back to article Study: AI designed to detect diabetic eye disease blinks in the real world, makes more work for doctors

A number of AI programs trained to detect diabetic eye damage struggle to perform consistently in the real world despite apparently excelling in clinical tests, say scientists in the US. Academics led by the University of Washington School of Medicine tested seven algorithms from five companies: Eyenuk and Retina-AI Health in …

  1. Chris G

    The AIs have it (or not)

    What I get from the makers responses, is that for their systems to function well, depends on humans trained to use the correct type of camera, to also be trained well enough to submit the right kind of photo so the AI is not really achieving the aim of simplifying, speeding up and making the process more efficient.

    I would argue that new AI diagnostic aids should have comparable testing and approval as that applied to drugs

    1. Anonymous Coward
      Anonymous Coward

      Re: The AIs have it (or not)

      Spot on - clinical tools with false positives have a massive detrimental effect on patients.

      30 years ago my PhD investigated the effects of noise / distortions in 'AI', or pattern recognition as it was known then. The reason being lab systems had a very good accuracy (as they'd been tuned to high heaven on the consistent data sets) which wasn't matched in the real world.

      A practical issue is labs / dev / research tends to go for the best sensors, cameras, mic's whatnot, real-world implementations go for the cheapest with differing response curves and SNR.

    2. AMBxx Silver badge
      Boffin

      Type of camera

      You'll see fewer cameras in your local optician now. They're mostly moving to the expensive OCT. Still a camera of sorts but provides far more information.

      There are also problems with comparing results between different manufacturers.

      the more things change, the more they stay the same.

  2. quartzz

    "It's better to be safe than sorry" - this phrase is being used to say "false positives" are ok. no they aren't. false positives can be just as negative as missing the real thing.

    it's like email security*. keeping the right people out is just as harmful as letting the wrong people in. no.....false positives aren't good enough. great for capitalism, but not for that diagnosis.

    *I can't let this one slide without mentioning instagram. who are currently/last few weeks locking loads of peoples accounts under the guise of "security", when it's actually under the guise of "your aren't uploading enough profit for us, so we're locking your account out". you know. pandemic. more people online. facebook wants more profit. (this is why your insta stats have gone down - cos other ppl have got locked out)

  3. Whitter
    Meh

    Anything that would increase the rate at which diabetic disease is flagged up from an eye exam would inevitably increase the workload of human reviewers. Is this result really a bad one? Difficult to tell from the article, which implies poor performance but doesn't show the data. e.g. How many true detections were made that would have been (or were) missed by eye docs? Maybe that is in the original research, just not the El Reg snippet.

    I guess it ultimately depends on just how many false detections were being made: alarm fatigue is a well-known problem in medical institutions.

  4. Anonymous Coward
    Anonymous Coward

    Image Quality

    I've had annual retinal screening checks for about 10 years now, arranged via the local GP practice, but performed by a contractor. Admittedly I am a difficult subject as invariably I blink when the flash fires and the operator will need to repeat the process several times, but sometimes only a single attempt.

    I always ask to see the images before leaving.

    Once during a regular eye test, a retinal screening was done. The optician took many images until she was satisfied of the quality of the images - many more than I have experienced in the past in the retinal screening service. When I saw the final images, it was a revelation - there was a big difference compared to what I had seen before - the blood vessels were clearly much better resolved and sharper. Back on the regular screening, the image quality was never as good,and in some instances, only a single image on each eye was performed.

    So, it is not at all surprising to hear “We believe one of the reasons for the discrepancy in performance was that Atlanta has a more stringent protocol for image quality at the time of screening,"

    I no longer attend the community retinal screening, as complications (not picked up by that service) meant I am now under the care of the hospital directly.

    1. Loyal Commenter Silver badge

      Re: Image Quality

      This is actually a pretty good example of the creeping privatisation of the NHS, and the negative effects of doing so. The motivations of private companies (in this case, the "contractor") and medical professionals are quite different.

      The contractor will almost certainly have been chosen either as the lowest quote, or as the only quote (in which case, the outsourcing has probably been put in place specifically to give work to that organisation, not because of clinical need).

      Private companies are motivated by profit, so the person doing the scans will be the cheapest available, and thus will have been trained to the minimum standard required to do the job. Spending any more money on someone more qualified, or on more training will be seen as wasted cash.

      Medical professionals, however, are motivated by a desire to try and help people. They will want to do the best job possible, to save them having to repeat the job later, and to make sure they don't miss anything and potentially get sued. The private individual doing the retinography doesn't have this worry, as it is their company that gets sued, not them personally.

      In theory, outsourcing such tasks gains an economy of scale (e.g. multiple NHS trusts using the same outsourcer). In practice, it doesn't save costs, but instead ends up with corners being cut to maximise the bottom line. In the end, this make the people who put that outsourcing in place very rich, at the expense of everyone else.

      1. Jellied Eel Silver badge

        Re: Image Quality

        This is actually a pretty good example of the creeping privatisation of the NHS, and the negative effects of doing so. The motivations of private companies (in this case, the "contractor") and medical professionals are quite different.

        Yup. Automate those tests! Isn't some Steve Jobs impersonator in a bit of trouble for trying that one? Contractor motivation in this one seems simple. Flog your systems, profit! Presumably there'll be an additional purpose in avoiding any liability when the AI fails to spot something. In which case, I guess the patient would have to try claiming compensation from the NHS, not the supplier.

        And if the AI asks me about turtles in the desert, I still reserve my right to cap that fool!*

        *I guess I should start using /sarc tags for comments like that now.

        1. Anonymous Coward
          Anonymous Coward

          Re: Image Quality

          I'm the AC who started this thread...

          About 3 or 4 times the imaging was done by the same technician/contractor - I remembered him. Let's just say he did not inspire confidence.

          1. Anonymous Coward
            Anonymous Coward

            Re: Image Quality

            Oh no you're not!

            1. Anonymous Coward
              Anonymous Coward

              Re: Image Quality

              Oh yes I am!

  5. Anonymous Coward
    Anonymous Coward

    AI trial sucessful

    AI software successfully interchanges shortage of trained ophthalmologists with shortage of trained camera and AI software operators.

    1. Loyal Commenter Silver badge

      Re: AI trial sucessful

      A quick google tells me that a trained ophthalmologist can expect a salary of up to £75k. The salary for a technical role in the NHS is roughly a third of this.

      Logic then dictates that those touting this software are going to be charging an annual licence fee of something in the order of £50k.

      Cynical, moi?

    2. Jon 37

      Re: AI trial sucessful

      The images are already taken by a lowly trained eyedrop-inserter and camera operator. Then an expert doctor will examine the image, later. So there is no change to the camera operator role, just replacing the expert doctor with an AI.

  6. Cuddles

    Mssing the point

    "Stephen Odaibo, CEO and founder of Retina-AI Health, told us in a statement he thought the researchers' conclusions were not supported by the study.

    "First, the study was a retrospective study based on heterogeneous unstructured data..."

    This and the following comments seem to be rather missing the point - the problem is specifically that the AI doesn't work in the real world where it has to face heterogenous unstructered data. If it's only useful when you already have a doctor sorting through all the input to make sure it's acceptable, you can just have the doctor do the whole job and throw out the AI. Either it it needs to work with the data that is actually presented under real consitions, or it simply doesn't work at all.

    1. Jon 37

      Re: Mssing the point

      Right, but what are "real conditions"?

      The AI model people are saying "if you want to use our AI then you MUST use images from an approved camera, and you MUST label the images you take "left eye" and "right eye". Which sounds perfectly reasonable to me.

      The "testers" are taking a bunch of unlabelled images from different cameras, throwing them at the AI, and seeing what happens. That's not how it's supposed to be used, and not how it would be used in real circumstances.

      (I mean, don't get me wrong, an AI that works with all possible cameras would be better than an AI that only works with one or two cameras. But these companies are not claiming to have solved that problem. And anyone deploying this is able to choose to use a supported camera).

      1. Cuddles

        Re: Mssing the point

        "Right, but what are "real conditions"?"

        Real conditions are the conditions that actually exist in the real world. Things get mislabled, pictures aren't always perfect professional quality, different equipment models are used in different places, and so on. You say it seems perfectly reasonable to insist that your system only performs when everything is perfect, but that's the whole point - nothing is ever perfect in the real world. If you want a system to actually be useful, it must be able to cope things not being perfect. Showing your system works with frictionless spherical cows under ideal lab conditions is great, but if it falls apart when the cows are no longer frictionless and occasionally turn out to be horses, it's not actually useful.

  7. Pascal Monett Silver badge
    Facepalm

    Oh really ? You don't say

    "This suggests that AI models may be more sensitive to image quality issues than human beings. "

    No kidding. A computer that just might be totally incapable of not taking an image input otherwise than pixel by pixel. What a surprise when said pixels have an inordinate amount of impact on the algorithm's interpretation.

    Computers do not see. They take the data they are given. Give them flawed data and the result will be wrong.

    If there are so many parameters to take into account to do retina diagnostics, then maybe the tool isn't fit for the job.

  8. Anonymous Coward
    Boffin

    Comments from a sufferer

    I have diabetic retinopathy. I go to a retina specialist every four months and exams consist of multiple photographs of multiple parts of the retina and more photographs after an intravenous injection of dye. No ophthalmologist ever noticed that I had a problem. It was only when my primary care physician diagnosed me with diabetes and referred me to a retina specialist that I found that I suffered from retinopathy. For now it's under control but eventually I will probably need monthly intraocular injections (which my specialist admits means hooking me up with eye restraints out of A Clockwork Orange).

    False positives in this case are acceptable if they move you to a specialist for an accurate diagnosis.

    Frankly, anyone with diabetes should be screened by a specialist.

    Note that I live in the US and this is covered by my Medicare and supplemental private insurance. I can't speak to NHS policies.

    1. Anonymous Coward
      Anonymous Coward

      Re: Comments from a sufferer

      I have intraocular injections - had one a week ago. Never gets easy, but it is a fairly quick procedure. In and out in 15 minutes the last couple of times. Covid-19 has actually caused the whole process to be streamlined and there is now much less waiting around in outpatients.

      Briefly, the process is:

      1. Confirm and mark with a "X" above the eye that to be treated

      2. Dilating eye drops

      3. Clean the eye - usually iodine solution

      4. Place surgical drape over the face/head, with opening just for the eye being treated

      5. Place Wire Eye Speculum to keep eyelids open

      6. Local aesthetic in to eye

      7. Patient is instructed to look down/keep looking down

      8. Injection!

      9. Cleanup

      10. Eye guard taped over the eye

  9. rh16181618190224

    I'm in UK, getting monthly injections in one eye for something similar. Its slightly unpleasant, and needs a couple of paracetamol afterwards - mostly because of the Iodine they use to disinfect the area. At the instant of injection it is numbed so I don't feel much. Last one a few days ago, took about 30 mins total then I walked home (with sunglasses on!) So don't worry too much about the clockwork orange frame. NHS has been great for this, free at point of delivery. You have my sympathy though.

    1. Anonymous Coward
      Anonymous Coward

      and needs a couple of paracetamol afterwards - mostly because of the Iodine they use to disinfect the area

      The iodine causes dehydration - they should have given you some moisturising eye drops, or buy some yourself - ensure it is a freshly opened for this purpose, don't use one you have already opened. They can be bought for as little as £1.50.

  10. rh16181618190224

    Agree about eyedrops.

  11. earl grey
    Happy

    Have hade my eyes photo'd for years

    Even before i was told i was pre-daibetic I was having pictures taken on an "optimap" (?) system and the eye doctor would look at the images before-during-after my eye exam. Insurance does not cover this at all, but i think my eyesight is worth the investment (same if i ever can get cataract surgery). No drops; no injections; just a nice sharp picture.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like