Today our theme is to discuss digital health technologies in cardiovascular medicine. We're going to discuss the success stories, and the opportunities.

I'd also like to introduce everyone to our illustrator, Alice, who will be scribing today's discussion. 

So today's topic is about cardiovascular disease and digital health technologies.

Cardiovascular disease is the leading cause of death in Australia and worldwide, and its incidences continues to rise with an aging population.

Effective treatments are urgently needed, yet recovery for cardiovascular events is often complicated by comorbidities like kidney disease or diabetes, and hospital acquired infections. 

These challenges are placing pressures on our healthcare system and highlight the need for solutions that are both cost effective and time efficient.

The panel today will explore the advances in a particular area of cardiovascular technologies, which is digital health technologies, and how they're helping with cardiovascular care.

Digital health technologies is a broad term that incorporates things like wearables, digital twin simulations of cardiac physiology or vascular physiology and also digital infrastructure for data, patient records and things like that.

So it's a pretty broad topic, but we will discuss a few of the exciting opportunities. 

I'd like to introduce our first panel member, Vicki Edwards. Thanks for joining us. 

Vicki has more than 25 years of health technology industry expertise. She brings cardiothoracic clinical experience and a deep commercial capability and understanding of how to successfully bring new interventional technologies for cardiac care to the market. 

Our next panellist, is professor Peter Ballas. Thank you for so much for joining us. After graduating from the University of Melbourne, professor Peter Ballers, trained at Austin Health, and completed a master of public health at Monash University and undertook a fellowship at the Royal Brompton Hospital London followed by a PhD at Rasmus University Netherlands.

His PhD was on optical coherence tomography in cardiology, and he led optical coherence tomography adoption across over 60 centres in Australia and New Zealand, and is internationally known for his work in interventional cardiology, advanced coronary imaging and novel cardiovascular therapies.

A fellow of multiple international cardiology societies, professor Peter Ballas has authored three textbooks, and over 130 peer reviewed papers in his areas of expertise. He leads a biomedical engineering research group developing new imaging and cardiac devices, including 3D printing and serves on several editorial boards.

Thanks Peter, for joining us. 

And finally, I'd like introduce everyone to associate Professor Lachlan Miles. 

Lachlan is a staff specialist and head of research in the Department of Anaesthesia at Austin Health, an honorary principal, fellow of the Department of Critical Care at the University of Melbourne, and an honorary principal, research fellow and perioperative clinical lead in the preclinical critical care unit of the Flory Institute of Neuroscience and Mental Health. 

For the benefit of the audience who have joined us, I think it'd be really good to set the scene by first discussing the burden of cardiovascular disease. 

Peter, maybe you could start us off on that.

Heart disease has probably touched many of us. As a practicing cardiologist. I work as an interventional cardiologist, and that is a cardiologist that manages patients with these little stent devices. And that's essentially 24x7 being available to treat patients with acute heart attacks.

And that's perhaps one of the more feared cardiovascular diseases, an acute heart attack, but it's really an umbrella term that refers to any disorder or condition that affects the heart. 

The blood vessels affect the heart itself. The leading cause of cardiovascular disease is stroke and heart disease.

The statistics are sobering, but nonetheless they are pretty dismal. And I guess, 120 people die in Australia each day, and that's just an enormous number. One person every 12 minutes dies. And, I see the deaths, and I see the disability that this condition brings forward.

And not only does it affect patients, but it affects their loved ones as well. And then they are unable to work. Just last week I managed a patient in their early fifties who had a significant heart attack. This person was an executive and now obviously is having major concentration issues.

Medications are becoming a problem, unable to work, and it has a huge impact. 

Lachlan, I might ask you a follow up question that I think, touches upon what Peter was mentioning.

Could you tell us a bit about what early signs, there's that feeling that heart attack is a serious one, but there are things that happen beforehand that have led up to that.

You raise a very important point, and in my area of practice, which is anaesthesia, and in the perioperative space, we can broadly divide these into two categories. 

The first category is overt cardiovascular disease, but there is actually a far greater problem, which is this concept of silent or hidden ischemia. And these are events that patients may not even know that are happening.

And we tend to see them in patients who are elderly, in patients who have other risk factors such as diabetes or high blood pressure. 

And in the perioperative space these are patients who are walking around, living their day-to-day lives. There are no problems whatsoever, but if they have an operation or a great stressor in their life, these then can become apparent.

And, and there's a term for this in the perioperative space, which is myocardial injury after non-cardiac surgery. 

Now these events are problematic for two reasons. The first is that they're common. They affect about five to 10% of all patients after major surgery. And the second problem is 93% of patients who have one of these episodes have no idea that they're having them.

So you might be at home, you get chest pain, shortness of breath, pain radiating to the jaw down the left arm. You say this is a problem, you call an ambulance, you go to hospital, you meet Peter. On the other hand, the events that I'm talking about are completely undetected unless you actually go looking for them with blood-based biomarkers.

And if you have one of these events after major non-cardiac surgery, they increase your risk of mortality roughly 23 fold. 

So these are major unrecognised clinical problems that nevertheless exist along the spectrum of cardiovascular disease that Peter was talking about. But what Peter deals with is largely at one end, and what I deal with is largely at the other end.

You both alluded to age as a factor, and Peter you were mentioning that you do see younger people coming to the clinic as well. 

Maybe you can talk about aging and how it's not even about aging, it's maybe even lifestyle and aging is changing who you see in the clinic.

Correct. Outcomes after acute heart attack have improved. Our stent devices are getting safer. Our pharmacotherapy has been instrumental at improving outcomes and thinning blood with certain drugs that we use and lowering cholesterol with statins and managing blood pressure. 

But we are seeing, day to day, younger patients presenting with cardiovascular disease and the challenges there are particularly genetic factors. 

So there are genetic factors that predispose us to building up plaque or what we call atherosclerosis. And that is these fatty deposits in the arteries that affect the heart, but also the major risk factors that we're seeing of high blood pressure, obesity, and diabetes.

We've all probably seen in the lay press about all these new drugs and Ozempic and the impact they're having. 

But the challenges of immense obesity leads to significant issues with high blood pressure, higher risk of diabetes, higher risk of obstructive sleep apnea. Another very important but under-recognised condition that contributes to cardiovascular events.

And as a result, we are seeing the younger patients and that means that these younger patients are living longer with the consequences of heart disease and the chronic nature of heart disease. 

I tend to manage the acute problem when somebody comes in hospital with a heart attack, but I look after them for many years later.

And managing those issues longer term are major challenges. 

We might have heard of a condition whereby the heart muscle becomes quite weak. While that also leads to not only disability, or inability to work, but also recurrent hospitalisations. And that's a major, major challenge.

Vicki, you work at Hydrix Medical and it would be good to introduce what Hydrix Medical does and also then talk about digital health solutions. 

Hydrix Medical is based in Melbourne, primarily a design and product development company. 

So we have a large group of engineers of various disciplines that work with a range of companies and personal cardiologists, vascular surgeons to develop solutions, not only digital solutions, but also active implantable devices et cetera, to help solve some of the problems that we see in healthcare.

And predominantly we do a lot of work in the cardiac space. But we do product development and work with a range of companies from the digital health perspective. 

But I think digital health really encompasses a lot of areas and we're talking about simple wearables like Apple device, apple watches, all of those Samsung watches, to medical type devices.

But then there's other areas of digital health that are less known. For example, electronic medication management, the electronic health record, but then there's also the electronic medical record, which sits inside the hospital system, and the healthcare system that provides details and continuity of information for clinicians.

But even wider than that, the plethora of data that's available, from digital platforms now is just phenomenal. And it allows better care, it allows a lot of data for research. 

And now with AI, or machine learning to actually be able to run over that data and look to solutions or find patterns in disease that allow the clinicians to look further into treatments.

Just simply at a patient level, the number of apps that are being developed both by hospitals and commercial entities tp allow them to manage their diseases better, manage their medications better, their symptoms better, or just provide health advice, the spectrum is massive across the range.

The nice thing about digital health technologies is the more data you have, the better you can understand it. 

It allows our cardiologists and our researchers to investigate better care and better follow up, and just provides us with such a huge opportunity to improve healthcare and clinical care across not only Australia, but globally as well, across populations.

So, digital health is just a huge opportunity across the spectrum for everyone. 

What's the most common pain points that the company's identified? 

I think there's a range depending on the technology that's available. It's great to have data, but too much data also then gets quite difficult to handle and understand. 

A lot of people in the health area talk about interoperability, so the ability to actually access the data and be able to combine data across platforms, whether that be the electronic medical record or into a patient app that makes sense. Cost is another aspect and most of these data sets are now cloud-based.

It comes with a certain cost to the healthcare system. And we all know the healthcare system is lacking in resources from a funding perspective, so that makes it difficult to be able to provide the solutions in a cost-effective manner. 

And change management, which is a bigger area, trying to implement platforms and changing the way people actually interface with information, how they use it in clinical care, et cetera.

And from a community perspective, I think the biggest thing is health equity, access to data, access to platforms. Being able to introduce technologies into rural and remote areas where internet access is not as great or whether health literacy or access to digital platforms just isn't available as yet.

I think just in the remote and rural area, that's probably where the biggest bang for buck is with digital technology and access to patients and being able to reach patients from a health perspective.

I want to kind of touch upon something that Vicki just raised, which is about access to data, the sharing of data, uh, and the potential to make rural healthcare more cost effective and just effective for the patient.

What do you think the challenges are in getting this ability to share data across platforms?

Within the critical care space, this can broadly be discussed as a series of three orphan subspecialties, which don't really fit very well into medicine or surgery or gynaecology or paediatrics. And broadly, we talk about anaesthesia and perioperative medicine. We talk about intensive care and we talk about emergency medicine.

And much of the work that's been done in this space in terms of digital health solutions tends to be in patient monitoring.

Now, the objective of any form of patient monitoring is twofold. It's prediction and prevention. 

So can we predict on the basis of a patient's vital signs, for example, whether or not they're likely to develop a cardiovascular event, and then is there a way having predicted that event that we could intervene and try and prevent that that event from occurring?

And this is a problem which is consistent within rural and within renewable health environments with primary healthcare, but also within advanced secondary or tertiary level facilities where we look after patients who are very, very unwell or whether or not they're having major surgery.

Now one of the major challenges that we've got is the sharing of data 

And the issue when we try and devise these digital health solutions, particularly those that use algorithms, machine learning, or computational modelling, those that use artificial intelligence, is that it requires a vast amount of data in order to be able to train these models to yield an outcome that we're interested in.

And what do I mean by an outcome that we're interested in? It's an outcome that ideally improves patient health outcomes. And by a patient health outcome, I mean something that means something to the patient. So I can say to you, I'm going to fix your blood pressure. Or alternatively, I'm going to prevent you from having a heart attack.

Now, out of those two, you don't probably care that much about what your blood pressure is on a day-to-day basis, but you care very much about having a heart attack. 

So just because I fixed your blood pressure in the operating theatre, I need to then show that that is linked to an outcome which matters to you, but also matters to my bosses, which is cost effective.

That does not result in an enormous amount of expenditure on a very small difference in terms of overall outcomes. Because at the end of the day, we are utilitarians. We need to have a view on what's happening on a health system level as well as the patient that's individually in front of us. Now, I alluded a second ago to the idea of big data and there is a creative tension here.

We need an enormous amount of anonymised de-identified data to train these algorithms and also link them to patient outcomes. 

But at the same time, each individual patient has a right to say what happens to their individual healthcare data. 

Now, it may be well and good for me to approach 20 or 30 patients that ask them for their permission to use their data, but we are talking about tens, hundreds of thousands of patients that we need to feed data into, and it's simply not feasible for us to approach every single one of those patients and seek prospective consent. 

Now, I don't know what the answer to this question is. 

On the one hand, I care very much about what someone does to my data, but on the other hand, I would very much like to use this anonymised de-identified data to improve overall patient health outcomes. 

It kind of comes down to a philosophical balance, whether or not you're a utilitarian, the greatest good for the greatest number, or where you are very invested in individual autonomy.

Either way, this is a problem that our privacy legislation is not really equipped to deal with and will need to be something that's solved on the policy level, before we can start using it, particularly when it comes to sending data overseas, 

I might just throw a bit of a spa spanner in the works with an analogy here, you know, like how we are a little uncomfortable about sharing our personal data on Google and all the shopping websites. 

We also don't like it when it's figured out that this is the typical thing that you look at. 

Even with a TikTok, YouTube video, you just keep getting these predicted outputs of things you need. That could be quite nerve wracking as a patient who feels is this really individualised right now?

Because you're just collecting a lot of data and you're making a prediction without really thinking about me as the individual. 

And, and that's a really great point. And this is something that Peter and I face every day in our clinical practice. Peter and I are both heavily evidence driven clinicians.

We believe in large randomised trials. But of course, you then take this large randomised trial and you need to take the lessons from that trial and apply it to the individual in front of you. You might have a 300,000 patient randomised trials, which is very common. The cardiologists are spectacular at running these very, very large trials.

We're not so good at it in critical care because perhaps we have a smaller recruitment base.

But we homogenise these patients within the database, don't we? And we assume that everybody within the database had the exact same outcome in the trial. And we try and then homogenise the outcome of the trial.

The only answer to this problem is more individualised data points from more individuals being inserted into the models. And for that we need big data and we need to train these models. An analogy is looking at it like a self-driving car. So do you think that Waymo just took a car and said, alright, off you go, have fun.

They don't. They train these algorithms on hours and hours and hours of individual driving data so that they know how to react in individual different scenarios. 

Now we need to do the same thing in our large database models. When we are using computational modelling or AI, we need to have that quantum of knowledge so that when an algorithm sees something, it says, I've seen that before, and I can make a reasonable prediction on how this patient is likely to react. 

Now it's a well stated maxim that all models are wrong, but some are useful. And in terms of utility, we need to ensure accuracy, but there always must be a human involved in the system.

This is not a self-driving car. This is not a closed loop. So there always needs to be somebody involved in interpreting these data, determining whether or not they think it's relevant to the clinical situation in front of them, and then instituting the suggested management if your algorithm allows for suggested management.

Peter, I'd like to learn a little bit about your experience with digital twin technology. 

You raised a point there following what Lachlan said about the data is great and digital technologies are wonderful, but we do get nervous about what data we are sharing and security breaches of our data become even more problematic. 

One example I'll give you about how patients have changed in the way that they feel towards digital health technologies.

Patients who have had a pacemaker, for example, traditionally they would go to the clinic every six months and have a pacemaker check with a device that is placed over their pacemaker. And then the, the technician and the doctor will make an assessment. Nowadays pacemaker checks are done all remotely from home.

You have a pacemaker. You are sent home, a little wireless box that sits at the end of your bed and the team tell you this is wonderful. Every night, your doctor and the company that your pacemaker belongs to is receiving your data. And we can adjust anything that might be necessary.

Now, that's exciting, but it's very scary and many patients are still very hesitant about what are you doing with all this digital technology and how safe is it? 

So that's really a major challenge. Vicki also raised the issue of rural health and how these technologies might impact rural health.

I work in metropolitan Melbourne, but I have clinics in Swan Hill, four and a half hours away. 

I'm also in Foster and Bendigo. And that is a major challenge. Now we are looking at these wearable technologies. There are 30-day heart monitors that are called heart bugs that can connect with a smartphone through Bluetooth and that instantaneously transmits heart rate and rhythm information to us.

So that's been a major game changer. We have implantable recorders that are little chips the size of AAA battery that get injected under the skin. These can last for about three to four years in battery and are suitable for patients who have had a possible faint blackout that potentially could be cardiac or might be at high risk of a condition known as atrial fibrillation, an irregularity of the heart rate.

While that, again, has also been a game changer in the way we manage our patients, because we can instantaneously detect whether a patient is going into atrial fibrillation and of course potentially prevent a significant stroke. So without these technologies, we will not be able to identify these particular patients.

So what have been some practical privacy solutions that have been developed?

Look, I think all of the companies are looking at significant levels of security across the globe.

Healthcare is driven by things such as HIPA in the US, the GDPR standard rules out of Europe, which are pretty stringent as well. And IT groups in hospitals are also very diligent with the security around the solutions that are being presented to them.

Recently I was doing some work in Singapore and we went through close to 12 months of question and answer on the platform that we were talking about and looking at all of the security aspects and the what ifs and what happens to the data and who gives permission and how, and, and all that sort of thing.

So, you know, everyone's looking at the privacy and the security side especially, you know, especially when you know, in Australia at the moment, we've had a number of data breaches, with different commercial entities outside of healthcare. 

So everyone is really nervous and the companies really work hard to ensure that the security and privacy are in place.

And then working with cloud such as AWS and Azure, which are the two main ones that we use here in Australia, their security levels and their fundamental work in security is all aimed at maintaining that security and privacy and ensuring that data is secure.

I think the other thing is de-identification of data. 

So if you are looking at remote monitoring, you need to have a degree of patient ID to be able to report on that information or to be able to access the information and talk to the patient about something that you've seen.

But when you're looking at the large data sets that we are using in the echocardiogram space as a massive Australian data set and that's de-identified data of echocardiograms that can be used for algorithm development and research. 

So I think it depends on the use case. In some instances, from a personalised perspective, you do need to be able to identify the patient and the data associated. But when you're looking at big data and you're looking at these massive data sets, the data can be de-identified. It doesn't need to be down to individual.

For a data set that belongs to a hospital and a hospital group who want to look at post-op complications or return to hospital or whatever the case may be, or whether it's larger data sets where we're looking at echocardiograms for AI or we're looking at medication management, or whatever the case.

It really comes down to the use case. But these data sets that you're accessing, are they set up by the hospital in agreement with you or is this something that they're collected? 

Is there one-to-one agreement before such a data set is collected, or are we looking at some government initiative that's collected that data?

What's happening? 

We work in the remote monitoring space for pacemakers and, and implants that Peter was alluding to. That permission is given at patient level to the hospital. 

And the cardiologist and then the hospital, and then that data is shared to the manufacturer's site, which then under license, can be used for monitoring.

That's part of the agreement with the patient. Other data sets, the echocardiogram data set that I talk about, that data set is a large national data set, and that has been licensed to companies or one company that I know of that, that can use the data set for de-identified data.

I understand that in the US some of the bigger organisations, part of being admitted to hospital, is a consent that your de-identified data can be added to the hospital data pool or data lake to be used for research purposes.

OK. Questions.

In the coming times, would AI be able to predict a disease or health issue before it sets in?

We're already doing it. There is a commercial product available called Hypotension Predictive Index, or Acumen.

It was originally built about 24 months ago by Edwards Life Sciences, And this is a device that was built in acknowledgement that low blood pressure during surgery was associated with worse outcomes.

So this was a machine learning algorithm that took into account literally hundreds of data points from advanced haemodynamic monitoring waveform, so the arterial waveform, for example, and gave a warning to the anaesthetist or the critical care physician. It told them what the likely cause of that hypotension was going to be, and then invited some strategies by which they could correct it. 

Now, there was a major problem with it. The device works very well. The issue is what I alluded to before, just because you fixed the blood pressure didn't mean that you changed the outcome at the end.

And so, whilst HPI is probably the best example of early iterations of AI technology in the perioperative and critical care space, association is not causation. 

And so it's important that we keep trying to find intraoperative warning signs that are associated with real patient outcomes. So the work continues in this space.

On a broader and perhaps less acute monitoring level, however, we are starting to see wearables enter into this space, particularly in terms of getting patients out of hospital earlier and back to their home environment where they're more comfortable. 

So you've had major surgery. Normally you would expect a four to five day stay in hospital just to make sure you're okay.

But if you're feeling well enough, you go home with a wearable now that continuously monitors you as you convalesce and recover at home. And if there are any problems within the algorithm that indicates that you might be going to get a complication or if there's an issue. It alerts the treating clinician back at the hospital, there's a problem and invites you to connect with them.

Now. It's a very, very early stage in terms of development, and it's still about those two ideas about prediction and prevention. We're still working on what the outcome is that we are looking at, and we are looking at the intensity of monitoring as well. So it's already entered the critical care and perioperative space, but it does require some refinement with subsequent iterations.

I might even add there is a field, the field that I work in, which is computational physiology, where it is about creating physics-based simulations of biological systems that was started in the 1970s and it's been growing since then.

We got to the point where we could create digital twins of an individual's heart, but it's still difficult to characterise that heart and then predict what it's going to do next. 

AI has come from another angle. It's much more pure computer science angle where it is a fantastic pattern recogniser, and it can handle lots of data, but it is still a pattern recogniser.

I think there's an opportunity where these two fields should meld and get to that next stage of predictive algorithm development. 

Peter, another register question I thought you might be great to address. 

My dad had an implant in one of his veins in his heart because there is a fat build up. How with current advancements in science and technology is this preventable? 

So that sounds like somebody's had what we call a stent device, and it's normally implanted into the arteries.

Or if they've had previous bypass surgery, they can be implanted into the vein grafts. 

And these devices are used to open up the artery to restore blood flow when there has been a reduction. Because of this condition known as plaque or atherosclerosis, these fatty deposits and these stent devices go in and act like a scaffold, like a spring to keep the artery open.

And improve outcomes. They are lifesaving if somebody's having an acute heart attack and they can improve symptoms in those patients who have chest pain or tightness. 

Now, what we are doing, and that's particularly my area of interest as an interventional cardiologist, is that we are using AI, not so much because we feel that AI is better than humans, but because AI is actually much faster.

So when we've put a stent in, we can analyse the stent by taking very high resolution images of this stent inside the artery using this technology that we talked about earlier called OCT or Optical Coherence Tomography. 

It's a laser scan and it gives us unprecedented information. Now, for me to analyse, say a 18 millimetre stent in an artery manually analyse and how it's functioning, takes me probably about 90 minutes.

Using AI and with machine learning, we have been able to get that down to 19 seconds. 

So the speed that AI is giving us, I really see as a huge opportunity to be able to personalise treatments in the procedure lab whilst we're implanting these devices and potentially learning how do we optimise them rather than doing an off label.

And after we've put the stent in and analysing it, with our postdocs and our PhD students over the course of several hours or days, that's really where I think it's going to be a game changer. 

So AI improves our speed and that will allow us to put this technology at the coalface, right in the clinical laboratory, in the ICU, in the lab where it's needed most.

Hmm. Lachlan, you wanted to raise? 

Yeah, I'd just like to add just a little grace note onto Peter's excellent summary and he's described is the ideal use of AI. 

I saw something from a creative on social media recently, which is that I want AI to do my washing and my drying so that I can write my poetry.

I don't want AI to write my poetry so that I can do my washing and my drying. 

So what AI needs to be is an interface by which the boring things that make up our lives are eliminated or made much faster so we can focus as humans on the broader strategic implications of the technologies that we are bringing to bear.

Put it another way. AI should do one of two things. It should do something that humans are just not able to do. Or it should be used as a labour-saving device for humans. 

And why is that? It's about clinician acceptability. You see online all the time, people talking about how AI is going to replace Peter and myself.

That is not an appropriate use of AI technology. It should free Peter and myself up to focus on you rather than eliminating us from the workforce. 

And if you can do that, it will massively increase the uptake and the acceptability by clinicians on the cold front. Peter and I don't want to be told that we're going to be replaced.

We want to be told that we are going to be helped. 

Yeah. That's fantastic, Vicki.

I wonder if you want to add to that at all? 

It does give us the ability to get those answers much quicker then that's the first thing. 

The other thing that we've seen ith the ability of digital health technologies and using AI and big data sets, is being able to identify things that may be visually difficult.

If you look at x-rays and the rise of the use of AI reporting in x-rays to identify things that aren't necessarily as easily picked up by written reporting on it. 

And that's not to do with the expertise of the reporters, it's just that it can do it better. 

We've seen that with Harrison AI.

I think there are specific areas and it's definitely not designed to replace clinicians, and clinicians need to be the decider of that end care, but it needs to help. 

As Lachlan alluded and do those tasks that free them up to actually spend more time with patients and more time doing that personalised medicine.

And something that is hidden away in making these training data sets is the manual labour of curating the training data sets. 

So there's a ton of boring work to be done at the stage to get that data set that you think is going to help make good predictions.

That is hidden away under the carpet. If you look at some of the technology outside health, it's the same thing. 

Question. Could we use the equivalent of organ donation consent for de-identified patient data for research preferable opt-out system like my health record? 

So with organ donation, you put down on your driver's license that you would like to be an organ donor, but that for you can be overrode by your family because if you do end up becoming an organ donor, you're not usually in it. You should never be in a position to be able to consent for yourself unless it's an altruistic donation.

We are exploring, in my institution, adding this to routine surgical consent. 

So in the perioperative space, it's actually quite easy to do. You present to hospital, you are having elective or emergency surgery, you are required to sign a consent form and there will be a tick box on that indicating that you are happy to use your data for research purposes.

Now that leads to a broader question around what are research purposes and do we include sharing data with industry as part of those research purposes? 

So back in the 1950s, if you went to the Alfred Hospital and you presented, there was a little plaque on the wall next to the entranceway that said, if you enter this hospital, you consent for an autopsy in the event that you died. Now, obviously that went the way of the Dodo people weren't happy with that in terms of the way that they were approaching this very thorny issue of personal consent.

So I don't think current privacy legislation will allow for an opt-out system like my health record when it comes to sharing your data with people who are not your individual treating doctor and who are not yourself. 

But there are ways around this, particularly in the perioperative space, and Peter might be able to speak as to whether or not it works in the cardiology space, that as part of routine consent, you would give your permission to use that data for research.

Peter, would you like a great question. 

I guess it all depends on the type of research, doesn't it? And I guess having that sort of overall umbrella statement of yes, I consent.

Research can just be purely data collection, which I'd be happy to participate in. But there is also research that we are doing, putting new devices in that have not been clinically approved and I would like a bit more time to spend with my patient and discuss this and with their family prior to enrolling them in a study like that.

I think depending on the research, yes, it might be an option to participate in some data collection and to identify data research. But in terms of interventional research and invasive research, I still think that requires a discussion. And I've been humbled by the support that patients give us.

I find them very eager to participate. 

I think we have time for one more question, and it's kind of heading us back to the AI topic, which is encompassing all of our minds every day at the moment. 

The FDA is evaluating AI to determine if cardiac deaths in clinical trials were due to test drugs or not. What is your perspective in AI making these decisions? 

It makes me a bit uncomfortable to be honest with you. And the reason I say that is that you would, you want to know how well the model performs. So let me give you an example.

In a recent trial that was run by our clinical trials network, we were looking at whether or not ketamine intraoperatively could prevent chronic pain, and in order to adjudicate that outcome, we had three chronic pain specialists review each outcome and the patient reported data and make a determination about whether or not it met with the predetermined endpoint in the trial.

Now, I don't see an a problem with AI doing that as long as there is precise alignment with the clinician opinion and the AI model. 

Put it another way, it's going to be right a hundred percent of the time. Why do I say that? Because trials that look at these new drugs in clinical trials, they tend to be very small and they tend to be very fragile.

One misjudged endpoint could be the difference between the drug being inappropriately approved or inappropriately rejected. It does not take many positive results, particularly for rare endpoints like death. It does not take many false positives or false negatives for you to reach the wrong conclusion.

So without seeing the data as to how good the model is at replicating the opinion of the adjudicating clinician, I would be pretty keen to keep a human involved in that process. And we've seen this, haven't we? With the Doge cuts in the United States where attempts were made to use artificial intelligence to make firings.

Subsequent to that, they've had to hire back almost the entire workforce because they're using a tool for a purpose, which it wasn't developed. So that's a cautionary tale, and I'd be keen to see the performance of this algorithm before I said it was appropriate or not. The stakes are very high.

Vicki, with the products that Hydrix Medical delivers to its clients, what's the process of going from, we want to see if this device works in our clinic to we're going to deploy it into your clinic. Is that a conversation that you have to go through?

In general, most sites when looking at digital health do want to trial it and have a look at the outcomes and the outputs that are coming through. 

They also, don't want to see a black box kind of effect. They want to understand how the decision, how the data is used and how the decisions are made inside the algorithms.

So we often have those discussions. In general trialling of the software is available. 

Alternatively, if other sites are using it, then peer conversations around using their solutions. 

So it really depends on at what stage in the market you are and whether you're first or 10th or hundred.

Clinicians like to talk to colleagues about using these systems as well. It really depends on what the system is, what the output is. But the idea of understanding what's in the black box, is incredibly important.

Do you think that the industry would value a third party doing that sort of work?

I think by the time you're taking it to the clinical team, the vetting has already been done. 

I think where research and where universities can come in is coming up with these data sets and algorithms and looking at the use cases.

If we look at heart failure for example, as a major issue in our healthcare at the moment because people are living longer and heart disease continues to increase, areas of research around how we can better approach or better predict or use early symptoms to better predict outcomes or help provide more information for those algorithms. That's where research in university is really good. 

And then if there's something that comes out of that, then taking that and looking at how that might be commercialised or used rather than looking to test existing solutions, 

Thank you so much. I really appreciate it. Thanks, Peter. 

I also want to thank our illustrator, Alice. I've been following briefly all the cartoon illustrations she's been doing. It's fantastic.

Thank you everyone. Stay safe and have a nice day.