AI retinal scanner can better and faster diagnose blindness
October 14, 2024 7:46 PM Subscribe
AI retinal scanner can better and faster diagnose blindness than eye specialists.
The Lions Eye Institute has won $5 million in state funding for an Australian-first invention to improve eye care in Western Australia's isolated and remote communities. (The Lions Eye Institute is a not-for-profit centre of excellence that combines an ophthalmic clinic with scientific discovery developing techniques for the prevention of blindness and the reduction of pain from blinding eye conditions.)
STARE INTO THE LION’S EYE, AND IF YOU TASTE THE CANDY
posted by Dokterrock at 9:58 PM on October 14
posted by Dokterrock at 9:58 PM on October 14
The classifier was published in the British Journal of Opthalmology. I did a PhD in ophthalmic diagnostics a few years ago; this seems in line with how things are moving.
Part of the potential issue I see with the methodology is that it is focused exclusively on diabetic retinopathy (DR). It’s unclear how it would react in the presence of other abnormalities. However, in a population where diabetes and DR are highly prevalent and of more concern than, say, macular degeneration, it may be that this is a genuine improvement in standard of care for rural communities. The top of the line diagnostic equipment is often in a weird spot where it is too expensive to throw around like a basic opthalmoscope but not broadly useful like an MRI.
Another interesting data point might be that the assessment of images is already done by non-MDs. The Duke Reading Center has specialist “readers” who interpret the images. One of my advisors is a Director there; his expertise is using ML and classical tools to analyze the images. I’m not sure what the SOP is there but there are lots of cases where the tools we developed are as good or better, or are are used to preprocess the images because the human eye/mind gets hung up on certain noise patterns.
posted by Maecenas at 10:58 PM on October 14 [7 favorites]
Part of the potential issue I see with the methodology is that it is focused exclusively on diabetic retinopathy (DR). It’s unclear how it would react in the presence of other abnormalities. However, in a population where diabetes and DR are highly prevalent and of more concern than, say, macular degeneration, it may be that this is a genuine improvement in standard of care for rural communities. The top of the line diagnostic equipment is often in a weird spot where it is too expensive to throw around like a basic opthalmoscope but not broadly useful like an MRI.
Another interesting data point might be that the assessment of images is already done by non-MDs. The Duke Reading Center has specialist “readers” who interpret the images. One of my advisors is a Director there; his expertise is using ML and classical tools to analyze the images. I’m not sure what the SOP is there but there are lots of cases where the tools we developed are as good or better, or are are used to preprocess the images because the human eye/mind gets hung up on certain noise patterns.
posted by Maecenas at 10:58 PM on October 14 [7 favorites]
In remote Aboriginal communities which can be a 12-hour drive from the nearest major teaching hospital, tech like this can be the difference between diagnosis and no diagnosis.
And the rate of poorly-controlled type 2 diabetes in remote Aboriginal communities is appallingly high, due to factors including (but not limited to)
lack of regular access to General Practitioners and Nurse Practitioners and Diabetes Educators;
Extremely high food costs, especially for fresh fruit and vegetables;
Low incomes;
High levels of emotional stress/distress (stress affects diabetes management independent of diet);
Conscious or unconscious racism by some doctors and some nurses;
High levels of problem drinking and/or alcoholism in communities.
posted by chariot pulled by cassowaries at 11:11 PM on October 14 [4 favorites]
And the rate of poorly-controlled type 2 diabetes in remote Aboriginal communities is appallingly high, due to factors including (but not limited to)
lack of regular access to General Practitioners and Nurse Practitioners and Diabetes Educators;
Extremely high food costs, especially for fresh fruit and vegetables;
Low incomes;
High levels of emotional stress/distress (stress affects diabetes management independent of diet);
Conscious or unconscious racism by some doctors and some nurses;
High levels of problem drinking and/or alcoholism in communities.
posted by chariot pulled by cassowaries at 11:11 PM on October 14 [4 favorites]
The company indentifye has been developing a commercial version of this (I have a friend who is a contractor there).
posted by ShooBoo at 10:23 AM on October 15
posted by ShooBoo at 10:23 AM on October 15
AI retinal scanner. . .
My mother? Lemme tell you about my mother.
posted by The Bellman at 11:42 AM on October 15
My mother? Lemme tell you about my mother.
posted by The Bellman at 11:42 AM on October 15
I’ve been a proponent of the CRADLE app and similar even since my son nearly lost his eyesight due to a gap in monitoring and diagnosis - we were late on an annual exam (but it may be the doctor missed it in the prior one) and by the time we picked up on his visual impairment he had lost significant time in developing binocular vision. Eye things are so important to catch early and yet vision care can be hard to access, especially in remote communities. This seems like a big win to me.
posted by warriorqueen at 11:55 AM on October 15 [1 favorite]
posted by warriorqueen at 11:55 AM on October 15 [1 favorite]
« Older 森林浴 | The Degradation Drug Newer »
posted by humbug at 7:55 PM on October 14 [1 favorite]