AI in Clinical Research Fellowships (13 Sep 2023)

SPEAKER_01 0:00

If you enjoy listening to career conversations, why not become a member of the Royal College of Physicians of Edinburgh? Our membership provides you with access to the RCPE educational portal, the live evening medical updates, and you have options to view the symposia both in person or online. If you would like to learn more about this, please go to the Royal College of Physicians of Edinburgh website. Hello everyone and welcome to another episode of Career Conversations brought to you by the Royal College of Physicians of Edinburgh Training and Members Committee. My name is Dr. Marilena Gianudi, and today I'm delighted to be joined by Dr. Josh R. Young, who is a specialty registrar in urology in London and has taken an out-of-program experience to work as a clinical research fellow in artificial intelligence. He will be talking to us a little bit about what his year has been like. So, Josh, thank you so much for joining me today.

SPEAKER_00 1:03

It's a pleasure. Thanks for inviting me, Marilyn.

SPEAKER_01 1:06

So, first of all, what's an out-of-program experience, if you don't mind if we start with that?

SPEAKER_00 1:11

So, you know, for clinicians that are in training who wish to partake in some other kind of experiences, you can take it out-of-program leave, right? So that can be either OOPR, so out-of-program research or OOP out of program experience. So technically, for this current post, there is a research component, but I'm also very involved in different operational aspects. So therefore it's an OP. So you apply to an OP through the college, and it's quite a simple, straightforward application. You just need to justify how this kind of contributes to your development.

SPEAKER_01 1:47

And with an OP, was this kind of a job that was already advertised that you needed to apply to, or was it something that you knew you went, found a potential supervisor and you kind of got the ball rolling yourself?

SPEAKER_00 2:02

That's a great question. So it's actually a bit of both. This post was put out as a public advert, but obviously prior to the post coming out, I was already interested in AI, and I reached out to this professor, Professor James Teo, who is the neurology professor and also director of AI in King's College Hospital. So I reached out to him and I was actually doing some work with him already, and he basically flagged up that this post was coming up. But you know, the post was public and was open to different applicants. But I think it probably did help that I was working with the team prior to that.

SPEAKER_01 2:39

And I do think you can go to any medical conference, you can read any journal, or even just, you know, reading the news nowadays without having something come up on artificial intelligence. So what does artificial intelligence kind of mean to you? Let's start with that question, then we can maybe go a bit into what you've done during this year.

SPEAKER_00 3:05

Sure. So artificial intelligence or AI, lately, as you know, you quite rightly said it's become a really high buzzword, which is used everywhere in media, in different organizations, in healthcare, and NHS is no exception. And part of the reason is AI means something different for everyone, really. So for lay people, AI represents probably a big vision of transformation with technology, and there are all these ideas about sentience and you know AI taking over the world. But for people who have delved deeper into AI, it's a lot more pragmatic. I don't like to generate too much hype with AI. At the core of AI, it is a tool for clinicians and people to use. And essentially a key component of AI is statistics. And you use statistics to implement a learning process in machines. And once you get over that barrier, you realize that AI is very useful, but it's not as hyped up or dangerous as the media would have you believe.

SPEAKER_01 4:11

I guess especially in healthcare, where we're using AI for good as opposed to for anything bad.

SPEAKER_00 4:18

Yes, exactly. And I don't know if you've had a chance to play on ChatGPT. You know, that's the talk of the town at the moment. But, you know, since ChatGPT came out, for those listeners who don't know, I mean, if you haven't heard about it, you've probably not been paying attention. But ChatGPT is a language model made by OpenAI. And essentially it's learned a representation of the human language, and you can really have any kinds of conversation with it, and it will give you very convincing conversational outputs. So it sometimes feels like you're talking to something sentient. At the back of it, it's actually just code. But when ChatGPT came out, there was a lot of hype and a lot of clinicians for saying that this will transform healthcare. But the thing I wanted the listeners to understand is there are many safety measures within healthcare and many regulations which you have to get across. So this technology is very far from being implemented in a healthcare setting. And there are very strong barriers in place to make sure that we implement AI safely and consciously and ethically within the clinical setting.

SPEAKER_01 5:23

And can you tell us a bit about the work that you've been doing with AI this year?

SPEAKER_00 5:29

So I'm working in a team based in St. Thomas's Hospital. The team is called Cogstack. It's an NHS homegrown AI team that focuses on natural language processing or NLP. For those who don't know, natural language processing is a branch of AI that focuses on human language. That's understanding and emulating and interpreting human language specifically. So the team works to implement NLP into the clinical departments, and that's been my main role this year: working with many different departments and clinicians to implement language AI to help solve a particular problem which they've been encountering. And importantly, as you know, in clinical notes, most of the patient data is actually in free text in the electronic patient record rights. So there's a lot of useful information there. But the way clinicians write in medical language, we use a lot of abbreviations or shorthand, or even between departments we may be writing differently. So it's really getting the AI to understand the nuances in medical language and how it can use that to structure data to make a difference in their workflow.

SPEAKER_01 6:44

So has it been a lot of coding, for example, from yourself?

SPEAKER_00 6:49

Yes. So there are many skills which are required, I think, for this role, coding being one of them. So you do need to build up technical skills. And I think that's the bit that's quite daunting to clinicians who want to join the AI revolution. So for the first part of my job, I probably spent the better part of two to three months just sitting there and doing coding courses, reading about machine learning, doing various online courses to get my knowledge up and get my technical skills up so I become a member who can contribute to this body of work that's happening right now.

SPEAKER_01 7:25

So before you went into this, you had absolutely no prior experience?

SPEAKER_00 7:30

No. So prior to this, I was working as an academic clinical fellow in St. Thomas's, just you know, running the medical take. And I really had to upskill for this position.

SPEAKER_01 7:41

I guess that's very good. And that's what I kind of want to highlight is that you don't need to have prior experience to do this. You just need to have a lot of willingness to learn, it sounds like, and patience when coding doesn't go right.

SPEAKER_00 7:55

Yes, uh absolutely. Coding is just like learning a language. The more you do and the more you engage and the more you practice, the better you will be at it. I think a lot of newcomers are looking for shortcuts, but really the best secret I have to offer you is to just take the time and spend those hours to really get to grips with how to code.

SPEAKER_01 8:18

And so what skills do you think somebody needs to be, I don't want to say good at AI, but successful within AI.

SPEAKER_00 8:27

Yes, so there's many different ways a clinician can get involved in AI. If you're like me who wants to get into the technical and research aspects, then yes, you need to learn the technical skills and coding, etc. But if you wanted more of an advisory or regulatory role, I know many clinicians that have joined from that position, so they don't necessarily know how to code, but they have gained an understanding of how the models work and how it should be implemented in healthcare. And then they can have a regulatory or advisory position in that sense. Also, there's a lot of money in health tech and AI at the moment in startups and different money pots flowing into the NHS for digital health specifically. So a lot of clinicians also position themselves as, for lack of a better term, kind of a health tech guru. You know, they post a lot about the hype of AI and they may get involved in raising capital for different startups, and that's an entirely different skill set. I would say that's more marketing focused, but I guess that's another way in which you can get involved in AI. That's not the path that I have opted for, but yeah, that is an option as well.

SPEAKER_01 9:42

And Josh, I think this is quite obviously it is the future and it's something that we're seeing a lot about. For me personally, the idea is terrifying. And maybe because you know, the idea of coding and you know, very complex statistics just doesn't sit right in my brain. But what made you want to explore this route?

SPEAKER_00 10:06

First of all, just addressing the comment you made about statistics and mathematics maybe not sitting right in your brain. Lots of people say that, you know, in high school and university, I've heard countless people say this. And I I'm not sure I fully agree with the statement. I think if people did put the time into it, or maybe people didn't have great teachers, which you know others had during high school, but it is actually quite intuitive. I think there is always a light bulb moment in stats or maps. When you engage enough, it will make sense to you.

SPEAKER_01 10:37

Of course.

SPEAKER_00 10:38

Yeah. So I think the first thing is not to be daunted and really put yourself in a position where you can, you know, learn more about this. I would say instead of being daunted, just have a go, reach out to you know, a team or a professor, and really just engage for you know a few months and see where you get, because it might be not as daunting as you think, and you may end up loving it.

SPEAKER_01 11:06

But what made you want to go into this?

SPEAKER_00 11:09

This is a story which probably goes quite far back to when I was a teenager. I've always been interested in mathematics, and actually, well, I'm not sure if I should say this, but when I was applying to university, I actually wanted to go for a mathematics course. However, my parents being kind of more conservative and you know Asian stereotype, they really pushed me towards medical school. So I ended up, you know, going through medical school and I did enjoy it to an extent, but at the back of my head, I always wanted to be more involved in mathematics and stats and computers. So I was working as an academic clinical fellow during the pandemic in St. Thomas's, and my research was lab-based. It was more experimental medicine looking at dietary nitrates and its cardiovascular effects. But I'm not sure if this is, you know, fates or I just made my own opportunity. But because of COVID, all of this research was put on hold, remember? So all the vaccine research and the COVID research was prioritized, and all the other deemed less urgent research was put on hold. So essentially, my lab time was ceased. So that was the time I took the opportunity to reach out to Professor James Teo. I did a lot of reading around to see who I should contact, and I basically just sent him a code email saying, look, I'm an ACF, I have academic time to spare, and I'm happy to engage with the process and learn more about AI. And I think this proved my commitment to the research. And then after working for them, well, working with them for a couple of months during my academic block, that's when he told me that there was funding for this AI clinical research fellowship post, and then the rest is history.

SPEAKER_01 13:03

And obviously, you know, you're doing it to being a neurology registrar. Do you think that anybody of any medical background can go into AI, or that maybe some specialties are more suited to it than others?

SPEAKER_00 13:17

I think the stereotype is that neurology is the most suitable for AI because historically there has been this virtuous cycle between neuroscience and artificial intelligence, because in essence you're emulating the human brain, right, which you study a lot of in neuroscience. But practically speaking, I think we have progressed a lot from that time, and it I think regardless of the specialty that you are in, you can definitely engage in AI, and really being a neurologist gives no difference, really. I think any specialty can learn and engage in AI.

SPEAKER_01 13:57

And obviously, you've done this out of program experience at the end of your ACF. Do you think there is a good time for somebody in their training, whether it's, you know, going down the academic route or the non-academic route, to undergo an out of program experience, or is it literally just when an opportunity may arise?

SPEAKER_00 14:18

In terms of time, I think it really depends on the person. For me, this was a natural break, you know, between IMT3 and ST4. But I would say for those interested, the best time is probably now or as soon as you can, because AI is such a rapidly growing field, and there is such a shortage of doctors who have the expertise in AI. So I would say if you are genuinely interested, it's good to get the experience as soon as you can. There are a number of posts which are funded by Health Education England called Fellowships in Artificial Intelligence, I think. So this was started only a few years ago in Guys and St. Thomas's hospital, but now it's brought out nationally, including Wales and Scotland, I believe. And each year there are a number of posts in which you can apply for. And usually this post gives you around two days a week of time dedicated to the AI Fellowship whilst you continue your training alongside. And this is a year-long post. So it's great to get a taster of AI that way.

SPEAKER_01 15:27

And that was actually going to be my next question. So have you been doing any clinical work during this year?

SPEAKER_00 15:34

I haven't been doing any contracted clinical work as such, but I have been still doing shifts as a locum every few weeks. Because I find that if I don't practice, I'm worried I might lose some of my clinical skills and edge. So I have kept startup on the side. But no, there's no designated clinical time as such.

SPEAKER_01 15:55

Okay. And Josh, what do you think has been your biggest learning point of this year?

SPEAKER_00 16:02

I think for me the biggest learning point is reaching out, taking the initiative to pursue what you want, and then getting engaged. So throughout my career, I've actually been doing this a lot, and I suggest listeners who want to do a specific specialty or research of mine do this as well. What I did was I sent a lot of code emails to researchers and professors in the past because I was interested in their research and I wanted to spend some time understanding and working with them. So I did this when I was in Newcastle University. There was a mitochondrial group headed by Professor Turnbull. You know, I didn't know anything about mitochondrial diseases, but it sounded very interesting. So I just reached out to him. I emailed him one day during the summer. I said, Can I spend a few weeks with you and work in your departments? And I'll be happy to contribute as much as I can. And this is the thing, I think throughout the course of my life, I've sent out several of these emails, and I've never had no for an answer. So he said yes, and then I spent a few weeks in the summer working with him and his team, and it was a really invaluable experience. It didn't translate to any research or any future opportunities, but I think being able to pursue what you are interested in is very important as a doctor. And I did the same thing during my time as an F1. I reached out to a multi-neuron disease group at the Welton Center. I just sent the professor an email, and then I ended up working with them, and we presented our findings in the international poster in the conference. And then likewise, this current post, I reached out to Professor James Teo during the pandemic. So I think that initiative and commitment is very important because it's a two-way process, right? You send out an email to someone who is accomplished and who is doing incredible work, and they agree to take you on as essentially a mentor, and you're joining as an intern, in a sense. And I think the unspoken contract is that, okay, they're taking their time to teach you and to, you know, help you learn the ropes. Your role in this is to try to deliver and commit as much as you can because I have seen this the other way now. Because this year I'm working as an AI clinical fellow. Many clinicians and medical students are very interested in joining me and doing some work with me. I've had countless emails from medical students and I've taken on a few. But what I find is they normally, after a week or two, they lose interest. They see the hype, they they want to get something on the CV, they want to get a quick publication if possible. But once they realize there's actually a large body of work to learn and a lot of work to be done before publication, a lot of them lose interest and they leave. So I would say number one is take initiative and number two is to commit and deliver. And I think if you do it that way, you're gonna have a very bright career.

SPEAKER_01 19:09

I think that's really important. And I think when you're seeing these really, you know, highly accomplished professors, you tend to say, Oh, you know, they won't look at my email, they don't have time to like help me, they don't have time, and you end up getting into this vicious cycle where you know, if you want to do something, you really want to do it, but then you almost talk yourself out of it. So I think that's really good. And thank you for highlighting the point of don't be scared, just go for what you really want to do.

SPEAKER_00 19:38

Yes. And the truth is, you know, if you're involved in research, you'll understand that it's always nice to get an extra pair of hands. So they'll rarely say no. There are always some kind of work which you can do and contribute to. So don't be afraid and yeah, make that first email or phone call and reach out.

SPEAKER_01 19:55

Great. Thank you. Thank you so much, Josh. I think that that's the end of my questions. I just want to thank you for coming on and talking to us today and for all your insights to our listeners. And hopefully you'll have lots of people reaching out to you in the future who do want to be involved in AI who aren't scared of statistics.

SPEAKER_00 20:18

Yeah, it's it's a steep learning curve, you know. Yeah, I'm sure there'll be more people coming soon.

SPEAKER_01 20:23

Yeah. Thank you so much. Thank you.

SPEAKER_00 20:26

Thank you, Marilena.