Is AI just what the doctor ordered?
Artificial Intelligence is being touted as the next big thing in medicine. Nikki Macdonald asks a visiting global expert if it’s really a global game-changer.
Imagine if healthcare was one continuous, global clinical trial – your every physical and medical detail recorded to fuel the world’s research engine. Would you want to sign up?
That’s the vision for the future of Philips’ American health tech chief medical officer, Dr Roy Smythe, a former cancer surgeon who was in New Zealand to talk about health and artificial intelligence at the Health Informatics conference in Rotorua.
If the phrase AI and healthcare conjures images of robo-docs cruising hospital wards dispensing advice and supercomputers chewing through centuries of research data in minutes and magically finding the cure to cancer, forget it.
That’s not the future Smythe sees.
‘‘Artificial intelligence, in general, is over-hyped,’’ he begins.
But AI is already in use in healthcare, and its potential stretches into every nook of the health system.
Smythe divides AI into three levels of sophistication:
1) Big data analysis – finding patterns in huge data sets that humans can’t see.
2) Machine learning – teaching machines to use and create algorithms to understand data, based on what they see.
3) Artificial neural networks – enabling machines to learn to understand – and find patterns in non data-based information, such as pictures and medical scans, and to train themselves from scratch. A breakthrough example of this was Google’s AlphaGo Zero, which taught itself how to play complex Chinese strategy game Go, without any human intervention.
Some hospitals already use basic data analysis AI to predict sudden, life-threatening patient crises in hospitals. Philips’ IntelliVue Guardian uses intensive care monitoring data to provide rolling risk scores, by looking for patterns that signal a change in the patient’s status.
In the next five to 10 years, Smythe expects such real time risk prediction to extend beyond hospitals, such as monitoring heart failure patients at home, heartbeat-by-heartbeat and breath-by-breath.
Other data will be added to improve accuracy: patient details, such as ethnicity, age, diet and level of physical activity; then a patient’s genomic profile and the genetics or characteristics of their particular tumour or disease. And suddenly the world of healthcare looks like a global clinical trial of 10 billion people, with machines searching for patterns in how different strains of disease progress in different kinds of people, and respond to different kinds of treatments.
‘‘The insights that would be delivered would be almost unimaginable – the accuracy.
‘‘Not to say we don’t really know our patients now, but you can’t really know them unless you know them through the experiences of everybody like them.
‘‘And what clinicians now know, is they know them through the experiences of the patients like them they’ve treated, and the ones they’ve read about. And that’s not the same. If you’re treating lung cancer in a 47-year-old white female with a certain genetic profile, a certain demographic, and I have access in real time to information on 85,000 other people very similar to her around the world, all of a sudden my inference and my recommendations become much more accurate, and much more powerful.’’
There’s just one flaw with this spectacular vision – it requires everyone in the world to pool health data, which is likely to raise both privacy and business problems. Health businesses would need to relinquish the idea of data as a valuable asset, and scientists would need to do a better job of convincing the public of the benefits of sharing their most intimate data, Smythe admits.
Success would also depend upon the world’s health systems changing focus from intervening in life-threatening problems to preventing future disease, he says.
Even with the power of big data, Smythe does not expect artificial intelligence to magically cure cancer in 10-15 years. In 50 years, maybe. The human body is as vast and complex as the universe – there is no human mind capable of comprehending even a single cell. Cancer is a derivative of the human body, with the same complexity of defence mechanisms.
What will change is medicine will become more individualised, with a dramatic increase in the likelihood the treatment you’re prescribed will actually work for you, first time. IBM’s Watson supercomputer already uses AI to suggest cancer treatments, but critics argue it has not lived up to the hype, and is too reliant on experts manually feeding it information.
Neither does Smythe expect AI to replace doctors.
‘‘Medicine is our most imperfect, but most humane applied science. And anything that involves humanity requires some degree of art to be effective.’’
While general practitioners might think what they do is mostly art, a recent North American study found the average GP visit lasted 13 minutes, and 5 minutes of that was spent documenting something. In Smythe’s ideal world, AI would replace those five minutes of mundane tasks, leaving doctors more time to care.
‘‘I believe, 50 years from now, technology is going to make medicine more humane.’’
New Zealand Medical Association chairwoman Kate Baddock – a Warkworth GP – agrees AI has enormous potential to help doctors with diagnosis. But she doubts a machine’s ability to deal with complex webs of chronic conditions.
She reckons acute healthcare is about 80 per cent science and 20 per cent art, whereas managing long term conditions is more like 80 per cent art and 20 per cent science. That’s the bit she believes machines can never replace.
‘‘Caring for somebody and the journey of their illness is much more than just the diagnosis... Where AI is most useful, is in people who have a single condition. Now the problem is that that’s not most people.’’
Smythe can’t say if healthcare jobs will go because of AI, but he believes we’re currently in a messy transition stage when it comes to technology. The golden age of healthcare is yet to come, and technology is the key to that.
‘‘With an ageing population in the medically developed world, a growing population globally, growing chronic and preventable disease in the medically developing world, unless we leverage technology, I don’t know how we’re going to get on top of it.’’
"Medicine is our most imperfect, but most humane applied science. And anything that involves humanity requires some degree of art to be effective."
Dr Roy Smythe