AI systems can provide more accurate diagnosis and treatment, without replacing doctors
How many of us have consulted Dr Google with our symptoms? With one in twenty google searches being medical-related queries, it’s got to be quite a few of us. In fact, last year Google updated its mobile app to incorporate ‘self-diagnosis’. By collating internet data and consulting experts at Harvard Medical School, Google said it allows potential patients ‘to get quickly to the point’, before seeking medical advice if necessary. Could this be the future of our healthcare system? Quite possibly, yes. But it must work with us and doctors.
The Google app harnesses an Artificial Intelligence (AI) algorithm to gather medical information, and it’s not the only one. Across the world, the benefits of using AI in healthcare are being widely researched. Most notable of all is IBM Watson, who claims to go ‘beyond artificial intelligence’ and ‘think like a human’. Although not directed at healthcare specifically, IBM Watson can structure, synthesise and analyse medical data, reading 200 million pages of text in 3 seconds.
But, before we delve into this topic any further, we need to dismiss the Sci-Fi stereotypes. That is, crazy depictions of AI as anthropomorphic robots. I don’t think we need to worry about walking, talking robots sitting behind the GP’s desk asking us about our symptoms. If this were to happen in future, then I’ll hold my hands up and admit that I was wrong, but that’s not the direction AI in Healthcare seems to be going.
So, how exactly can Artificial Intelligence be used in Healthcare? In recent years AI technology has seen a paradigm shift. Programmers no longer relentlessly waste hours of their time loading computers with information that will only ever give out what has been put in. AI is now all about deep-learning. This is a technique that uses algorithms to train AI systems to think in the same way that we humans do, whilst holding more information than the human brain can. As Omar Latif, Vice President at Care Manager says, “one of the biggest challenges facing healthcare today is information overload”. An individual doctor cannot know everything, even within their specialist field. Using AI will not only gather all existing knowledge, but process new information and produce diagnosis and treatment suggestions to facilitate decision making.
As a recent example, a team at Stanford University have developed an AI system able to detect skin cancer. To train the deep-learning algorithm, the system was fed over 128,000 images of skin lesions corresponding to various diseases. It was then presented with 2,000 previously unseen images, where it was asked to differentiate between those malignant, and those benign. The results were tested against the opinion of human dermatologists, and found that AI was in fact more accurate.
It’s no surprise that the higher accuracy of AI compared with doctors has led to the speculation that they will soon replace doctors; an unsettling thought. Unsettling not only because we would be putting our lives in the hands of a ‘non-living’ machine, but that it undermines our human ego as the most powerful group on the planet. Now, with our dominance over nature, we could potentially be creating a technology able to take over our control.
This is perhaps the scariest thought of all – that Artificial Intelligence will one day become more intelligent than us. Though this might seem more ridiculous than a robot prescribing us drugs, it’s not as far out there as we think. As philosopher and technologist, Nick Bostrom, points out, the limits of machine processing go far beyond biological tissue. The human brain is limited in its size, as it physically has to fit inside the cranium, whereas AI could take the space of a whole warehouse if it needed to. Additionally, neurones in the brain travel at a speed of approximately 100 metres per second. For AI, it’s closer to the speed of light. Suddenly, the reality of AI overtaking human intelligence not only seems likely, but imminent.
If this were to happen, AI could start to favour its own interests over human values. What these interests or their impact would be, nobody knows. But before we lose sleep over picturing such a world, I don’t believe this is something we need to worry about if we are careful about how we introduce Artificial Intelligence into Healthcare. That is, consciously implementing it as a tool for medical professionals to facilitate diagnosis and treatment. ‘Tool’ is crucial here, as doctors must retain their decision-making power. Though AI might be able to think like a human, it cannot empathise or possess human values.
The next step could be public access to AI. In a similar way to the google app, it is hoped that AI can be used on portable electronic devices covering various aspects of healthcare. Andre Esteva, co-author of the Stanford research on skin cancer detection, is optimistic that this technology can be applied to an app. He called this the “eureka moment” of his work, with the confidence it will make healthcare cheaper and more accessible.
Naturally, the practical issues have been identified. They crop up with every technology, but it’s about understanding them, as well as if and how they can be overcome. For example many doctors, including Dr Anjali Mahto from the British Skin Foundation, have expressed their concern over using AI in self-diagnosis. Potential patients may not be aware of ‘warning signs’ and serious conditions could go ignored. On the other hand, the technology could cause false panic, turning us into maniac hypochondriacs constantly interrogating the AI.
Yet, what AI does indicate is that the boundaries of healthcare go beyond hospitals, surgeries and laboratories. It’s every bit a part of society as it is science. Regardless of who is given access to this technology, everyone is their own expert on their body – we live with them all day, everyday. Doctors can only ever get a snapshot of your health during the occasional appointment, and this is never going to be good enough. It makes diagnosis and treatment extremely difficult, often scary. When your doctor provides you with a possible treatment, all you can do is hope it will work. Nothing is certain.
When we do seek medical help, having an AI system in place has huge potential to improve the care we receive. The proven accuracy that AI is able to suggest diagnosis and treatments cannot be ignored. Additionally, if we had a way of getting reliable medical knowledge ourselves to assist any health concerns before going to see a doctor, then this must lead to greater efficiency. If it also meant we didn’t have to wait weeks on end for a GP appointment, then that’s got to be a good thing!