S2 E14: S2 Ep 14 – AI in Healthcare (Part 3): Where AI Can Actually Help

Okay, welcome everyone.

This is the third episode in our AI series on therapy and healthcare.

The first episode was about the tools, the early years of AI and where we are now.

The second episode was really about the pitfall.

Pitfalls and things to avoid was.

Now this episode is about how what AI can do really, really well and if it's done right, what benefits can it bring us as a population?

So, Bob, over to you.

Where does it work for you in physical therapy?

Right.

Okay, so after, as you said after our last podcast about some of the pitfalls of AI, one of the things that I mentioned was the fact that AI doesn't really do nuance.

It's not.

It doesn't think.

So AI is very much about.

It's a prediction machine, so it's very much about thinking about.

Okay, this is a question, this is the answer.

Now, AI can't think yet, but there are AI models which I can see as being very useful.

I don't.

They will never substitute for a physical therapist because I can't do hands on.

And regardless of how advanced robotics get, they may get to a point where, where it can do hands on.

But what.

Why would you do that when you've got people with those skills anyway?

But that's a whole nother story again.

So, yeah, where I would say one of the things that I'm working on at the moment is building an AI version of myself because I can only see X number of people per day, X number of people per week.

If I could have an AI model that do a lot of the kind of triaging.

So when we talk triage, it's an initial assessment of what's going on.

So the idea is that I do a, is that I create a cyber Bob who can do, who can do the.

Take a case history so it can ask all the right questions.

AI can sense movement to a point.

So potentially we could build in as well as taking a case history of, okay, what's the pain, where did it happen, when did it start, what medication are you on?

All of those kind of questions that can be done by AI because they're fairly standard questions.

So that could be done then potentially physical assessment, not so easy.

But we can get AI with motion sensors that could then pick up, okay, I want you to squat, I want you to bend forward, I want you to bend to the side.

And then it can look at, it can compare their movement with ideal movement and just see what, what comes up from that.

And then what I can do is I can give it an understanding of how you do a diagnosis.

So it will go through all that information, it will pick out two or three of the most likely problems that caught, likely issues that cause that problem, and then it can come up with a potential solution to that.

Now, if it's purely a physical exercise point of view, then again I can do that.

It can be, it will be able to scan the person and see whether they're doing the movements right or not.

If it's something more complicated, like an example I spoke about in the last podcast, but if it's something more complicated where the person had an ankle sprain 20 years ago and they're now getting knee pain on the opposite side, technically the AI, based on the information that I give it, should be able to make that connection.

Because I've done it, I can put in these are the things that I looked for.

Is that the case?

And if the AI gets something that's too complicated or that it doesn't know what to do with it, or it's set or it's, it's a red flag to say that, yeah, you'll get, you know, cord require syndrome where you, where you're getting raging pain down both legs, or you got saddle area anesthesia, or there's numbness around the groin area.

And there are various other things that you can put in as standard rules.

AI can go through those and it'll, and it can make a decision.

So if it gets stuck, then refer to a human being.

So that's kind of one of the things I'm working on.

And from what I can see, the potential is there, just scraping the surface of that at the moment.

And hopefully in six months to a year's time, I'll have a better idea of whether it can actually work or not.

But that's one of the things I can see AI doing.

One of the other things that it's very good at is repetitive tasks.

So some of the backroom stuff that I really hate doing, like writing up notes, we've got voice to text, so I can have a conversation with somebody in a session, I can have full on eye contact with them and the AI is recording the conversation with the person's permission, obviously, and then it just translates that into text and then that goes into updating their notes.

So AI can save lots of time with things like that.

It can also do documentation, writing GP letters, that sort of thing.

Key thing about all of this is that initially at least, there will always be human oversight.

I will not let AI do anything without checking it first.

And then if there are errors.

I can correct those errors within the AI and then it'll get it better next time.

Another use for AI is diagnostics.

So where you've got.

They've come up with some AI software which can detect, which can listen to heart patterns.

So there are certain, certain conditions where a human can't necessarily hear the subtleties in change in heartbeat.

But you can train AI to be able to detect really small changes in heartbeat and come up with a diagnosis for those.

And I believe that that's being rolled out.

If I can find a link to that study, I'll put it in the show notes.

But there are lots of uses for AI in medicine, but it's an ongoing process.

It's not going to replace us and it's not going to replace, or it should not be used to replace gps, because there's.

Because however advanced AI gets, there's still that.

There are still benefits in having that human interaction and I don't think AI is going to be good enough to replace that.

But there are lots of things that AI is, are quite good at.

So that's kind of it from the therapy perspective.

Claire, over to you.

Yeah, I mean, I would agree with you absolutely on it not replacing the human empathy element.

And that is one of the things that's one of the key skills in terms of a coaching therapy session is being able to understand, pick up on emotions, being able to pick up on nuance, be able to pick up on body language and tailor the session accordingly.

I don't believe that AI would be able to do that, but that could be something in the future that it may be able to do, I just don't know.

But certainly the fact that it's a confidential conversation, it's professionally managed, and I have access to so many different frameworks depending on the kind of issue that the client presents to me.

Again, it's about a bit like Bob talked about, it's about triaging, it's about understanding what that client actually needs in that moment.

And again, I don't think AI is there yet in terms of the chatbot therapy tools, but I absolutely believe that it's better than nothing at the moment.

And some of the cognitive behavioral therapy tools that are out there at the minute are actually very good because they help people reframe thoughts and feelings.

And then the idea would be, is if they get to a point where actually I think.

I think I'm getting somewhere, but I need some help with this, potentially that's when they would come and see a human to be able to take them further with it.

And you would hope that the AI could say that at this stage we need to route you to a human being.

Yes, exactly.

I would be very concerned about AI, you know, talking to someone who's got trauma or PTSD or an eating disorder.

And certainly in the last episode we talked about Adam Rain who where there was mentioned about suicide.

I think there is a limit to how AI can interact with somebody.

Where I do find AI really helpful is to extend the therapy session beyond the session itself.

So there are some really helpful tools out there.

So for example, helping someone with journaling.

Journaling is a really good way of someone understanding patterns and triggers.

And you don't necessarily have to sit there with a piece of paper and do it that way.

You can do voice notes, you can do all sorts of things.

But what the AI within the, the software would do would be able to help you identify those patterns and triggers that you might not necessarily see yourself.

And whole part of therapy is being able to become aware of these patterns and triggers.

That's starting point really.

So I find that really helpful.

There's a wonderful little AI tool called Finch which quite a number of my clients have used, which is effectively a well being wellness tool.

It's a bit like if you remember the Tamagotchis, where you'd actually have to look after this little thing.

Finch is very much the same, but you set its own parameters and it's very much about well being and wellness.

And I have had clients that have found that to work really, really well for them.

Tapping as well.

There's various apps out there that you can use for tapping and it will help you identify your, you know, your tapping points that work best for what you need.

There's also other apps out there about mood tracking.

Okay, being able to help you with your mood.

But again it's all down to being able to keep a note and identify patterns.

And that's particularly what AI is trained to do as a therapy coach.

We are trained to identify patterns and pick up on those things.

But if you've got this, this tool or app in the meantime, that whole process is going to happen faster and you're going to see the evidence yourself.

So that's certainly very, very, very good.

I haven't actually gone down the route of transcribing conversations yet because I haven't found the software that I would feel confidently that, that that information is going to be confidential.

I know that in all my notes I don't put names.

So if, if that might be the way around it.

But I.

But I know other therapists do use that because it's very.

It helps in terms of time and accuracy as well.

Yeah, there are tools out there that have got that level of confidentiality.

So for me doing Cyber Bob and setting that up, it will take the notes, transcribe, and it will store them in a secure way that is GDPR compliant.

So they do exist.

Right.

Well, yeah, that's another conversation.

I'll have to talk to you after the episode, but I haven't gone down that route just yet, so that's something I. I need to do.

But essentially it gives people extra support between therapy appointments, which I think is really, really helpful.

And so.

And there are just more and more new apps and more.

More AI support available all the time.

Yeah, and that just kind of reminds me of.

There's also.

That's that the similar aspect in the physical therapy side of things, because I'm very keen on exercise.

Movement is medicine.

As if you've seen any other podcasts, you will know.

I give people exercises to do, and one of the things that I can do is to keep a track of what people are doing when they've done it, and it can also give them progression.

So key things with exercise are if an exercise is too challenging, you regress it back to an easier version, and if the exercise becomes very easy, you can progress it and make it more challenging.

So one of the things that AI can be very good at is keeping track of when people do exercise, what they've done and whether they need to regress and just make things a little bit easier because they've not had a great day or to progress because they've done it and it's fine and they're ready to go on to the next level.

So AI is very good for that sort of thing.

And the other advantage of it is it's almost like gamification, which is the Finch tool that you're talking about.

Yes, yes.

So with games, you know, you can give somebody like an archery type contest.

So.

And it's all about movement of the arms, movement of the shoulders, and you can say, okay, here's a target.

Let's hit three things on that target.

And if they do that, then the AI can sense they've done that.

Okay, that's time to move on now.

Let's give them 10 ways to do something.

So making something more of a game is going to get more exercise compliance because it's fun and people will do it.

And it's also a nice way Once people are out of the clinic, then they tend.

And once they're out of pain, they tend to stop doing the exercises that got them out of pain in the first place.

And it's another way of communication, keeping track and just letting people know that you've got their back and that you still care enough to make sure that they are doing what you've asked them to do.

So, yeah, that's just another advantage of using something like an AI tool that can make decisions.

Yeah.

I think the thing.

I heard the phrase the other day that rather than thinking of AI as artificial intelligence, to think of it as assistive intelligence.

And I really like that.

I think that's how we should reframe.

If we're talking therapy, we should reframe it as that.

So it's not necessarily a threat.

I don't think it's going to get rid of either myself or Bob anytime soon.

I would hope not.

But we can use it as support in different ways and we have to be open to how it can help us both in the administration of what we do, but also with our clients.

Yeah, absolutely.

Because the thing is, AI is out there.

It's not going away.

And so rather than dismiss it and go, oh, no, we're Luddites and we're like, it's never going to catch on.

I remember saying that about mobile phones and look where we are now.

Rather than being Luddites, it's actually better to look at what is available and see how we can use that now.

And then as things advance, we can progress with it.

So, yeah, I think AI has massive potential.

Some of that is already being realized, but I think there's so much more that can be done.

And I think it's important for therapists like myself and like Claire to be in a position to take advantage of wherever it goes next and just be ahead of the curve rather than sort of trying to catch up.

Yeah, exactly.

We are excited about by this, really, aren't we?

We are.

We're not.

We're not serious.

It's going to.

It will make a massive difference to a lot of people as long as it's used.

Right?

Yeah, I think that's the key.

That's the key.

That's the key.

Yeah.

Are we.

Are we done with AI now?

Yes, I think so.

Well, this part of AI, there's so much more we could talk about.

Yeah, there is a lot more we could talk about, but, yeah, hopefully you got a lot out of this one.

And as we said in the three episodes, we've tried to give the history of AI from a therapy perspective some of the pitfalls that you need to be aware of and hopefully give you some ideas of where it's going to be going in the future.

And we will be going along with it.

So that's goodbye from me.

And goodbye from me.

See you at the next one.