#456: Half of babies with heart problems are missed - AI in ultrasound is changing that

0:02Can we fix this postcode lottery where some places you might end up with 80 % of kids not being picked up? My own daughter had a major heart problem that wasn't picked up. Found out that, you know, the pickup rate goes up to sort of 97, 98%. Suddenly you can change the workflow, but it allows basically a second check to happen very quickly.

0:27hey ready i am delighted this week to be joined by professor reza razavi and he's the professor of pediatric cardiovascular science uh at king's college he's also an honorary consultant pediatric cardiologist at guys and tommy's uh was the vp of research at king's college as well director research king's health partners director of the king's welcome trust epsrc center for medical engineering uh resi done loads of stuff you're also the ceo of freya uh a very interesting fascinating in fact um ultrasound software company leveraging ai to help uh obstetricians and sonographers with a lotus tools that can reduce scan time by 42 apparently which is incredibly significant and improve the detection of congenital heart disease um something very close to my heart actually as well um i remember my time doing neonates very fondly i talk about it on this podcast a fair amount actually it's the the placement that really kind of sticks with me um since leaving medicine it's the one that i think about more more than any other specialty but I'm sure we'll come on to it.

1:43Rosa, it's a pleasure to have you on the podcast. How are you doing, sir? I'm very well and it's great to be here. Excellent. Whereabouts are you speaking to us from today? I am at home today at the tube strike. So anyway, so I'm working from home today but it's actually nice because it's more relaxing in terms of doing it. Well, indeed. Indeed. Absolutely. Exactly for this sort of thing. You've obviously achieved a huge amount in your medical career um you've done lots in and continue to do lots in the hospital lots outside the hospital and obviously now ceo of a software company as well um i'm intrigued if you were to start at the beginning and tell us that story where does all of this begin so yeah i trained initially in pediatrics and then in pediatric cardiology um it was an area which of all the sort of specialties i rotated through i found more sort of if you're like comfortable with of chined with me because we got very sick babies but made them better and then you would see them as well kids and so that's actually very rewarding but also it's quite a technical specialty both in terms of being able to work out what's wrong with them because there's so many different complex conditions you can have as a congenital heart problem you know more over 50 different sort of permutation it's not one disease and actually trying to work that out requires often fancy imaging, which I've helped develop over a career, which now is mainstream, actually, things like using MRI.

3:29And also, in a sense, for me, it's a very multidisciplinary sort of specialty, which also is good, because I think I like the idea of working in teams and working together to get good outcomes. But it's obviously a super specialist specialty and a condition is relatively uncommon. So only one in 100 babies are born with a congenital problem, and only about half of those have a very serious one. So although, obviously, there are a lot of babies born, these things are relatively rare conditions in the newborn babies. And then, of course, the way it's delivered in the UK is given in a few specialist large centres.

4:17And I now, you know, I've worked actually in the same place for 35 years, but it's now the biggest center alongside Garongman Street because it's recently joined up the Evelina, which is where I work. And the Royal Brompton have joined up, just recently come into one site and one unit physically, although it's been one hospital for a few years. And so it's, you know, I mean, we provide a very high quality care to babies from all over the southeast of England. So that's my clinical journey. And actually, early on in my career, I became very interested in research. So, you know, did a research higher degree doing, you know, sort of taking time out of clinical work to do, you know, a research full time for two years.

5:11and then after that went on to an academic route so basically i've been employed by the university all of my clinical career as a sort of consultant with an honorary position in the hospital to draw basically doing half my time doing research although it's always not half because for many years i did pretty much a full-time clinical job alongside i was gonna say you end you end up with a 200 percent role rather than a 100 percent and uh and then like later sort of i'm actually pretty much earlier in my career but i ended up in leadership roles partly because i realized to for us to be able to do really effective research you needed good infrastructure good resources and that required you to be engaged in in how the university and the hospital deliver clinical services and research.

6:08And it also enables you to have a broader reach towards more people, enabling not just your own research, but others helping others to achieve their goals. And that's been something that I've done for many years. So you mentioned some of the roles I've had. But going back to babies with congenital heart problems who actually can present super sick at birth because basically there's two ways they come to us one way is that they are diagnosed during pregnancy and every mom had a scan of 20 weeks so if something picked up on that scan they get sent to a specialist unit and we have the biggest one in europe a fetal cardiology unit where people scan the pregnant moms but they're sort of people like me and I've done it before myself or don't do it anymore but it's basically looking in the baby in the womb around 20 weeks and in fact these days we often do it even earlier around the 12-week scan looking at the heart and blood vessels and seeing what the problem is and so that means that those parents know about the problem we chat with them they can if this kind of requires surgery they meet the surgeon they go and see the intensive care and the ward actually also importantly they make a decision about whether they want to carry on with the pregnancy because some parents decide the baby's got a major major problem and we they fully explain to them what the consequence of that is about the pregnancy but basically that's to the way if you like the gold standard of care and those babies then when they're born they get born in our hospital and literally for the minute they get delivered and as the junior doctor i've had to do this but also as a senior doctor you're there as the baby comes out to provide them with literally you know very specialist care from that moment and sometimes they need an urgent procedure really very soon after that but certainly drugs and then plan they're already having planned their major surgery because you know when they were born therefore was planned and so they can then have really good outcomes.

8:25The other way that they present, and unfortunately in the UK, this is half of them presents the second way. So it's only one in two chance of this happening is when you don't know about them. So basically they're not picked up on those scans. And then they get born in their local hospital, wherever that is. And the thing with congenital heart problems is when a baby's born it's not like if you have a sort of a cleft lip or you know a problem with the abdominal wall you know something that's very obvious a short leg when the baby's born you don't see anything and actually after birth for the first few hours days of life sometimes because some of the blood vessels and holes in the heart that are open inside the womb are still open you might not even notice anything by all the baby checks that are done on babies and these babies get sent home so quickly after birth so that it might not be till they're home and then suddenly the babies get very poorly or it can happen in hospital but basically nobody knows what's going on and you've got a sick baby you know who's not feeding who's uh maybe blue who's and then the The thing is, there is no specialist in those local places, in the A &E or in the neonatal unit who knows what's wrong, who can diagnose what's wrong.

9:50So then what happens is, you have to get in touch with our unit, we send an ambulist, the specialist, the ambulist team, the specialist and the health care team to go over and retrieve the baby, that baby, and bring it back to us. And those babies can be very sick by the time they get to us. They're even sick when, of course, the team arrives. So they're intubated, ventilated, you know, the sort of things you used to do. And so their outcomes are quite different. And there's lots published on this that babies have a longer stay in, for example, intensive care. They have a sort of more difficult ride.

10:26They even have actually a worse survival rate. Although these days we've got very good at looking after these babies and survival is excellent. but you know latest figures are if you have an antenatal diagnosis maybe it's one percent on average across you know not surviving uh if you don't three percent so it's actually you know three times as much and but the thing is it's not just you know even if they're sick and we get them through it and they have surgery they survive of course the great majority do what we find is those babies who haven't had an antenatal diagnosis who have been very sick you know during that initial birth surgery end up with long-term problems and i see them in clinic you know the ones that i've looked after when they were babies who were very sick and i find you know in the conversations with the parents they're struggling at school there's sometimes behavioral issues and yeah and over the years now we've developed a real research program to look into this because initially we yeah it was great that they survived and they were well in terms of walking into your clinic room but you start worrying about these other things and what we find is actually if you image their brains their brains are affected by being very sick yeah and that that consequence of that isn't necessarily cerebral palsy which is very obvious but these learning difficulties and behavioral difficulties which are just that you know hugely impactful for these children, you know, for the rest of their lives.

12:02And so that's, if you like, in a nutshell, the problem that I was hoping to address through our research and now through Freya, the company, is can we fix this postcode lottery where some places you might end up with 80 % of kids not being picked up and some places actually 70, 80 % are picked up, but on average it's 50 % not being picked up and having less good outcome and I suppose one thing I might add um is that it's always you know your patients are really important to you and when you see them not doing as well that obviously impacts you but for me there was a sort of dabble whammy in that My own daughter, Poppy, who is now grown up but was born in London, in a major teaching hospital in London, having had all these sort of antenatal scans, had a major heart problem that wasn't picked up.

13:09and that was very traumatic because being a dad and a mom my wife and i who my wife's a doctor as well suddenly being in a position that you are having to deal with it as parents was you know even more harrowing and impactful and we actually unfortunately poppy was the same as many of the other kids not diagnosed she was very sick because she wasn't antenately diagnosed in fact but she was on what we call ECMO, which is basically a heart bypass machine that keeps you alive. But imagine a little tiny baby for about a week just to keep her going. And, you know, the chance of survival, very slim, but she survived.

13:56And so for me, being an academic, a researcher, I mean, you do research a number of things, But this became really very much the top of my priority list. And I was lucky enough with a great team of engineers, computer scientists, to get a big grant, a£10 million grant at the time, this was back in 2009, 2010, to fund a research, a big research program, which ran over seven years called iFind, to try to solve this problem. And when you get a substantial amount of funding, this was from the welcome trust and epsrc you know huge thanks to them and the great support they provided we were able to explore multiple avenues of how can we do this better and improve the outcomes of children in the future and in effect that's how we came up with the research that led to eventually to freya at the company thank you reza for telling that story that's really really kind of you to share that about your daughter as well i know that's um it's obviously a very personal story i think just just listening to how to how you've uh how you've told that story one thing that i've written down here is um that it's very it i find it very different hearing the origin stories of a company when it's a patient story there's a there's a level of like motivation i think behind the desire to solve the problem that it transcends something i don't know whether that something is there's a lot of problems that you encounter being the CEO of a tech company let alone a health tech company and you know there's a lot of problems that you don't want to face or that you find difficult to face and I think where you've got where you've got that motivation and desire like you have actually not only from being like you know the patient side of it but actually the way that you told the first you know 80 % of that story the clinician side of it there's this there's this I think desire that i hear in your story or foundation to your story that's so rooted in what the actual problem is for the people that feel it the sonographers the obstetricians the neonatologists the patients that you've lived it you've lived the problem and what one thing i actually thought was quite nice in your story was that you know you playing on the research side as well as the clinical side You've clearly got a part of you that wants to not only solve the problem as it is felt today, but you want to change tomorrow so that we don't face the same problems again.

16:43And I think that's quite common to entrepreneurs and people that end up in, you know, C-suite positions of startups, particularly health tech startups, where, you know, we're not exactly fintech over here with, you know, an exit every five minutes. do you know what I mean like we're we're trying to do a thing in a space that's very you know difficult to do that thing which is you know as much a liability as it is a gift I think in in at least we've got some we've got a rock to keep pushing uphill and that's what helps us wake up every day but it's such a meaningful rock and it's a meaningful mountain that we need to do that and I don't know I I just think it's really nice that that you've got all of that in your story and I feel like I've got a really nice grasp of who you are and what you're trying to do and where this company comes from um that you're building now which i will come on to but one thing i do actually just want to mention as well um it reminded me you talked about the confusion i'm talking specifically about the problem that you're trying to solve here you talked about the confusion you know dgh district general hospital that doesn't have specialist you know level three neonates or any of that stuff going on in the hospital you know they've got a labor ward and and this this type of thing might be going on and there's that confusion about what why why is the baby physiologically like this like that I get it and I get it because I can remember when I did neonates and bearing in mind I did my neonates in a level three unit so I was used I was at the end of my neonates placement so you know had a you know a couple of months behind me of doing this stuff and seeing this stuff so by no means an expert in dealing with it but you know knew a little bit and I got called to see this baby but I got caught every now and again used to get called to see a baby that wasn't high risk for anything it wasn't suspected sepsis it wasn't meconium it wasn't caught around the neck it was just a baby born on labor ward everything was normal apart from something which you now needed to go and see so I wasn't actually that vigilant you're not that vigilant to anything seeing you know a baby like that and as you would be that was you know 20 something weaker anyway got called to see the baby and they were tachypneic so breathing quickly tachycardic and um as you as you say like looked a little bit say a little bit blue or a little bit gray and i did all the usual stuff and i just couldn't get the sats monitor to work properly because the sats kept coming up in the 70s and i was like there's no way this baby sats are in the 70s so i kept i just kept i tried different different limbs and like just tried to just try to get this thing to work properly anyway both parents were anesthetists and they're looking at and they're looking at this going yeah the sats monitor's not right but they're also like this baby's unwell this isn't normal um you need to fix this get like get another sats monitor so i'm like i'm getting three or four of these sats monitors and i'm like trying that anyway they're all starting to say the same thing i'm like do you know what there's actually something wrong here i go upstairs speak to the consultants on neonates there's something going wrong here the sats are low they're like no there's no way the sats are low so we're all we've all got this bias that this is a healthy bait this is a healthy baby this is a healthy baby completely normal pregnancy born at term there's there should be no problem here there's no reason for it to be like this so i get it all of this by the way is time this is all time where there's a pathology going on that the pathology isn't being solved and albeit i'm acting incredibly quickly here because i'm start and clinicians listening you'll know this feeling as it starts to dawn on you something is actually going wrong here and then and then you're trying to like put the panic at bay and you go into problem solving mode so i'm like right i now need to convince level one neonates that this baby needs to be up here but i can completely appreciate what you're saying in that where that happens in a district general that doesn't have neonates on on site all of this is time so what you're saying about there are changes in the brain as a result this time anyway this baby did not end up having congenital um a congenital heart problem what they had was a pneumothorax well okay again where's like where is that in my differential diagnosis you know like unbelievable it was out of absolutely nowhere anyway you know straight to level one like you know chest strain all the rest of it um both parents incredibly grateful and actually i can remember on one of my last days on the rotation just to finish the story positively they brought the baby back and they were i guess we're emotional but they were very thankful um and it was a very nice end to the story anyway um all very nice but i relate massively it's tough even in a level three unit so our daughter was born in you know you know specialist level three unit in with a hospital but not with pediatric cardiology in london and the registrar had the same thing you know she was born at two three in the morning at nine where i came to take baby and mom home because she was you know she'd be a healthy happy normal baby our third child he was trying to get the satsala i understand you know and then of course she was very obviously blue that's a 50 percent effect and then they took her to the neonatal unit and the consultant just happened to be off the unit which is unusual they couldn't reach her she'd left her bleep behind and and they wouldn't do anything because they didn't have the consultant safe they were in a sick baby that they know normally yeah they're looking after so i was trying to say well you know this is an urgent thing And of course, you can imagine now our place.

22:23In the end, eventually, I got hold of her and told to them. She said, actually, asked the dad to say it was just lucky that I happened to be a pediatric cardiologist. They asked me to scan her to work out what's wrong with her. Wow. So that then I was able to organize for the team for my hospital to come pick her up. But the point was that's just chance that it was my baby. And the outcome would have been quite different, perhaps, if it wasn't. but even in a specialist unit with surgery you know even if you take your child up there's still it's very different outcome to one way there's an anti-intelligent diagnosis and then cardiology and cardiac surgery i mean my daughter needed urgent surgery uh literally you know with every minute counted a sort of keyhole surgery but that you know the team came from guys to say well should we do it here right now but we don't have any surgical cover so they rushed her in ambulance back to guys to do it as soon as she literally got through the door wow so so there is a sense that even even in the thought of specialist units unless you're one of the lucky eight hospitals where yeah the two coincided the pediatric unit on site you're still going to struggle with these babies and we need to basically the way to tackle it is to pick them up antenatally so that they get the right yes the care pathway you know we're at the time of delivery rather than afterwards yes which brings us very neatly into what you then had as i guess an idea at the time to solve the problem um or what's your what's your run into to to now being ceo of this company yeah so so with this grant would i find we tried lots of things i had lots of ideas and of course naive some of them perhaps or maybe not so naive and lots of them progressed to things but some uh you know didn't so much so one thought was well you know i my expertise was an mri and where we get three-dimensional pictures of things you know of the heart for example not in babies in the womb but in newborn babies and onwards and i developed programs in that which were sort of first program in the uk and then i set up at our hospital at the evelina then set up the program at great ormond street and now you have these programs all across the uk we've trained i've trained people who now deliver them across in every units that does congenital heart disease in the uk and internationally it's become a major way of imaging these babies and children but at the time of course that was my no so i said well let's try to get the same 3d pictures of the baby than the womb's hearts and so all sorts of ideas so for example we developed technology that instead of one probe because i can only cover a bit of the baby we could use up to four probes simultaneously oh wow so that we had them all together and so let's see if we can just do like an mri cover the whole baby because babies move they don't hold still and just lie there and let you scan them and they can be all sorts of different position backups sideways and so so but that technology and in fact we developed MRI technology for the babies at that gestation and that's actually now is a clinical service.

25:42Wow. First up the world's first clinical service again at St. Thomas at the Evelina and now it's spreading around the world but you can't MRI every pregnant mom so we do that for specialist cases where there's uncertainty about the diagnosis where we know there is a problem but to get every baby screened and pick up the ones with the problem what we found was if you do 3d and fancy things you just can't roll it out because you have to change the way people scan and this scans are done by hundreds of people across the uk in fact you know just thatst thomas's there's like 30 sonographers who do these scans and you imagine every hospital has 10 15 sonographers doing them so you can't get all those hundreds of people even thousand plus people change the way they scan they've been doing it all their lives they've worked so in the end the way and in fact it's interesting because the ai revolution sort of was coming along i mean we airs been long a long time in terms of imaging but we have it's progressed so rapidly and so we basically were able to say that sort of track really does allow us to do something right now and again we've developed technologies to you know put a drape of ultrasound across the tummy to image the whole baby with a completely new way for doing offside but those are going to be around in 10 years time we want something now so for now we found that by just using ai in real time because the computer is also really improved to be able to get the images as they come during the scan and run different ai rhythms on them is something that you could do right now because it doesn't change the way the scan's done and in fact it really helps the way the scan's done because at the moment when you do a scan you're having to stop the scan freeze the picture to save it for your report you have to stop the scan to make measurements put calipers on and think that takes about half the time with the scan and so the AI can do all of that make your life easier let you focus on the real problems can you say this baby's normal or not and because the person doing the scan is at the same time reporting it you know it's a difficult cognitive problem and in fact after that initial grant we've had multiple other grants to build this technology you know test the prototypes until we started the company about 18 months ago but during that journey with a prototype we did a trial which was published in the general medicine ai about a year ago to show that you know 42 reduction in time but and you know it made the cognitive load of the people doing the scan less make with less stressful doing the scan made the life easier for them improve the precision of the measurements and although that wasn't powered because we didn't we had only about you know 75 pregnant months there with a third of them having a congenital problem uh which we had a trend that it improved the detection rate as well because you had a more easy focused time and what we found was also the AR allowed you to do something which isn't possible at the moment to have a certain person have a look yeah and we've shown again if the ai structures this completely unstructured exam and allows the pictures to be looked at just the key ones the clips and you don't have to save these because it's automatically every single image all hundred thousand are looked at sorted and you know if you like structured organized so you can look at the best pictures of the key structures very quickly, that in three minutes, somebody can review that.

29:33And then we find out that, you know, the pickup rate goes up to 97, 98%. So suddenly, you can change the workflow, but in a way that doesn't change the way people are doing the scan. It allows basically a second check to happen very quickly and easily. And in fact, what we've now got, again, thanks to grant funding from the National Institute of Health Research and IHR is a large randomized trial, which is ongoing now in 9 ,600 mums across four hospitals, where we're randomly either, you know, doing the scan in a normal way or doing the scan, both using the AI during the scan, which sort of makes the scan much easier.

30:17That's called freya scan and also allowing this rapid second review using the ai called freya detect and looking at the two key things you know can we make the scan much faster and easier and can we really improve the detection rate and because this is properly powered now you know it's not 75 patients 9600 mums we can look at real improves in detection rate as well i think this is fantastic So one thing that I just want to pull out here for listeners is what you explained there very, very briefly. But I think this is where the principle of innovation in health tech. I just think we need to shine a light on this.

31:06So when you come in for an ultrasound as a patient, you lie on the bed, the jelly's put on you, and then the scan starts. Now, for that sonographer or ultrasonographer or whatever you want to call them, or whoever it is, actually, for whoever is doing that scan, you're absolutely right that what you do is you're moving that probe around to get a certain image on the screen. And then you decide what the best view is. And then you've got to hold yourself unbelievably still while your other arm now comes up to the computer to then roll a ball or tap a thing or use the mouse or whatever. And then the thing's moved.

31:41and you're like oh then you need to get the probe thing back again and then you've got to look at and then eventually you'll you'll align the two things click a button and you've then taken the image you might tap a few letters as well after that you're going to find the next image and find the next image for the next image but every single time you've got to do this pause capture pause capture pause capture uh pause capture data input even i would add to that so pause capture data to input pause capture data input and so as a process it is robust in the as you're taking that image you're sort of verifying okay that is a good image because i'm adding these other things to it so it's robust from that perspective but god it you're right the amount of time that that takes versus what is the actual skill of the sonography the skill of the sonographer is being able to look at the body part, the abdomen, the limb, the neck, the head, the whatever, know where you're putting the probe, at what angle, as quickly as possible, and you know that.

32:41And that truly is the skill that's practicing top of your license is just doing that bit, not the data capture bit. And you're like, well, okay, could someone else do the data capture? Well, that's expensive. And it's so funny how in the last few years as AI has made the progress that it has that more and more people would be able to look at this now and be like can't AI just do that and the answer is a resounding yes and the people like yourself have been thinking about this for a very long time and have the tech understanding and probably were in AI when it was called computer vision for this particular bit if you know you know all that sorts of stuff um god I remember going to stuff and speaking stuff and it was like that but um yes, this is that, correct.

33:27And our ability to, once you've got the image on screen, the data's already in there and therefore can we do X, Y, Z, yes. And so it is absolutely no surprise to me that when you link that idea with an amazing 10 million pound, a huge amount of money grant in order to make this happen, but combine all of those things with your story and your motivation to actually get to the core of this issue, which is increasing detection rates. that's ultimately it with the fact that because you've done this you understand the process you understand that this is how people work and again core principle of health tech innovation this knowing how people work knowing that they're not going to do any extra clicks one thing that i'm sure you will resonate with here as a is that i used to do quality improvement projects when i was a clinician that's what i liked as the sort of budding entrepreneur entrepreneur I was I love doing quality improvement projects finding an issue and solving it but my bar for a good quality because I I did it to get the amount of points I needed for whatever thing like that was ticked that was ticked off long ago I was now just doing it for fun in inverted commas um but it was my bar was is it a net reduction in work because so many people would do a quality improvement project and and you must see these all the time for your people you mentor or you know junior doctor or resident doctor sorry i should say um it's always the the output is always always there's there's now this new form to fill in and you're like oh oh really like ah that that's not that can't be a good quality improvement outcome that there's just a new there's a new thing for me to do that okay makes things but there's a new thing for me to do so whilst it might make it safer or better or closer to guidance or whatever it is it's a net increase in work as soon as this person rotates this is now out the window why what the thing that i love about what you're talking about here the innovation really that you're talking about is this is a piece of technology that goes into a system that already happens which is an ultrasound happening not only is that is this okay end of day net net net going to increase detection rates the way it does that and a huge way that it does that is actually a massive net reduction in work it's a net reduction in the pauses that people have to take in order to take the imagery which is below their license that is not work that they should have to do so now they don't and how does that work well actually that what's the second order consequence of that well it's the reduced cognitive load and what is the third order consequence of the reduced cognitive load it's the increase in detection rate because a it's getting the best image at the best time through technology and ai but b they're not cognitively reduced exhausted on the route to burnout so they don't miss anything they are more likely to practice better in inverted commas and get the best imagery because they're not rushed they're not exhausted they're not tired and so it's so interesting isn't it that there are in healthcare these these wins win win win win win and then you get this flywheel and it's and it's fascinating to me that the way that you told the story and right back to the way that you started i'm a clinician i did this i do this i get it and i just think it's great how all of this has come together no but i think you make a couple of really important points that are worth you know emphasizing one is that actually a lot of this to do with workforce and of course we need to embrace the training of the workforce and have them but the workforce is on a lot of pressure you know these sonographers are doing these scandals they're doing some of the gynae cancer scans and also the further scans under the real shortage of them in the and And these days it's pretty hard, particularly as we get increased, you know, body habitus in some of our pregnant mums, you know, physically hard on the wrist and on the neck.

37:36And, you know, so, you know, many are having to only work part time. Many are actually leaving. So there's a real big issue with retention. And it's stressful because if, you know, on average half are being missed. imagine every time that happens the sort of feeling that you have oh my god you know didn't go right i did my best but it didn't work out and that's that's not a good feeling and so actually where there's shortages there's a real problem getting something where the workflow is makes it if you like the process much easier and less effort and to be honest less cost which is also really important in our health system actually you know having you know being in sort of senior on the board of one of the you know our biggest hospital trust for about eight years it's very clear that we want innovation to come to improve outcomes but it's hard if that means more cost because we're so cost risk constrained you know we're so financially constrained in our health system with big wait times.

38:41And that's what's basically we're being judged on at a hospital is, are you needing your six-week wait or your diagnostic waits and all of the other things? As a board, as leadership of a hospital, that's what you're looking at. Nobody's looking at what's your detection rate for congenital heart disease or other congenital problems in pregnancy. That's not what you're being judged on and the chief executive might lose his job on so if somebody says oh have a new technology it's going to cost you so much it'll improve your detection rate great that's a really good thing but if it's going to cost me more money i'm so constrained about money and that means that actually it might even increase my wait times and things because i have to take money from somewhere some activity will have to happen there that's that's a problem and i think that makes technology hard to get into the work you know mainstream of clinical workflow and particularly in the national health service but if you can deliver a technology that improves outcomes but also reduces costs reduces effort improves retention and you know the workforce challenges that we really have then to be honest that addresses two key things and it's much more likely still might not get adopted but much more likely to get adopted and sorry once i'm just finished writing these notes gosh i wish i can talk to you about um but just going back to the point you made about the clinical background actually you know i'm sort of the specialist who sees these babies when they're picked up but in our team both grant and since then we have had a number of sonographers who do these day-to-day scams full-time involves as part of the research team and to be honest it's them there's a lady called jackie who's our sub chief medical officer who's one of those sonographers but it's done a phd now and you know it's amazing but it's people like that who are at the if you like forefront of doing these day in day out on normal pregnant mums where the challenges are who have driven how our technology has been developed you know our advantages have come from that's probably the one of the most important one is is is that strong clinical uh problem ethically know-how and understanding and trying to really solve it not just from a you know ai yeah your academic point of view or even a pure commercial point of view but also from a sort of real true deep clinical point of view but the second i think key thing is with that with all of these things it's the quality of the air so as you know some of the major ultrasound companies like g and samsung have some ai technologies now in our machine for obstetrics and they've developed those and but they're not used as much certainly some of them are not used as much and i was on a call with the dutch um uh if you like public health body who looks after of these screen programs in Holland just earlier this week, yesterday.

41:45And they were saying, we did a prospective study of those technologies with sonographers, and we found that actually they weren't good enough, really. Although these are CE mark products that are sold, and we're worried about that, and we want to warn our sonographers if something isn't really, you know, doesn't mean that anything with a label AI is good enough and should be used. and so and then we did we talked about our technology because art is just coming onto the market so and so uh and the evidence we have about how good it is which is prospective studies in patients and where also we've done studies with users you know and again in our big trial that i was talking about a big part of that is qualitative sort of studies talking to the personographers interviewing them but also doing surveys and things and what the reason maybe our technology works perhaps better is because of with ai it's not just about how the product is made to look and how easy it is to use because it doesn't require any buttons our technology work with any ultrasound machine you have to buy a fancy new one it's very easy to use there's no very intuitive to use you know when we did our previous studies for example it's a five minute training to be able to just get on and use it and show that big reduction of scan type.

43:08But also, the AI has been developed on huge amounts of data, much more than others have done, because prospectively, as part of these grants, we collected data, full scans, you know, 100 ,000 images per scan on tens of thousands of months. That allowed us, if you like, to build AI, which is very robust and works. and that's one of the biggest challenges that's been in the of AIs come into healthcare is when it doesn't work well clinicians lose confidence and then they don't use it and then you don't get the value that you could get it's like anything you know when something new is being developed lots of things quickly come on but people are more worried about quickly getting something out because they want if you like the marketing benefit and you know first mover but actually you've got to get something out there which really works well and i think that's that's also a key challenge that we've hopefully trying to address it just feels so robust i think that's kind of the word that comes to mind everything that you said about the way that you've conducted you know yourself the way you conducted your studies the way that you've brought the right people on that were clinical that kind of understand the workflow like it all just feels very robust and you know i actually had a conversation um prior to recording this podcast actually with someone that said hey like you run so x like i've seen your personal brand stuff i know the podcast like i followed you for ages we as a company want to build trust with the nhs as quickly as possible um how do we do it uh this is the budget that i've got to spend and it's and it was a really interesting frame and challenge to be like well actually what you need to do first is care about about all of that but you need it's not care about getting the trust you need to care about the thing that that enables you to get that trust and so I guess what I mean by that is you care deeply about the patients you care deeply about the process you care deeply about reducing the scan time about reducing workforce issues and you care about increasing uh the detection rate so like these are fundamentals to what then becomes well actually all of our company activity now is contributing to these types of things and the fact that you're then doing those things is what builds the trust so you telling that story of like these are all the things that we're doing and have done and this is why that's what builds the trust it's like there's no budget in the world of things that you can set trust is is is something that you show or or the building of trust is gained by things that you show not things that you say in my opinion it's the same with people you know you if someone shows you who they are believe them don't believe what they say it's that kind of thing it's like i've got an 18 year 18 month old son and dads that are you know further down the line than me are saying that you know you can tell them anything in the world you can tell them not to smoke but if you smoke you can guarantee one day when they come home in their 20s they're going to be smoking and so it's that for me it's like you've got to show this as a company with your actions and actually if you've got a set of values then behave in accordance with those values and the trust will come and I think that's that's what I'm I'm you know getting from there but you're absolutely right trust is absolutely huge um the tech is interesting now you've talked about you've talked about the uh the technology there and how it integrates it's interesting to me that it's you know machine agnostic actually and that this kind of just layers in however you know with with with any machine how do you have to install it is it is it something that just sort of lays over like how how does that practically work so that's one of the challenges in getting new technology imaging is because someone with a usb stick you just deploy them into hospital is because you know you have to work with the manufacturer of that imaging device particularly something you should use in real time it makes it hard to and so we had worked with a couple of big companies in modifying their ultrasound machines so that we could get the data in real time and then be able to process it but again that was hard because that's a bespoke ultrasound machine that you know it's not standard common everyone they're sitting everywhere so in the end what we found was a great solution is actually let's use what ultrasound machines already do every ultrasound machine certainly ones used for obstetrics already spits out a video of its images in real time because moms like to look at the picture of their babies and they can't see the screen the person doing the scan so you have to have a little separate you know tv for them to look at that while the scan's happening so there's an hgmi out for every ultrasound machine so we just pinch the pictures from that and that means that we don't have to do anything to the ultrasound machine and basically then we take the data to a tiny little computer that has a you know nvidia gpu because a lot of computers what doesn't have an nvidia gpu these days so honestly and then it's basically a little uh tablet screen you know uh that allows you to the sonographer to interact with the software but also importantly we found for example we tried lots of different things and you know competitors do this where they show lots of pictures with the ai you know put onto it but that's very distracting we found we found the best thing to do is not to you know the only thing we show is something that you can briefly glance at which tells you which views you've collected and you know yeah just the confirmation that you've got the images minimum amount of information you need and your measurements where they are on a centile thing and how full confidence the thing is and we do the way that we do these things we develop new ways of doing it for example we realize people there's a lot of uncertainty about measurements And in fact, with AI, it can even make that worse with publications to showing that.

49:27So what you need is, again, use what you have. So we've patented this. When you do a scan, you don't take one look at a picture. Of course, if you freeze it and save it and then do calipers on it, you look at one thing. Maybe you'll do a beat that twice. Most people don't. Rarely three times. You're supposed to do three times, but nobody does. if you're not doing that and you're just scanning and you're getting every image the ai is doing it you're getting hundreds of that picture thousands even yeah so why not measure every single one the ai doesn't mind yeah it's all happening in real time and actually when you do that and you use bayesian statistics you get a very precise accurate single measure made up of hundreds of thousands of measurements that is reproducible much more than a better than a human again let's take advantage of the way that the AI can really help you and get the improvements that basically come because technology can basically repeat something multiple times without being tired or worried or things like that.

50:36So we do lots of things in a way that it makes it really easy. And for the person who's doing the scan is not distracted. They're focused just on doing the scan in real time a baby's moving the ultrasound probe's moving they're looking at things they have to cognitively look at it and work out if the problem or not do i go back look at that we just leave them to that well and they have to manage the patient as well to some extent i imagine as well because that that is a room probably full of anxiety for some people and you know they have to manage their own emotions of what they show to the patient you know as i say i got an 18 month the old son so you know this is very fresh in mind you know these these moments unfortunately you know he's very healthy and everything was normal but you know you go to those meetings and um you know someone described i saw it written down incredibly the other day someone someone described they were they were they were in a waiting room and they and they described it as like like they had schrodinger's health at that point you know that that in that moment you are simultaneously completely healthy or dying frankly and it's to say and that feeling of you know this scan is is showing me a real fork in the road for me and my entire life i've never never felt that more acutely than in that moment and thank goodness you know that the sonographers in that moment that i didn't realize how fast they were like as soon as you come into the room i don't know if it was this one specifically but i thought it was wonderful as soon as you go into that room it's it's it's lie down belly exposed gel on thing yes there's the heartbeat baby's healthy it was all within like 60 seconds i was like wow but it's all part of the management is what i was going to say like even that it's all you know another thing that they're having to cope with at the time and and do that they're freed up to do you know no absolutely i mean you know the early scans where they look to see if the baby's you know heartbeating thing which is great because you're worried about that as mums and that but those laser scans you know 20 weeks where they're looking to see if my baby going to be born with i don't know missing eye or a short limb or of course a heart problem or a lung problem you know you're worried about every organ because you know that something could go wrong it's rare but it could and of course it's such a huge relief when they say oh no if everything's fine the thing is that they could say to you everything is fine and it's not that's even in some ways worse yeah freya detect i just want to talk to you about this because i think this is really interesting um the automation of the secondary review is this what turns this kind of momentary ai and value and i say moment you know meaning that whole consultation is that what turns it from that momentary value into more of a service well it's interesting because if you look at uh how um a lot of screening tests are done be it histopathology to look at cervix smears or mammography you get two people to look at it and you know there's a real benefit to that and we know from imaging if you're looking at something real time where you're focusing on trying to get a picture and at the same time trying to basically report it in real time in your brain as hard.

53:55Well, if you're not doing the first part, the best images are just being shown to you, but not just a tiny bit of the best images, all of them are there, then you'll do perhaps even a better job, not just because the two people looked at it, you improve detection rate, but also because of the way it's done. And in addition, we have developed and are developing AI models that can help, if you look for potential problems in the images. So if you look for the abnormalities, now these are harder to do. And again, we're testing them in this big trial. But that can basically advise you, look more carefully here because we think there might be something wrong.

54:38You know, you're a little bit of an extra sort of help where they say all about the human AI interaction. Now, we thought actually as a company, when we set Freya out 18 months ago, when we were first setting up, oh my gosh, this is great. We'll have this software, the medical device in real time is happening and then we will provide a service where we can do the second reviews very rapidly and people will just come in, you know, if you like, in effect, have that to improve their detection rate. But the problem is, it doesn't already happen in the UK and in many European countries. In places like the US, there is often a second review but it's only a name in the sense that the photographers can't sign off their own reports so a radiologist or a sort of obstetrician has to sign it off to be honest they don't really look at the pictures or anything but sometimes they do but the point is when there isn't an existing workflow it's hard for us to provide a service because you have to employ people to do these second of course yeah and you have to have cqc and all of those other things and do it for each country because you know providing service in each country is different so in the end actually we pivoted we said we'll develop all of that software and we have we're testing it in this trial but we will also provide that as a medical device to the clinical sites we're to be able to then do it themselves if they want to and that way it's something that they can do it's like a quality improvement things it's all run through the cloud it's all anonymized pseudonymized so that you know it can be done from anywhere you don't have to be in the hospitals because again that's something the way that people work now perfect um but you know secure aws we use but the thing is it allows people the choice and they can do part of their service um and and for us it's much simpler being a software company it's much it's it's much cleaner, isn't it?

56:39And I think it allows you to focus on what you're good at and, you know, any, any, you know, your time and your effort and your energy and your expertise going into, you know, the next thing and the next thing rather than trying to build out, you know, hiring a load of radiologists or whatever it is. And yeah. And taking on that clinical risk as well, as you say, it requires CQC and all that sorts of stuff. And then you've got all of that stuff to deal with and the litigation's ultimately distracting from what you're very good at, which is clearly this side of things um this has been fascinating reza thank you so much um no i my my question for you my final sort of question for you as we're wrapping up here is um where is this now in terms of your scale and what can we expect from this because you know I'm not going to lie to you.

57:33I think this is something that I speak to a lot of companies doing a lot of things, right? There doesn't seem to be any holes in this in the sense of, like, my comfort level of this reaching market, reaching patients, is I'm so comfortable with it because of what you've talked about, which, you know, is relatively rare. There's always some risk, I think, being carried by particularly software as a medical device and AI as a medical device. Like, there's always seemingly some level of risk somewhere. but I think everything about this makes things more robust rather than introducing any extra risk and I think that the steps that you've taken and obviously your research background being key here but the New England Journal of Medicine is the one that's you know published your statistics so like in terms of peer review you know everything is pretty robust here so I'm interested in where your scale is currently what you're looking to do what you're looking to achieve and indeed anything else that you're looking to do in this sphere or indeed any other sphere in your career because i think you're you're someone that's worth watching from here on in oh thank you well i mean you know i've said we're 18 months in as a company uh and so it's good because we're literally literally about to in the next week or two launch our product across europe at least in six major centers our country if i should say with a partner you know we haven't we're a small team for 14 15 people we haven't built a big sales and marketing workforce but we're working with the major alt-time companies because they see our technology is really good for their customers that's the obvious one because they've got distribution big sales sales workforces in each of these countries So one of them is launching our product now through all their sales workforce across these six countries.

59:24And in fact, they've been kind to us, they've already given us some of the expected sales proceeds up front because we're a small company, so that really helps us. Excellent. That's good negotiating. Perfect. Another one of the really big ones is a central contract to do something similar. And we're just exploring that and negotiating. The third one is we've just had a very good conversation with, and then we just, again, sending a contract back and forth. So there are avenues, alternate companies, the software for the reporting companies, even PAX companies and others that have workforces that we can work through to get this technology out there.

1:00:02And that's our approach. So this is in Europe. We, you know, be aware we have CE marking. and what we want to do next is move to the US and because that's a really important market. So we just submitted for the FDA and we're now hoping to raise some funds to invest in launching in the US as well because that, of course, you know, it's a very attractive market to us, but it's also a big place that requires some investment to set up. You know, you can't just be European, UK company and say, well, I'll just sell in the US. you literally have to become a bit of a us company as well yeah to be successful in that market yeah oh god and you should think the amount of impact you can have over there as well and the size of it you know in the us they're even worse than europe really see there we go yeah there we go despite so much more money they're spent yes reza thank you um i've really enjoyed this genuinely i i've thought it's absolutely fascinating to hear what you're up to the way that you have conducted yourself and your company and in the way that you've grown it the robustness around evidence um and frankly how that's going to pay back to you i think is worth looking out for any health tech entrepreneur so i really appreciate you coming on to tell us to tell this story and for being so open about your personal background as well in your motivation for starting this for people that want to learn more about you or they want to get in touch with you or indeed they want to learn more about freya what's the best way for them to do so well uh we have a website which is very easy freya is spelled slightly differently to the norse scholars that is named after is basically f-r-a-i-y-a with the ai in the middle of course a freya uh just freya.com uh you know come to our website and there's a way for you to get in touch with us amazing the website has more information videos and things and you know the team etc it's all perfect i appreciate you thank you very much all right lovely to chat bye-bye