Close
Creating an artificial pancreas to improve diabetes management?
Diabetes is the leading cause of death and disability worldwide, and diabetes prevalence is increasing [1].
In this episode we talk to Ken Li, Assistant Professor of Artificial Intelligence in Healthcare at UCL, and Melissa Connolly, who has type one diabetes and blogs about her experience. Ken’s doing something amazing – he’s working on a device that just makes life easier!
As a type one diabetic, Melissa always has to keep an eye on her blood sugar levels. If it’s too high or too low, she has to adapt her day according to the readings. Imagine wanting to go for a run but you can’t because your body hasn’t been able to regulate its blood sugar for that to happen.
In today’s podcast we talk about all that. We discuss what it’s like living with diabetes, what Ken’s device can do, what devices are already out there and what the regulatory landscape looks like and why we can’t think of a device today and get it tomorrow.
About Ken Li
Ken (Kezhi) Li is a lecturer (Assistant Professor) of AI in Healthcare, at Institute of Health Informatics (IHI), University College London (UCL). He is passionate about solving physiological, medical, clinical, operational problems in healthcare using AI techniques, in particular in diabetes management and biomedical time series analysis using monitors/wearables. His team was the winner of international blood glucose level prediction challenge for type 1 diabetes in 2019.
About Melissa Connolly
Melissa's been managing her diabetes since the age of 6 and was recently put on the pump (you can hear about what that is and how its like on the podcast). Melissa's a friend of Alice, our producer, and after reading through her very well written podcast, we thought she would be perfect for the show. Melissa gives a different perspective to the story - that of someone living with diabetes and who benefits from the type of devices Ken and his team are creating.
Listen now
Other ways to listen and subscribe
[1] Source: The Lancet
- View the episode transcript
Ferdouse Akhter 00:05
Hello, and welcome to health in a handbasket, your podcast about the sexy world of Healthcare Engineering. I'm Ferdouse Akhter. And I'll be your host. I'm the Marketing and Community Manager at UCL's Institute of Healthcare Engineering. And although I don't always understand what's written in the research papers published by academics, I know that what we're doing in the world of healthcare engineering is important and impactful, and I want to share that with you by speaking to those who know a bit more about him than me. Today, we're picking out the topic of smartwatches. Everyone's got one these days, I'm not wearing mine, because I always forget to charge it. But they're really clever. You can track your heart rate, your steps, and manage your diabetes. So you probably didn't think I'd say that, but that's what we're going to be talking about today. My dad has type one diabetes, so I'm familiar with it. It honestly takes over your life and the things that would be unconscious daily actions really aren't things like eating, sleeping, you know, managing how hot you are - you really have to think about those things when you have diabetes. So for anyone living with a chronic illness - is a hard life. So having something that helps you manage your condition is honestly such a significant step in terms of healthcare. So let me introduce you to our two guests today. It's a really busy recording studio today. I've got Ken Li with me who's an Assistant Professor of Artificial Intelligence in Healthcare. He's passionate about solving physiological, medical, clinical, operational problems in healthcare using AI techniques, and in particular diabetes management using monitors and wearables. His team was the winner of the International Blood Glucose Level Prediction Challenge for Type One Diabetes in 2019. And I'm also joined by Melissa Connelly, who is a type one diabetic. She was diagnosed when she was six, and is currently using an insulin pump that she started using in February of this year. She's been blogging about her experience, and you can read her blog - we'll put it in the show notes. And so hello, both Ken and Melissa. How are you guys today? Good, Hi, everyone. Yeah. Super hot today. So we're all a little bit sweaty here. Okay, let's start right at the beginning. What is diabetes Ken?
Ken Li 02:19
So diabetes is a group of metabolic disorders characterised by high blood sugar. People who have diabetes normally will experience like, when your diabetes is too high, you will experience what we call hyperglycemia. And when your blood sugar or glucose level is too low, we'll say like experience or hypoglycemia. People's target is to maintain your glucose level. Not too high, not too low - in a very central middle range.
Ferdouse Akhter 02:48
What is glucose, Ken?
Ken Li 02:49
Glucose, you can also call it like a blood sugar.
Ferdouse Akhter 02:51
And you get that from food.
Ken Li 02:53
Yes, you get you get it from food, your your body digests the food and the blood sugar into your blood. And it like goes to your every corner of your body to send energy.
Ferdouse Akhter 03:04
Okay, yeah. And I know that there's two types of diabetes. There's type one, which is...
Ken Li 03:09
Yes, there are mainly two types, or actually are three types. Yeah, type one, type two, and gestational diabetes. So gestational is for pregnant woman.
Ferdouse Akhter 03:24
And so only pregnant women get it?
Ken Li 03:27
Yeah. Only pregnant woman will get it
Ferdouse Akhter 03:29
And after their pregnancy it's gone?
Ken Li 03:31
Yes, that's correct.
Ferdouse Akhter 03:33
Okay, so what I understand about type one and type two, so type one is a lifelong condition. So your immune system attacks, and destroys the cells that produce insulin. And that's what Melissa's got. And type two is where your body doesn't produce enough insulin, and their body cells do not react properly to the insulin that's being produced.
Ken Li 03:51
Yes, that's correct. And normally, type one is more severe than type two diabetes. And the most of people who have type one diabetes are because of family reasons. But for people who have type two diabetes, some of them are due to family reasons, or some of them due to lifestyle reasons.
Ferdouse Akhter 04:12
I mean, diabetes is quite common. I think a lot of people listening will be familiar with it, because it affects one in 11 adults, so Melissa's got it, my dad's got it. And 10% of people have diabetes have type one diabetes.
Ken Li 04:25
Yes. I think there are more than like 500 million people in the world, globally, who have diabetes. So that's a very large number.
Ferdouse Akhter 04:34
A huge number. Yeah. So Melissa, you've got type one diabetes, tell me what it's like living with it.
Melissa Connolly 04:40
So I was diagnosed when I was 6 - 25 years ago now. And I've obviously I've grown up with it and it just feels normal now.I don't really remember what life was like without being diabetic. Obviously, when I was younger, my parents would sort of manage it for me. So as I've got older, and I've had to learn how to do it myself, that was a bit of a challenge. As a teenager, I sort of pretended that it wasn't a thing. And I think that's quite a normal thing. I think that's quite a common thing for teenagers to do - you just put it to the back of your mind and sort of get on with your life. But as I sort of got into my 20s, I realised that it was sort of time to take the control back over it. But at the same time, that's when I started sort of getting into running and exercise. And obviously, that helps with health and weight. But that sort of brought challenges with my diabetes and I've sort of learned in the last sort of 10 years, how hard it is to balance those things. Yeah then in February, I got my pump to hopefully help me with it. And that's when I started writing about it to hopefully help other people that are struggling with the same things that I am.
Ferdouse Akhter 05:43
How do you manage your diabetes? How do you inject insulin?
Melissa Connolly 05:47
So I started using a insulin pump that sort of attached to either my leg, my lower back - may touch my stomach or my arm, and that administers a little bit of insulin every hour. And then I use a device like a smartphone to inject a top up of insulin every time I eat carbohydrates.
Ferdouse Akhter 06:07
So you have like an app on your phone that kind of monitors it, or do you log your insulin on that?
Melissa Connolly 06:12
So it's a smartphone in itself. So it's not an app on my phone. It's a device that looks exactly like a smartphone that has just one app on it. And that manages how much insulin I administer.
Ferdouse Akhter 06:27
...you have to always carry around two phones.
Melissa Connolly 06:29
Yeah, yeah. Yeah, like everyone else has one phone that they have to keep a hold off, I have to and I'm not very good at keeping a hold of my one phone. So two is really fun.
Ferdouse Akhter 06:38
Ken, there's also other ways of managing it? It's not just using a insulin pump? What in what other ways can you manage your diabetes?
Ken Li 06:46
Yeah, traditionally, like five or 10 years ago, people normally use like insulin pen to inject insulin, certain units of insulin into their bodies, along with meals normally, but in the recent, like five years, a lot of new technologies developed including the insulin pump, so more and more people now use an insulin pump and any other like variables to help them injected insulin dose,
Ferdouse Akhter 07:11
I guess, with the insulin pen you have to be very active in managing your own condition because you have to remember to take it and you know, take it after every meal, but with the pump it's a bit different.
Melissa Connolly 07:21
So when I was using the pens, I was I would have to inject a background dose of insulin when I woke up and when I went to bed, and if I didn't, my blood sugars would spike quite a lot, despite sort of doing the top ups of fast acting insulin if you like. And with the insulin pump, you don't have to remember to do that because the background is happening all the time anyway, but then you do still have to remember to administer the top up doses so it's easier in most ways, but there's still that oh I've eaten a pizza and I've forgotten to inject that still happens sometimes which is really fun.
Ferdouse Akhter 07:54
What are the consequences if you don't take your medication properly.
Melissa Connolly 07:57
So if you don't do enough insulin, your blood sugar goes pretty high. If I'm if my sugars are high, I get very thirsty I need to go for a wee a lot because obviously I'm drinking a lot I feel quite sick headachy and I don't get this but I know quite a lot of people get so your vision can go a bit like blurry and then if you take too much insulin and your sugar goes low it's it's almost a bit like being drunk so you can I can't think straight again my vision does go a little bit blurry is the wrong word but foggy I guess feel a bit weak feel quite like sleepy. I can feel a bit sick. Yeah, it is. It is just a bit like being drunk. You say some weird things. And also I get quite angry. So yeah, there's there's sort of two sides of it. If if you do too much, it goes one way. And if you don't do enough it because complete other.
Ferdouse Akhter 08:45
I mean, I find it so crazy that a hormone like insulin can affect your mental and physical capacities, like is crazy. Like it has such a big impact on our lives. And I think with people who don't have diabetes, you take that stuff for granted. You don't realise how important insulin is and how like, I don't have to take insulin after I eat or, you know, do that kind of stuff. Or like I don't have to worry before I go to sleep. Is it too low? Will I wake up the next day, because I think with my dad, he had like a few insulin coma. So like in the morning, he wouldn't wake up. So it's crazy. It's like, it's super scary that you have to really, really manage your life in such a particular way. What are some of the surprising ways diabetes has affected you?
Melissa Connolly 09:24
So obviously, I've sort of touched on exercise and things like that. But I started exercising to lose the weight that I'd sort of put on at the end of my teens when I lived with my friends. And probably about five years ago, I started to suffer with a little bit of anxiety and I always thought that was sort of that thing where you know you're trying to keep weight off and you're worried about what you're eating and it was just sort of generalised anxiety, but when I actually spoke to my diabetes nurse about it, she told me that people with diabetes and other like chronic health conditions are a lot more likely to suffer with anxiety and I never knew up before but it just makes sense because obviously I'm trying to make hundreds more decisions every day than someone else's. So it's it's not really that surprising that weighs on your mind a lot. And you're worried because one unit either way when sorry, one unit of insulin can like completely tip the scales, and you don't want to end up hypo because it's scary. And yeah, it's just awful. You don't want to end up high because you feel rubbish. So there's no surprise that it can cause anxiety. And I guess the other one was, I got married a couple of years ago, and they always used to ask me at my diabetic clinic, you have to let us know if you're going to think about having kids. So after we got married, I sort of said to them, it's probably on the cards at some point soon. And they were like, Oh, you have to go to this - I think they called it a pre Pregnancy Clinic. And I just I never expected that they don't tell you about it when you're younger. Because you know, if you do get pregnant, it can, for whatever reason, I don't really understand but it makes your blood sugar's less in control, you have to go to this clinic to get your control really tight. And you have you have to go to this different clinic until you get pregnant. And until you have a baby, which is crazy. I didn't know that you had to do that. Because no one tells you that when you're six, and you are told that you're diabetic, so.
Ferdouse Akhter 11:15
So how has the insulin pump changed things for you recently, like going from the pen to the pump?
Melissa Connolly 11:20
I think the most obvious change is that you don't have to do injections anymore. So whereas when I was doing injections, I was doing at least six injections a day.
Ferdouse Akhter 11:31
After every meal, kind of thing?
Melissa Connolly 11:32
Yeah, yeah. So I was doing an injection when I woke up and an injection when I went to bed and then an additional injection with every single meal or snack that I had. And now that I've got the insulin pump, I touch it every three days. So I only have one tiny injection, I guess every three days. And that's amazing for someone that's done five injections a day for 25 years. You know, if I if I wasn't doing so much running, it would make life so much easier. Because when I do nothing for a day my blood sugar's are so steady. And I never got that before there was so much ups and downs regardless of the fact I wasn't doing anything. It's a little more challenging when you're trying to exercise but yeah, it's it has been a massive change for me. And it's been great.
Ferdouse Akhter 12:19
So Ken, how are you working to make Melissa's life even better, like the insulin pump is already great, clearly, but you're working on something to make Melissa and everyone else who's got diabetes - their lives even better.
Ken Li 12:32
Yeah, my job is generally try to help people maintain their glucose level in a tabular range by using different like wearables, including like CGM: continuous glucose monitors, and insulin pump and wearables like a smartwatch etc. So there are several ways to do that. The first way is to do a glucose prediction. So if we can predict the future glucose level, for example, in the next 30 minutes or one hour, it will help people to intervene in advance to avoid or prevent hypoglycemia, hyperglycemia. For example, if you predict that you will experience hypoglycemia in the next 30 minutes, or you can do some pre acts proactive actions, like to have small snack - something to avoid that. Another way to achieve this is to develop a personalised insulin advisor. So generally people need to estimate the hypo intake and then use a formula to calculate how many units of insulin you need to inject along with meals several times per day, but later or in the future, or we can use this personalised insulin advisor to do this calculation. So you don't need to do this estimate by yourself because normally, sometimes it's not accurate. And yeah, this personalised or individualised tool can help you to achieve that.
Ferdouse Akhter 14:00
So is it kind of like so you'd have a smartwatch on your wrist? Most likely your wrist, right? Yes. And it will kind of like monitor your blood glucose levels and then that the AI or the thing inside your smartphone will be able to tell you hey it's too high or too low.
Ken Li 14:17
Normally now people use CGM continuous glucose monitors to monitor their real time glucose levels. Well, CGM side is a continuous glucose monitor. Yeah. So it's a small device that you can wear on your belly or under your so it can measure real time or instant glucose level every five or three minutes.
Ferdouse Akhter 14:38
And does that like send it to an app?
Ken Li 14:39
It will send transmitted the information to an app or to your like a smartphone so you can monitor your current glucose level in real time. That's very convenient. And for smart watch. It collects physiological data from your body, like your heart rate, like accelerate metre to measure your actions or activities like sports or like exercise, etc. So, we will incorporate or integrate all this information as input to smart algorithm or AI enabled algorithm, then this algorithm will calculate how many units of insulin do you need in the next few hours. So they will send this information to the insulin pump and insulin pump will deliver the dose of insulin directly to a body. So this will form a closed loop system we call it artificial pancreas.
Ferdouse Akhter 15:34
Oh, I guess we should have mentioned this at the beginning. But like your pancreas is the thing that produces insulin. Oh, yeah, yeah.
Ken Li 15:42
So diabetes is because your pancreas was damaged. So your pancreas cannot produce enough amount of insulin, or your FrankerZ does not work properly. So
Ferdouse Akhter 15:55
it's like this little thing that sits on top of your liver or kidney or somewhere in your stomach are or something?
Ken Li 16:02
Because this area.
Ferdouse Akhter 16:03
Ken's pointing, like rubbing his stomach. And the left side, that's what you were rubbing, right?
Ken Li 16:10
Yeah, under your liver.
Ferdouse Akhter 16:12
Okay. So that's where it is. If you guys wanted to know,
Melissa Connolly 16:15
I never knew that. Yeah. Because mine doesn't work. No reason to know that.
Ferdouse Akhter 16:21
Yeah, I didn't know where it was going to. So how does this all sound to you, then?
Melissa Connolly 16:25
Yeah, I think it sounds it's amazing. It would be life changing for so many diabetics, I think, like one of the challenges with currently, I'm using a CGM separately to a pump. And the human error is the biggest challenge that I have. So for that to be removed, would just be amazing. Like, for instance, if I eat a jacket, potato for dinner, I will deliver my insulin for that jacket potato 10 minutes before it's ready, and I start eating. But then if I get halfway through, and I don't want to eat anymore, I'm stuck with that insulin on board. So to have something that could stop that from happening would be amazing.
Ferdouse Akhter 17:05
So you have to like really forced that jacket potato down you, even if you don't want to eat it.
Melissa Connolly 17:09
So yeah, you'd like how, how am I going to sit in another half of the potato, it's not like it's a small meal. But if you don't Yeah, if you don't eat that, then you're hypo. And actually, that just means that you have to eat more, because you have to have sweets or Lucozade or more carbohydrates. And it's just a bit of a bit of a nightmare. And the other thing about the sort of algorithm is, say, for instance, this morning, like I woke up and I planned to go for a run before I got the train to come down here. And my sugars were too low. But if I could have an algorithm that predicted that I knew that I go for a run in the morning, and then I didn't need that, or I need to stop my insulin or lower it like it would be amazing. It would be so much better than waking up and having to change my plan for the whole day because last night's insulin was a bit too much. So I just Yeah,
Ferdouse Akhter 17:56
So you didn't go for your run this morning.
Melissa Connolly 17:58
I did. But I had to leave later. And I could go as far as I wanted to, which actually was quite nice. Because yeah, he wants to run that far anyway.
Ferdouse Akhter 18:07
I was going to say - in this heat.
Melissa Connolly 18:08
Yeah, yeah, way too hot. But yeah, it's just yeah, had to change like the plan of my whole day. So yeah, it would be great not to have to do things like that.
Ken Li 18:17
Yeah, exactly. I think those are two ways you can tell the algorithm that you're going to do exercise in the morning. One way is that you announce it in advance saying that in the next 20 minutes, I'm going to take an exercise so the algorithm will automatically calculate the how many insulin Do you need something like that, another way that if you wear like a smartwatch, or like other wearables, and this will record the activities in real time, so they will transmit or send this information to your algorithm and your algorithm realise that okay, you are doing exercise, so automatically it adjusts the number of units you will inject in the next few powers, for example, yeah,
Ferdouse Akhter 18:58
so cool. So it really is your artificial pancreas, but you don't have to think about it like you do with your actual pancreas.
Ken Li 19:05
Yes, yes. But also there are a lot of other challenges. One challenge is that the inter personal or intra personal variability of glucose is very large. I mean, for for different people, or even for one person during different days. That variability can be very large. One way to cope with that is to train a personalised individualised model by using your own data. Basically, we have built like a general or universal model by summarising a lot of people's data, but that one is just a universal model. It's not for your habit or lifestyle. So then we can use your personal data recorded, for example, in three weeks to fine tuning the universal modal to make it like customised or like very personalised, and that's why we call it transfer learning. is a technique in deep learning. That's the way we try to refine tuning the model.
Melissa Connolly 20:06
Yeah, I do think that's really cool because I, you know, when I was getting my pump, I've got my little handset and they say work out how many units I would typically need per hour based on what my sugar's were for the last week. But that's not what my sugars are always going to be. And it doesn't always work. So like when I was on holiday, it was a bit hotter, and I probably needed a bit less. So I was constantly having to fiddle with it. And instead of changing my background insulin, just working out how many units I needed less when I ate when I needed to, like turn down my background insulin and things like that. And if that was just all done, for me, that would just Yeah, I keep saying it would be life changing. But it would be it really would be life changing to just not have to think about that sort of thing all the time.
Ken Li 20:49
Yeah, exactly. Exactly. Is the core of our work. I think technology needs to, you know, make people's life easier and happier, not, you know, add more like extra burden on people's shoulder.
Ferdouse Akhter 21:01
Yeah. So I guess for a lot of people listening at home, especially those with diabetes, this sounds super promising, and they want it in their hands tomorrow. But where's this technology at right now? At what stage are you at?
Ken Li 21:14
That's a very good question. We're developing the algorithms now in our lab, and many other labs in the world. And we're confident that it works. However, because it's in the healthcare sector and healthcare sector is a sector that is relatively conservative. So we have to we have to go through a lot of like, procedures to make it open to public use. For example, the MHRA,
Ferdouse Akhter 21:42
the UK regulating body. Yeah, I don't know what the acronym means. But you can Google that.
Ken Li 21:49
Yeah, they just announced the last year, I think they will regulate AI algorithm and how the software works just as normal medical devices, it means that if we want to use it, we need to go through several clinical trials to ensure that they are 100% safe. So I think that's a good decision, but actually there have good or bad. The good side is that we can ensure that the product or the algorithm is very, very safe. But the bad side is that we need to wait longer like yes, because normally clinical trials takes a lot of time from like applications approval, patient recruitment, conducted the trials, and analyse the data report. So you have to go through all of these procedures. And then you have to not only one trials, but several trials to show that your algorithm is very safe. That can take years. Yes, I mean, at least like two or three years. Yeah. Or even longer. Yeah,
Ferdouse Akhter 22:56
yeah. So with AI, I learned this very recently off another podcast, by the way, our podcast with Peter, that AI data comes from what's already there. So if you're feeding, if you're feeding AI with data from white males of the age of 40 to 50, you're only going to get personalised data for that subset group. So how does it work for someone like Melissa, or like my dad who don't fit those categories?
Ken Li 23:24
Yes, that's a very good question. Because that question relates to like the bias and ethical ethics of AI algorithms. There is saying AI will called rubbish in rubbish out. It means that if your data source is rubbish, the model you're producing rubbish. So you need to ensure that your data source is unbiased. It covers all the population, all the ethnicity, and all the ages to ensure that your model is unbiased. Yeah, that's very important.
Ferdouse Akhter 23:58
I guess that's where the regulation the recent regulation, government regulations come in. Yes, yes. Yeah, to help it what happened before then they haven't before regulation or like regulatory systems came into place.
Ken Li 24:10
Actually, there are many existing algorithms already in the market or online produced from different like companies or even big companies, you name it. And many of them are using like a private data set or like some data set are not public accessible. And they claim that they are unbiased, but you need to go through a lot of like assessment to ensure that they are unbiased because some people find that they are biassed for like minority group of people. And as a big names realise that then they update the system was algorithm a cracker that
Ferdouse Akhter 24:50
is, I guess, like regulations take a lot of time because you have to go through so many different processes, but in the long run, it helps people helps, especially ethnic minorities. It make sure that their data is also being fed into the model. I agree. I agree. Yeah. So can What is the future then hold for diabetes treatments?
Ken Li 25:10
I'm personally I'm very positive actually. Yeah, I think because now the technology is developing very fast. We just got the news or from like Chat GBT for like, related like technologies that blow people's mind. And a lot of people are using it, and people call it the iPhone moment of natural language processing or AI technology. Personally, I think the similar iPhone moment will happen in the diabetes management in the next 10 or 15 years. I think there will be some very powerful like a CGM or insulin pump that can be developed in the next few years that enable people who have diabetes can lead very, like normal life, like normal people. Happily, I think, yeah, that's achievable in the next 10 years,
Ferdouse Akhter 26:05
I think that's really positive. And I think that's a really nice sentiment to wrap things up on - that those with diabetes and other chronic disorders aren't monolithic, that personalised help using ai - I know AI kind of gets a bad press here and there. But AI is really useful here that you can help you with personalised treatment with people like Melissa and others with type one diabetes and lead a more I don't want to say fulfilling is the right word with leader more
Melissa Connolly 26:33
easy, easy yes.
Ferdouse Akhter 26:36
is that the right word is easy's a good word, lead a more easy life without having to second guess what's happening with your body, without having anxiety filled days. So thank you both for speaking to us today. So thank you, Ken, and thank you Melissa. Health in a handbasket is produced by UCL's Institute of Healthcare Engineering and edited by Cerys Bradley, the Institute of Healthcare Engineering brings together leading researchers to develop the tools and devices that will make your life better. We're using this podcast to share all the amazing work taking place. You can learn more by searching UCL health in a handbasket or following the link in the show notes. So share with your friends and family if you found this interesting, were available everywhere, especially where you just listened to us