AI predicts cardiac risk from a photo – Get down to your optician to help your heart!

One of the things they teach you at medical school is that the back of the eye is the only place in the human body where you can see blood vessels with just a lens. Learning fundoscopy is one of those things that some grasp really quickly and some struggle with. I never found it difficult and have a great story about a short case in an exam where I saw something in a patients eye, that the consultant examiner hadn’t, luckily for me the patient chirped up that I was right and I passed that section!

One of the other things it takes time to understand is the concept of risk. While we talk about risk in many contexts the most common is in predicting cardiovascular risk. We use scores based on actuarial like algorithms that take peoples’ baseline scores on a range of variables and try to predict their risk based on large cohorts. E.g. the average 55-year-old 20/day smoker with a BP of X/Y has a risk of 25% and you fall into this group. A lot of this seminal work in England has been done by Prof Julia Hippisley-Cox from Nottingham, one of my heros, and perhaps worthy of more fame than she has.

Risk scoring by its very nature is population-based but it has a lot of evidence behind it. The problem is are you the average? or is there something about you that makes you higher or lower risk. if you think about it in that group 25% will have an event and 75% won’t – if there was a way of knowing which of those groups you are in — great.

Now there have been some “invasive” tests like angiography that actually measure the coronary arteries and there are other tests like calcium scoring CT scans that give an indication of true disease state but these involve radiation and are costly.

Google researchers have published some data from an AI that has been reviewing fundoscopy pictures. They reckon they can predict risk. This is fascinating stuff as although doing fundoscopy by hand can be difficult – most opticians now have retinal cameras that take great images really easily and non-invasively with no radiation.

While for ages opticians have spotted things in peoples eyes that might have suggested high blood pressure or cholesterol now they might with AI help be able to accurately identify people who need risk reduction.

Perhaps even better – the right type of people go to opticians, it is not the young fitness mad youth its the slightly older patients. There are opticians in most locations and there is no stigma in going. I’m not sure about the socio-economic distribution of opticians and perhaps this needs looking at.