Interview with Professor Mark Westwood

I am working on a start-up company that brings artificial intelligence into cardiology. One of my two areas of expertise is using MRI to look at the heart in detail (I also undertake CT scans of the heart too). There are tools available to assist in interpreting the pictures, analysing the MRI scans and making a diagnosis. Those tools have evolved, but I think perhaps not evolved as quickly in the field of MRI of the heart as they have in other areas. This is particularly true when we look at the use of artificial intelligence and machine learning.

There is an opportunity to develop something that really uses those new tools that we have in computing science to their full potential for the benefit of cardiologists. It means I can report faster and be more accurate and precise in my diagnosis, but it’s also therefore better for patients and that’s what we’re all here for in medicine.

I was also looking for a challenge, to do something slightly different. I’ve had various leadership positions both nationally and internationally on professional medical societies. This new opportunity came along with a long-standing colleague and friend of mine. He is more academic, I’m more clinical. We thought if we put our heads together then we have a very good combination.

Our start-up was founded about a year ago and we’re still very much developing the product and working with computer software engineers to create it. We’ve got a lot of work to do and we’re still early on in our journey. We want to be out to market in some form this year although we are hampered by the regulatory approval that you need for any medical software product and that can take a relatively long time.

You’re not only looking for somebody who can be a specialist in software engineering and computer science, you’re looking for somebody who can understand what the images mean. On top of that they also need to understand the language of a doctor and clinicians. So as clinicians we have to learn some of their language, they have to learn some of ours. We need good communication skills and to spend the time with that team, to go through what they’ve done, what they see as the priorities and also give lots of feedback.

Every time something’s developed or refined we need to get it in front of clinicians to see what we think and what needs to be improved. It’s a learning experience. That’s part of the challenge I enjoy – finding out how I convey and idea of who a piece of software should work when I can see it in my head and then communicating that to somebody who’s not a doctor.

I think the technology is probably only starting to come of age in medicine and you see that in many fields. There’s evidence now that technology may allow us to go beyond what a human could interpret from images of organs. There’s even a whole specialty area of radiology dedicated to this. Looking at images is pattern recognition and a machine can probably recognise patterns that are so subtle I as a mere human cannot see them.

It’s also pervading many other areas. We’ve heard previously about the design of very specialist drugs particularly in the field of oncology or cancer, but that’s coming to other areas as well. with antibody-based drugs for example a lot of that is driven by technology and computing. This also applies to treatments. We’ve seen robotic assisted surgery etc so everything’s evolving.

I think the only thing that we can know on this journey is that medicine in ten years’ time is going to be
unrecognisable from the medicine that I and my colleagues currently practise. We are on the crest of a wave of a period of fairly rapid technological advancement leading to changes in delivery of healthcare and improvement of outcomes for patients.

If you take ChatGPT 3 for example, it’s passed law school, it’s passed medical school, it can code computer programmes. It can also write the computer code, so it could challenge areas like law and accountancy where we really thought we would always need humans and now it appears maybe we don’t, or at least not in the way we have done previously.

This can be very empowering if used correctly as we can get some tasks done by technology and spend our time doing things that only a human can do such as telling a patient bad news. Will a machine ever be able to talk compassionately and sympathetically to a patient in a way that they would want to convey that kind of information? There is talk that technology could remove the need for all doctors but it’s not going to happen. It will be an adjunct or an additional tool that we have in our armamentarium.

As an example, I’ve started using Duolingo and again, you can see how technology can aid learning. It’s a very good example for everybody, but certainly in terms of practising something. The world of simulation is coming of age but now rather than using real simulators, there’ are virtual simulators. We can see how technology is going to completely change the way in which we train our doctors. At Barts we look after about 200 junior doctors – not just cardiology, but across the other specialties that we have here.

Another question is how do we know somebody’s a good doctor? It’s getting to a point that I think the university essay is dead because now you can cheat very easily using tools that I have just mentioned. A machine can write your essay for you, and you can set that machine’s parameters to say ‘write it as if I’m a 21-year-old student’ and ‘please add ten percent of errors and some grammar and spelling mistakes’.

I conducted a semi experiment with this. I had to give a presentation to a group of highly skilled doctors. At the end, I reminded the delegates that at the start of the presentation I had said that I had written it with the help of some friends. I then told them it had been written by ChatGPT 3 and not one of them had noticed. Medics are generally quite sceptical and dismiss things but here was the proof of the future possibilities and also how radically with these new technologies things will change.

One certainly needs to exercise caution. In medicine, we always tend to programme these things with a high degree of redundancy and safety. We’ve already had this for years with electrocardiograms that come with a report – the programme is deliberately designed to over report so that when it says it’s normal, it really is normal. It virtually never gets that wrong but it will overcall things that the clinician might then verify as being fine. We’ll build these programmes so they are overly sensitive and dependent on human checks – and of course, they will be heavily regulated. One would also hope that in the medical sphere we can work together in unison with our European, American and worldwide partners to achieve that.

We want to embrace this technology – particularly if we look at the backlog we have in healthcare at the moment in the UK, where we know people are waiting for tests or they can’t get results. Some of this work could be accelerated whilst being safe and done with compassion, kindness and sensitivity and taking into account individual patient preferences.

It’s not going change overnight – it will evolve because there is quite rightly some concern that these programmes can do their own thing. I think they’ll be built quite carefully in medicine in order to gain clinician and patient trust.