Hazel Tang A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.
Experts from the medical profession and technologists gathered in central London today for a breakfast meeting to discuss how artificial intelligence (AI) can help the NHS.
Dr Anthony Chang addressed the audience
The meeting was kickstarted by a video message from Dr Anthony Chang, the Chair and Founder of AIMed, who told the audience: “About five years ago I started a series of meetings to focus on AI applications in medicine with the intention to form a multidisciplinary movement to embrace and understand AI in medicine.”
“I realised how critical data science and AI can and will be for the future of medicine, and I’m sure that’s why you’re here on a Tuesday morning. There’s a lot of things happening now in Europe with GDPR, and several national leaders really putting attention and focus on AI, and I just wanted to welcome you to the AIMed meetings and movement.”
A panel of NHS experts
After Dr Chang’s message, a panel moderated by the Associate Chief Clinical Information Officer of NHS England, Dr. Harpreet Sood, and featuring Vishal Nangalia, a Consultant Anaesthesiologist with a PhD in Medical Machine Learning, of the Royal Free London NHS Foundation Trust, and Matthieu Komorowski, a Locum consultant in Intensive Care and Anaesthesia, of Imperial College Healthcare NHS Trust, tackled three key issues:
Dr. Harpreet Sood, Vishal Nangalia, and Matthieu Komorowski
- Applied examples of where AI could make a lot of difference
- Addressing some of the practical barriers in accepting AI (i.e. trusts, transparency, privacy)
- What does this mean for the workforce (how do we practice and function moving forwards, and what does this mean for clinicians)
Dr Nangalia said: “It costs about £40 – £70 ($53 – $93) per patient just to see them, yet for the majority of these consultations it’s just a simple questionairre. AI would take the drudgery out of repetitive tasks, and save everyone a lot of money as well as increasing efficiency.”
Dr Komorowski said: “There’s a lot of talk about AI versus humans and there’s a lot of concern, but I think that’s because people don’t have the expertise yet, they don’t know what AI can do. Go and look at what Gary Marcus says about deep learning, and you’ll see the truth is we are very very far from achieving anything close to general AI systems.”
Vishal Nangalia (right) and Matthieu Komorowski (left)
“Most of the concerns and the far people have revolves around a lack of understanding. It is about putting intelligent tools into the hands of humans.”
Dr Nangalia said: “As clincians we are very good at treating patients. We aren’t data scientists and we need to learn those skills. To say that as clinicians we can develop the best algorithms. I think we know we can’t.”
“If you look at the mathematics behind the algorithms in medicine, they are single variable. They assume that every variable is independent from everything else. There are scores that we use in medicine that say protein and potassium are independent variables. They are directly linked together, yet clinical algorithms say they’re independent.”
“Medicine already has algorithms everywhere we look, we change the way we treat patients on these. Many decisions are determined by an algorithm already. We can just take that one box of an algorithm and replace it with something a hell of a lot more intelligent.”
As the talk went on the panel addressed the question of where hype around AI in medicine comes from, the working relationship between the AI and the clinician, and regulatory issues.
Questions from the audience
Kamil Tamiola, the CEO and Founder of Peptone
A question came from an audience member, Kamil Tamiola, the CEO and Founder of Peptone, who said: “I’m very sceptical about AI or data driven automation because at the end of the day it’s just mathematical modeling, and this debate we are having today about patient data, about predictive outcomes and about predictive medicine, will still be in the hands of policy makers and in the hands of people doing data science.”
Dr Komorowski said: “I do absolutely agree that as much as achievable we need to understand what’s happening under the hood, but sometimes in healthcare we don’t actually understand how things work. Paracetamol, as far as I know, the full mechanism is not fully understood, but the fact that we do not understand the technology does not mean that we shouldn’t use it as long as we know that it works.”
Another question was asked about when will AI be on the curriculum for upcoming clinicians, and Dr Sood said the Topol Review, being led by Eric Topol, will shed light on how the curriculum in the UK will change.
Dr Naila SIddiqui Kamal, a senior lecturer at Imperial College London
Dr Sood said: “This may not be for everyone, not every clinician is going to have an interest in data science, but they need to have some insight on it. I think we need to find the balance for that.”
Dr Naila SIddiqui Kamal, a senior lecturer at Imperial College London, told the panel: “It’s extremely heartening to see that the NHS is engaging with the adoption of technology to help in our achievement of our innovation objectives, and the main thing is clinical health.”
“The point that I would like you guys to take home is that the clinician mindset has to change. We are not preparing, or maybe working, with the same speed as you are working towards engaging with these technologies. When I talk to my peers about this, they laugh and joke about it – artificial intelligence is a joke.”
Dr Nangalia said: “I agree with you, but I think that has changed. I remember when I went to get my PHD fellowship application, I went to the Welcome Trust, and some eminent professors on the panel, when I told them I wanted to look at data, turned around and said: “You’re just data fishing, that’s not real science.”” Today, I doubt any of them would say that, they would be keen.”
“You will always have people who are naysayers, you will always have people who say there are other priorities. And there are other priorities. I talk about AI, but I work in an office where it takes five minutes to log on to Windows. We’re not gonna fix everything.”
The audience joined in a spirited, interactive debate with several people openly disagreeing with the panel. The discussion was an hour in length and was followed by a networking session.
Attendees were encouraged to sign up for AIMed Europe, 11-13th September, for a wider ranging discussion and even more high-level debate.
