Artificial intelligence has the potential to revolutionise healthcare all over the world. It could even replace or at least augment doctors in the future. Most importantly, it will hopefully help to alleviate the problems we are facing here in the UK, where the system simply collapses under the massive patient overload.
In my hospital, patients coming to A&E departments have to wait frequently for more than six hours. There are too many sick people going into hospitals not being treated fast enough because we don’t have enough doctors.
If we could use artificial intelligence, the computer could be the first point of contact and it would determine, which patients don’t actually need to see a doctor and which do.
People are not going to hospitals and to see doctors for fun. They are doing it, because they are concerned. And they are genuinely concerned that there is a problem. So if we could reassure people that what they are experiencing actually is not a problem that perhaps they should just go to a pharmacy, or just rest, it would prevent certain medical-seeking behaviours that are essentially burdening the system when no action is actually needed.
- A demo of a diagnostic chatbot
There are three major streams of development of artificial intelligence for healthcare. First there is image analytics. Only last month, Stanford University researchers announced that an algorithm they had created could diagnose skin cancer as accurately as a human doctor. There are many research groups and companies developing this type of technology, which essentially uses an image of a mole or a rash on your skin and based on data it had digested before determines whether it is or is not a problem. Similar technology could be used to analyse MRI or CT scans. In November last year, there was a paper published describing technology that analyses images of human eyes and identifies signs of diabetic retinopathy. This technology also proved to be as accurate as human doctors.
In future people may be able to download apps into their phones and every time they would feel concerned about something, they would just take a picture of it and consult the app.
In hospitals, we are already sending CT scans and MRI scans to Australia in the middle of the night for doctors there to interpret them. It would be even easier if we have diagnostic computers analysing the scans first and deciding whether a doctor actually needs to see it or not. That would reduce cost.
Another stream of research is big data analytics – having an artificially intelligent computer look at large datasets and identify patterns that might be associated, for example, with the patient’s prognosis.
- IBM's Watson computer has also joined the artificial intelligence healthcare revolution
As part of my PhD, I collected anonymised data from about 20 different hospitals from the past five to ten years. There were two data sets – one dataset were blood results of the patients and the other was data about what happened to the patients after they came to the hospital. I built an event-based analysis engine, which is effectively the same sort of technology that bankers use to do algorithmic trading when a computer trades stocks and shares without human intervention. By comparing the two datasets, the computer was able to recognise, which patterns in the blood samples are associated with worse outcomes for the patients. This way, the computer would determine, which people in the hospital need the most attention.
You won’t be able to learn from the computer what that is that makes a particular patient a high risk case but the doctors and nurses will be able to prioritise patients, which appear at a higher risk of complications.
The third stream of development are diagnostic chatbots. The NHS has been trialling such a chatbot, which essentially asks you question to determine what’s going on with you and whether you need to see a doctor. To be honest, it’s not very good at the moment but I believe that with the fast progress in research, we may soon see something more painless.
The advantage of having a doctor in your pocket is that you will be able to have someone proactively keeping you healthy by being able to monitor a whole bunch of variables.
Chinese internet giant Baidou is probably the leader in this field. They have a massive AI division and they released a chatbot that is artificial intelligence driven and you can open this app and tell the chatbot what medical problems you are suffering from. It processes all the information, figures out whether it needs more information, in which case it asks you more questions, and then it gives you a diagnoses or tells you what you need to do next. So far it only works in Chinese.
Artificial intelligence systems would also help us start considering data that we don’t normally look at – how much you have walked any given day, what have you eaten, how your mood has been and how that impacts all other aspects of your wellbeing.
A human doctor would never be able to process all that and the patient would probably not even remember all those details. The thing with computers is that they can remember everything.
The problem in current medical practice is that we treat symptoms; we don’t aspire to keep people healthy. And I think the advantage of having a doctor in your pocket is that you will be able to have someone proactively keeping you healthy by being able to monitor a whole bunch of variables.
Most of us can’t afford a yoga instructor every day, a masseuse, a cook who cooks you perfect balanced meals, someone who walks next to you and doesn’t let you smoke and makes sure you exercise. These virtual agents will help democratise this sort of support so that everyone has access to it to a degree. That is especially important in third world countries, where people frequently don’t have access to any medical care at all.