 
One big change consumer tech has made in the eye care space that I've noticed is the need for instruments to have good industrial design.

Before, instruments can (and did) look entirely functional -- or downright ugly. Their marketing value to practices was secondary, if thought about at all.

So now you get perimeters like this:



... or devices like the Daytona, which are sleek and very obviously designed to impress patients --

daytona-plus-2.png
 
Adam,
Yes and an AI company that could conceivably read those OCT scans.

I'm skeptical -- look at this other thread with the 11-year old and the mystery lesion. Even with the OCT and other diagnostic info (and case history), we still have experts disagreeing on what we're seeing.

If this is used as a screening test, i'm certain it would pick up pathology, but also lots of incidentalomas. Is this too expensive?
 
a recent study looked at AI/fundus photos for ONLY diabetic ret, and the results were really high false positives. The authors were not impressed.

AI/oct will be just as fickle if not more so. In that context, a stand alone oct is as unhelpful as a stand alone refraction.
 
Last edited:
  • Like
Reactions: John Mills
a recent study looked at AI/fundus photos for ONLY diabetic ret, and the results were really high false positives. The authors were not impressed.
I suspect the machines will be able to get it right eventually, but in the short-term, my concern about incidental findings remains high. It reminds me of the 'full CT body scan' fad of the early 2000s.
 
  • Like
Reactions: John Mills
I suspect the machines will be able to get it right eventually, but in the short-term, my concern about incidental findings remains high. It reminds me of the 'full CT body scan' fad of the early 2000s.

I disagree that "it will get it right". At best it is a vastly oversimplified "screening", no more sensitive then checking visual acuity with a paper chart at 20ft. At worst it will further mislead patients.

BTW have you heard of anybody using AI to look at CT scans?
 
BTW have you heard of anybody using AI to look at CT scans?

Yes indeed!

As the algorithms get better, it will be interesting to see when and if humans are beat at reading various tests.

I will note that it took over 20 years from the time the first computer chess program was released (1960) until one was devised that could beat a chess master (mid-1980s).

By 2009, a program running on a mobile phone became a grand master. Now competing against a computer is nearly hopeless for all but the very best chess players.

Now of course the issue with diagnostic tests is that when run in isolation, they yield imperfect pictures. (What is that old aphorism about diagnosis being 80% history?)
 
comparing chess to reading an oct is apples and oranges, and frankly I'm surprised at how long it took for AI to master chess, whereas I don't think AI will ever be able to master oct/photo.

Also, I think its a bit of a misnomer to label these algorithms as "AI", although I know its common to do so, but it isn't "AI". Either the oct/photo is similar to the one in the "AI" archive...or its not. Reminds me of how we use the term "textbook case"...and how relatively rare those actually are.

I agree isolated testing isn't very helpful which I think reveals their true motive...money (as usual)
 
I'm skeptical -- look at this other thread with the 11-year old and the mystery lesion. Even with the OCT and other diagnostic info (and case history), we still have experts disagreeing on what we're seeing.

If this is used as a screening test, i'm certain it would pick up pathology, but also lots of incidentalomas. Is this too expensive?

Adam,
The function of a screening test is to screen out those who are "highly likely to be free of a particular condition". It doesn't do as well to screen out any and all condition. I agree that a comprehensive eye examination is necessary.

However, timely visits and timely detection are two separate issues. Since society cannot afford a comprehensive examination for everybody and everybody cannot go to the doctor at one time, therefore, another method to better identify those who can be treated in a more timely fashion. I might also add that the OCT might not be the only procedure or instrument that can be "boxed-as-a-remote-sensor".
 
  • Like
Reactions: Frederick Frost
Adam,
The function of a screening test is to screen out those who are "highly likely to be free of a particular condition". It doesn't do as well to screen out any and all condition. I agree that a comprehensive eye examination is necessary.

However, timely visits and timely detection are two separate issues. Since society cannot afford a comprehensive examination for everybody and everybody cannot go to the doctor at one time, therefore, another method to better identify those who can be treated in a more timely fashion. I might also add that the OCT might not be the only procedure or instrument that can be "boxed-as-a-remote-sensor".

I presume these tests would not be offered for free and whatever the cost is the difference can’t be all that significant from an eye exam offered at a Walmart or some other low fee corporate place.

I just don’t buy into the whole lack of access argument when it comes to eye care. An OCT kiosk is not needed nor do I think many people would use it anyway.
 
Adam,
The function of a screening test is to screen out those who are "highly likely to be free of a particular condition". It doesn't do as well to screen out any and all condition. I agree that a comprehensive eye examination is necessary.

However, timely visits and timely detection are two separate issues. Since society cannot afford a comprehensive examination for everybody and everybody cannot go to the doctor at one time, therefore, another method to better identify those who can be treated in a more timely fashion. I might also add that the OCT might not be the only procedure or instrument that can be "boxed-as-a-remote-sensor".

It is very likely these kiosks will have a visual acuity test as well as an ultra wide field fundus photo as well.