Max Little, Associate Professor Of Mathematics At Aston University, UK And
Visiting Associate Professor At MIT, USA - Smartphones And Wearables In Clinical Research
I’m delighted to be able to welcome Max Little. He’s going to give us a presentation on technology as an enabler or a hindrance to participation in self-management. Max Little is a professional applied mathematician and statistician based at Aston University in the UK. He’s best known for his multi-disciplinary research, which includes the use of consumer technologies, such as telephones and smartphones, to detect the symptoms of Parkinson’s remotely. Delighted to have Max here to wrap up these two days. Thanks Max.
Thanks very much. Really appreciate it. So Paul really set the scene brilliantly for what I’m going to be talking about. But I want to take a step back basically and to say that I think all of this consumer tech is fantastic. I mean I’m a geek at heart and I love this kind of stuff, it’s amazing. But on its own it doesn’t really mean anything for clinical research. So to set the scene I want to describe a particular disease area that I’m heavily involved with, and it will help to bring out, I think, the many issues that I want to discuss. Well I spend a lot of time working with Parkinson’s disease. There are estimated to be around 7 million people living with Parkinson’s worldwide. One of them includes my friend here, whose name is Jan Strickling—here he is with my mom. And as Paul says, I’m not a clinical researcher, but I’d like to try and help. And maybe I could just ask of you to put your hands up if you know somebody with Parkinson’s disease. And you obviously will want to help and you know what a devastating disease it is. And being able to have this opportunity to work in an area which you know is making a difference is what motivates me as well. And then obviously you’ll know that it’s an incurable disease, and so in a way it all seems rather hopeless, you know, to be able to try to do something here. But let’s not give up of course.
The point is, if you want to do something in Parkinson’s disease, you’ve got to know something about it. you’ve got to know what’s the science, what do we know about the science, what’s the science of Parkinson’s disease. Very little, unfortunately. The proximal cause is that dopamine-producing neurons in a particular area of the brain called the substantia nigra gradually die over time. Perhaps though, over a very long period of time, and so that what we actually see when somebody gets diagnosed is kind of like the end stages of the disease. That’s of course incredibly frustrating. And it’s not at all obvious because it's such a heterogeneous disease and different people have different kinds of symptoms and progress at different rates. There may indeed not be one type of Parkinson’s diseases, there may be many. And it’s not obvious why one person gets it and another person doesn’t. So given that kind of variability, and the first thing that everyone points to, they say ah, this is genetics, it’s genetics at work. And there's been intensive research in this area. So we’re getting to the scale now where we can look at whole-exome data for a large number of participants, the kind of trials that I’m working with are looking at that. But unfortunately, again the evidence is rather inconclusive on this, because although it is the case that there are some genes, some genetic effects, that lead people to have an increased rate of getting Parkinson's in the population, there are only a few genes known to do this, and it may actually be not one of those genes on its own, a single expression of a gene, will actually lead to the disease, it maybe that it’s multiple genes acting in concert, because it's a hugely complex area. So the evidence for genetics is somewhat inconclusive. So then it's just random, yes. And when I say random, what I mean is biology is random. And you’re going to say well we’re all the same, you know, pretty much. Yes, this is true. On average, cellular effects and molecular effects produce people that look pretty much the same. But if you ask a molecular biologist they’ll tell you that of course all this jiggling around of molecules is actually just random, it’s just quantum behavior. And so it means that actually what happens over time is that some people will accumulate deleterious phenotypic accumulated effects that lead them to get the disease and someone else doesn’t.
So just looking at the molecular level, what do you know about this disease. Well so far, for example, we know about these alpha-synuclein and like prion-like proteins that accumulate in the cells. Now, these alpha-synuclein proteins, they’re essential to function, so it’s not that these proteins, if you have this protein in your brain, that means you’re abnormal. You’re normal, it’s just that probably you have more of this protein than you should accumulating perhaps in the wrong place. And there are effects of this. For example, we know that there’s a symptom called hyposmia, which just means loss of the ability to smell, and it seems to be a precursor, it seems people who have this particular symptom are more likely to go on to get the disease. And there’s yet another clue, which is called REM sleep behaviour disorder. And understanding sleep clearly very very important there. So if you could potentially pick up these episodic behaviors before the so-called motor symptoms, the tremors start appearing, chances are that you might be able to pick it up early. But otherwise, because of the fact that we don’t know who is susceptible actually, we’ve never really been able to study people to find out whether for example which specifically are symptoms that we should be looking for, these phenotypic effects. So the evidence there is somewhat inconclusive. So then you’re going to say okay it’s just down to behavior. It’s your lifestyle is different from someone else’s lifestyle and that's what leads you to get Parkinson’s, you know, the effects of your behavior end up by accumulating over your lifetime and that’s what caused it.
Unfortunately we don't know anything about that at all. What we do, we know a little bit, what we’ve been able to collect in surveys. But as this session, I think, has brought home quite clearly, surveys can be somewhat misleading in fact and of course they only give you a tiny little snapshot of behaviour over a period of time that’s probably way too short to actually pick up these effects. So we don’t really know anything because it hasn’t really been possible to objectively measure what is going on in someone’s life over this period of time.
So maybe it’s just environmental effects, so it means that actually what it is, you need to know about someone’s environmental exposures and you need to know about that in order to be able to understand the difference between whether someone gets to have Parkinson’s or not. So does where you live actually explaining the risk of you getting Parkinson’s? Well there actually is some evidence to this effect because farmers who are exposed to a particular kind of toxin are more likely to go on to get the disease than others. But again, unfortunately, with environmental data, we just don’t really know enough.
I’ll give you an example. If you say, you ask to survey somebody and you say, so have you been exposed to this particular agricultural toxin. And they say no of course not, I’m not a farmer. But they live in an agricultural region, and every day they drive backwards and forwards past a field which is exposed to this particular, say, pesticide. We would never know that from the survey. So this is the point, right, which is that a lot of this information could be hugely missing and we don’t know about it. So environment, again, is confusing.
So obviously, and this is the point I want to make, we’re going to have to get all of this data in order to try to understand this disease. We’re going to have to look at genes, we’re going to have to look at phenotypic effects, random expression profiles, we’re going to have to look at proteomics, we’re going to have to look at behaviour and environment, and all this information is going to have to be combined in order to try to make progress in understanding this extremely frustrating and complex disease.
So my research, and what I’m going to be talking about next, is about focusing on addressing the many technical problems that arise in trying to record behavioural and environmental factors, one of the biggest problems being the sheer expense of just contemplating collecting the kind of data on the scale that we need in order to try to understand this disease. Because actually, with the disease, we need to be able to measure behavior and environmental effects as often as possible. And we also need to do that over very long durations, and I mean upwards of 20 or 30 years. So you start thinking about doing that right now. To make that possible, every part of the measurement process is going to have to be ludicrously cheap in order for that to happen. There’s just no other way it’s going to happen.
But I actually believe that we’re living in a rather miraculous time in some ways, because I think we have these miraculous technologies that are just sitting there, just in our pockets, we use them all the time. And as Paul pointed out, these technologies are advancing at an incredible rate, and they’re starting to be able to collect things that were just unimaginable even just a few years ago. And our approach is to adapt whatever measurement technology is actually available to participants right now. What sort of technologies are currently available that are in participants’ life that they know already how to use and that have the capability to collect the kind of information that we’d be interested in. And what I mean by that are things like commodity smartphones, the telephone, internet-connected PCs, whatever we can. So I should take the step back and say, what sort of consumer technology has been and is now available to do this kind of remote telemonitoring in Parkinson’s. And things have come an enormously long way, they’re moving very very fast, and they happen in such a way that you don’t really notice. You know, there was a period of time, perhaps I’ve got listed here between 1990 to say 2006, before the emergence of the smartphone on a large scale, so that is something like 15 years of development in technology, but it’s taken well maybe, I suppose, six to seven years for smartphones to mature. So that gives you some idea about how the technology is speeding up. And smartwatches are already maturing, they’re sort of recapitulating this same technology development cycle, but they’re going to do that in two to three years. So that gives you some idea about how these development cycles are speeding up.
So it’s no longer feasible for you to think about what is available now, you have to think about what’s available in a year’s time. If you’re thinking about what is currently available you’re missing the point. Because by the time you plan the studies, the technology that you could use to do it is almost guaranteed to be obsolete by the time you actually implement it. So you have to think about what’s coming up.
So for example, smartphones, Paul very ably described a list of sensors that are incorporated in smartphones. So I won’t go over that again. But what smartwatches are doing are starting to complement all of that same sensor suite with this rather amazing thing, that they are actually right next to the skin, so very intimate. And that means that we're getting to this era where we could be talking about full continuous vital signs monitoring. And I think that that is going to change an awful lot of things. Because for example, just one example in Parkinson's disease, people also believe that there sort of an autonomic dysfunction, that might be a precursor symptom of this disease, things like orthostatic hypertension, or even gut problems. But the issue here seems to be that we haven’t been able to actually measure this. So if we can get continuous vital signs data, this opens up an entirely new window on this disease that has been completely unavailable to us. And that is potentially perhaps less than a year away in a feasible world.
So what about the back-end service side. So Willie very ably talked about this need for standardization. And I think the key to the answer to that is that we already have the standard technology available to do it, we just have to look to the open-source community because it’s already all there, pretty much. What I should say is yes, there are lots of bespoke technology companies, but in my view, none of them are going to be around in any length of time to make this kind of project that I’m talking about feasible. They’ll come and go like that. And the reason why is because the force of standardization that exists, which is backed by gigantic corporations like Samsung and Google, etc. is pushing open source standardized hardware and software platforms to the extent that it is impossible for anyone to compete with that on any feasible scale. When you’re talking about a combined market capitalization of nearly a trillion dollars, you know, a little startup somewhere in California has got no hope whatsoever, basically. And of course the open source community is this mass of cognitive surplus that's working away to already produce these standardized technologies. So we can actually exploit it to our own advantages and use it to create platforms that are incredibly cheap and, in a way, already cover what we need to do in collecting large amounts of data.
So for example there’s Apache Tomcat, there’s a system called Twilio that is an interactive voice response system that’s incredibly cheap to use and programmable online using standardized interfaces. And of course there’s Android operating system, which is a rather miraculous thing, because it pretty much levels the playing field for all these kinds of devices. Apple maybe, in some ways, is going to be a bit of an anomaly, actually, in that respect. But we’ll see of course with the 750 trillion dollar market capitalization, let’s not be a fool to bet against that. Anyway. They use Linux. So they’re already using open source. I mean OS X platform is actually Linux. So you know, that gives you an idea about where we’re at. So we have these ubiquitous collection technologies that allow us to set up backend servers in say a couple of hours. That’s all that’s really needed in order to set up a server to collect this kind of data. So it’s incredibly democratizing in many ways.
So by combining all of this consumer tech, it has allowed us to run, for example, very very highly accessible trials. And what I mean, it’s kind of like bring your won device but on steroids, you know, like we can reach out to very large numbers of participants so that anyone with almost no training can actually participate in behavioural data collection trials, pretty much like that, you know, with very very little setup.
So for example, this is a trial that we started a few years ago called the Parkinson’s Voice Initiative. And there we recruited nearly 17,000 participants in under six months to record vocal bbehaviorsrelevant to Parkinson’s disease, and we were able to do this for under $1000. So it gives you some idea about the scale and the expense of doing this. And I’m talking about, we really should be able to get to the point where these clinical trials can basically cost zero dollars. If they can cost zero, then we’ve hit the margin, we’ve hit the effective marginal cost of the technology, which means that it’s now no longer an issue. We can just know that we can do this, we can do it for no money, so that ceases to become an issue in things like budgets. And to reach out to very large numbers of participants I’ve worked with my friend Paul Wicks, who’s done some pretty impressive work in community-led patient-reported outcomes. And together we set up a trial in which we used PatientsLikeMe, their Parkinson’s community, along with a Twilio voice recording system in order to collect patient-reported outcomes and match them up to behavioural data collected through telephone. So you can do that kind of thing, you can set that up very very cheaply and get very large numbers of participants very quickly. So we were able to get about 1000 Parkinson’s patients in four months, who participated in the study, and we were able to get pretty good quality data, just that way. And very very cheaply.
And so talking about standardization now, a lot of the world’s largest technology companies are thinking about this, putting effort into this question, because I suppose if you have a cynical point of view, they want to sell more smartphones, right. And they want to get researchers using smartphones, yeah. So for example, with the support of Apple, we were able to participate in launching a system that they call Research Kit, and that aims to be a standard platform for mass-market smartphone-based clinical research. So it’s what Paul’s basically talking about, which is to democratize the process such that anyone can do it, you know. One of the problems is that, I suppose, the closed world of clinical trials doesn’t inspire large numbers of participants to take part, because it would seem that it’s niche, you know, it’s not something you do every day. But of course as more and more people are participating in these trials, we’ll all know somebody who’s participating in them, so it becomes less alien, I suppose, in a way. But if the technology is actually available in the mass market to do this right now, with minimum of effort, then we can recruit large numbers of participants. That's the kind of thing we need to do, to get that N up to be as large as possible such that we can reach the holy grail of statistical significance. And if we don’t get N to be large enough, then we’ll fail. And of course we have to get N to be very very large because of the adherence and compliance issues. Because down the line, participants will just not participate continuously. That’s the nature of these studies, that people are interested at the start and lose interest towards the end, I mean all of us are like that. So we have to build that in, we have to understand that.
So for example the research kit was launched with the five apps on the iTunes Store on the first day it was launched and it recruited tens of thousands of trial participants in a matter of days across several disease areas, including Parkinson’s.
So I hope I’ve been able to convince you that the consumer technology that’s at our disposal right now, and that will be available just in say a few technology cycles away, perhaps maybe even a couple of technology cycles away, is approaching the point at which it is serious, high quality, clinical-grade, objective symptom data, which can be used in the mass market to get extremely large numbers of participants contributing data on an almost continuous basis. And when I talk about on an almost continuous basis, so to give you some idea about the frequency, with accelerometry data, we talk about 100 times a second. But with audio data, we’re talking about once every 22 microseconds so that’s 44,000 times a second. And that’s the kind of frequency of data we’re talking about here. You know, it’s down to this very very high precision data continuously about what is happening in your environment. This is essential information about you, about your behaviour, about your context, about your environment that says enormous amounts of information about you, your behaviour, that’s relevant, that’s directly clinically relevant and directly scientifically relevant. And what I hope is that actually with this data available in the scientific community—you know, we open it out, we make it freely available for participants to use, because of course we as researchers are very unlikely to be able to hit gold, you know, and find the answer. We have to let other people use that data, because the truth is that we as researchers have one view on it, and it’s almost certainly the case that we don’t have the right view on it. That’s just the nature of scientific discovery. So the data has to be reproducible, it has to be be out there, it has to be standardized, it has to be continuously available, and it has to be openly available. And by doing so, I hope we can actually get accelerating progress towards curing diseases like Parkinson’s.
So I’ll end there.
Thank you very much, Max, for a very interesting take on this field in a much broader take on it. So we’ll open it to questions.
AUDIENCE MEMBER: Thank you very much, Max, that was really interesting. And let me be a bit selfish for the moment, because you started mentioning prions, and before I get into health outcomes, my PhD was on the prion protein. And whenever you do research, you see it linked with Parkinson’s, Alzheimer’s, CJD. And you mentioned lots of things there, you know, prion protein, REM sleep, and gut problems are potentially one of the functions of that protein, although we don’t know for sure. You mentioned farmers with toxins and getting sick. The cows are also getting BSE in those same fields. With the research you’ve done looking at Parkinson’s, do you think, is there a way of maybe transferring that into Alzheimer’s and CJD? Because a lot of maybe Alzheimer’s cases are diagnosed as Alzheimer’s because patients are old, they have a certain set of symptoms. But they’re actually CJD or prion-related diseases, but we just don’t test properly. Do you think you’ve an application in that?
Not myself, but I’ll mention another area which I think is really interesting on that line, and that’s for example in separating out things like Parkinson’s from other Parkinsonian-like diseases, like say MSA or progressive supranuclear palsy and things like that, that are much much rarer, but they do mimic the symptoms of Parkinson’s. Now how can you really differentiate? What we’re really talking about is, how to differentiate these different diseases. And context is everything in that, clearly. You know, it’s not really enough to wait for five years to discover that, oh yeah, actually that person did have MSA, you know, because they would go in decline much more rapidly. So the key to it, I think, is it’s the same mantra. Get as much data as possible. Get all of it. And get it at high enough frequency and with enough context that you can actually start to answer those questions retrospectively. It may not be you that comes up with the right hypothesis that links that data together. But if the data isn't available, we certainly can’t test your hypothesis.
So Max, I’m wondering, is this anything you’ve—either yourself or heard of other people having said—that the big thing we always need to be conscious of within our industry field is issues around regulation, you know, the people who oversee healthcare and whatnot. Is this something you’ve managed to have conversations with regulators about or heard of discussions and their thoughts on this, which at first blush certainly seems to make sense, to capture all of this data, to try and understand these diseases.
Yeah, I mean there are all sorts of issues. Yeah, so I should directly answer your question. Yes, absolutely. I mean organizations like the FDA, it’s not like they want to make things worse. They want to make things better. They’re on the same side as us. The truth is that we really just need to educate them about what it is that is possible. I think a lot of them are somewhat technologically blind. They don’t really understand how revolutionary this technology is. And they need to have that explained to them. And I think it’s our role really to explain it to them. But there are significant efforts in that direction. So stay tuned, I think is probably the best way to put it.
That’s resonates with a lot of people here. Maybe more, I guess, the other people you have to convince is just the general public at large.
Yeah. You do. But of course, you mentioned two things. You mentioned this in your talk. The first one is, why would participants even bother to take part in trials like this. Well the immediate answer is, because you’re contributing towards hopefully curing these diseases. And that’s an amazing motivator for a large number of people. So would anyone want to participate in these trials? Yes, the motive is there. Clearly, and actually when we survey people and we ask them this question: would you like your continuous smartphone data to be available to the research community to help research into Parkinson’s disease, there’s just nobody who says no. You know, again, it’s because everyone is on the same side here. The idea that there are, you know, that people want to block research is nonsense. But of course, if you ask a different question, you get a different answer. And that is: are you afraid that your data could be misused in some way in order to harm you. And they say yes. So there is a tension, right, there’s a tension between this kind of very invasive data and the need and the want to help. But participants very often don’t think about it that way. They don’t really know how valuable this data is. I mean the FBI does, the NSA certainly does. So you know, it’s not as if we’re all clueless about this. I mean we know what can be done, we know what metadata like this will do and how invasive it can be. The difference is of course that we’re not the NSA nor the FBI. I don’t give a whit about what somebody does. All I care about is, you know, how is that relevant to helping the science. And they appreciate that and of course that is what they sign up for, when they sign up for informed consent to participate in these trials.
Of course we can make noises about security and encryption, but the reality is of course, if you’re moving data around over the internet, then the NSA is also able to tap into the data. Because they pipe all the data off the internet. So you know, it’s a silly thing to think that your data is perfectly secure. It’s not. That’s just a fantasy. But of course what you can do is you can do your best to try to put as many locks on it with encryption and industry standard process which we do in order to try to make it much harder for nefarious actors to do something with it. And of course that's what you have to convince your IRBs of. But down the line, yes you know, this data is unfortunately continuously available as you put it, in a way our smartphones leak this data all the time. And in fact if you talk to organizations like Samsung, they won’t tell you, but it is a fact that they are collecting continuous GPS data. Well Google is collecting it continuously. So they know where you are, as you say, all the time. So you know, given that that’s the state of affairs as it is now, we have I guess walked into it ourselves. The question that comes up is, you know, are we willing to throw away that data, the potential for that data, purely for the sake of the invasiveness of it, or are we willing to accept that risk and at the same time contribute towards potentially curing ourselves down the line. It’s a complicated question. I think the answer is everyone has a different answer to that, and therefore we can’t—you know, all you can do is you can present people with the option to be able to participate and the option to be able to withdraw. And of course that is a reason why one of the things that we do when we do these kind of trials is to give patients the full access to the data and the option to be able to totally delete it if they want at some point in time. At least totally delete it from our servers. There you go. You know. So I don’t know. Those are the ethical issues of course that come up with it. I don’t know if that answered your question.
AUDIENCE MEMBER: So you mentioned the data harming people. Insurance industries love the idea of collecting data, finding out if you’re going to get a disease because your parents had it or your dad or your mom or somebody had it. And then they stack the deck nice and high and they give you a really high premium. Do you think that’s going to impede the kind of work you’re doing where nobody wants a high premium, nobody wants to admit they smoked or they’ve had a bit of chest pain, especially before you get an insurance premium?
So is it going to impede the kind of—I won’t wander into the ethical minefield that is that question which is of course, why is your premium larger than somebody else’s, and should it be, is it fair that the premium—should it be priced at the average or should you price it more individually, you know, because that’s kind of an ethical difficulty. But does that hinder? I would suspect it does to certain people. And in fact when I’ve given talks like this before, like I’ve asked this question of the audience, you know. And one participant in the audience said, well I see all that is going to happen is that insurance companies are going to use it to price insurance premiums and set that as being like okay that’s the end of the story, we shouldn’t do it. That seems to be the conclusion that I got from that. But I think that’s a rather sort of adolescent view on it actually, it’s much more complex than that. It clearly is, you know. So I think my view as a mathematician, is that without the data, you cannot do the science. If you can’t collect the data, you can’t collect numbers, you can’t do the science, then we might as well give up. That’s the way I view it, you know. Then the question is how do we deal with that down the line. And it’s much more of an ethical question, I think that comes up. But yeah, it probably is a hindrance, yeah.
AUDIENCE MEMBER: You mentioned quite high recruitment in quite a short number of months. Did you do that solely by working with PatientsLikeMe or did you do any advertising or any other campaigns to get people?
A lot of this is about having good relationships in media. Because the media wants to help. Again we’re all on the same side here. We really do want to help. And they want to help. And a lot of the motivation I suppose if you’re cynical about it is a lot of journalists would rather work on something that’s more interesting and more potentially beneficial down the line than work on the latest stories about, I don’t know, Charles and Diana, I mean, I don’t know, Prince Harry. It’s kind of like—so that’s largely how we’ve done it. It’s a big effort to try to rope together all the necessary agencies involved in getting that kind of recruitment. And you know, but we can use that to our advantage, you see, in order to reach out to large numbers of people. So there are already existing channels there if you like to reach out to a large number of people. The way I see it is that we’re just, you know, pushing that attention away, a bit more towards where we want it to be, you know, where it would be beneficial, because you know, maybe people want to just watch entertainment all the time, but you know. But if you ask, again, if you ask people this question, what do you want to do, do you want to be part of the solution or part of the problem, you know.
AUDIENCE MEMBER: I think it’s interesting when they released the Parkinson’s app on the Apple health kit that that got about hundreds of patients in less than 24 hours.
Thousands. Tens of thousands I think.
Something we, again hearkening back to the industry, our struggling with the wearables point of view at the moment is—this was kind of covered in detail in Jessica’s workshop, but Willie also touched on it, in regards to accuracy and concerns around the accuracy and differences around the accuracy of those devices. And this is something you kind of touched on but didn’t go into any more detail, but I was wondering if maybe you would talk on the fact that technology gets better, that’s really what technology does and seems to get better at an accelerated rate. And that’s just a feedback loop, that’s what technology does. And I guess we can apply that to wearables as they exist at this point in time.
Yeah exactly. I mean, well my view on this is that, so if you look for example at the accelerometer in a smartphone, versus an accelerometer in a wearable device, they are the same hardware. In fact, they may even be made by the same manufacturer in China. These are just IMUs, yes. There’s a bit of a fallacy going around that these are not accurate. This is this, you know, the curse of validation, the curse of the idea that seems to be going around in clinical minds about what validation means, they are hopelessly confused about what that means. The technology is exactly the same across these different devices. That’s not the issue. The issue is how to turn it into clinically meaningful information, scientifically meaningful information. Because an accelerometer is exactly the same, it will always be the same across all these devices, that’s not the issue.
So how do we turn it into useful information? Algorithms. The problem then comes that—the problem is proprietary algorithms and a lack of access to raw data. This is the problem. Proprietary actors want for various reasons want to seal up their algorithms because it is their IP. And they also want to seal up the raw data because they don’t want it out there. But that’s fundamentally against the scientific principle, which is reproducibility. But if we have the raw data and we have access to the algorithms, then we can test how good these algorithms are. We can apply whatever, you know, whatever philosophic approach to validation you think is suitable for you in your clinical practice. So without the raw data and without access to the algorithms, we’re stuck, we will always be stuck. But I see some real hope in there, in the sense that the market forces that are acting to push towards standardization are actually acing in the right way for other reasons that have to do with scientific reproducibility. What I mean by that is, for example, the miracle of Android. There is a standard platform that works across all these different devices and it has an API that’s standard that collects accelerometry data from these IMUs, these inertia measurement units, which are all the same devices, and they’re all standardized across these different platforms. So therefore, if we have the raw data continuously available, we can create accurate algorithms, not necessarily us, someone out there will be able to create accurate algorithms that can be validated against various gold standards or metrics to create accurate clinimetrics, and that can be a process of indefinite improvement. But without the raw data and without the standardization that won't happen. However, I see there being very positive signs in that direction.
AUDIENCE MEMBER: So my question is towards the behavioural side that you mentioned. So in terms of how to quantify that, so if any particular sort of social activity for that day was a trigger of stress, which is one of the symptoms you’re capturing, which causes a flair, do you find the underlying cause? How do you do that, and how do you quantify that into your statistical analysis at the end of the day?
Yeah, I mean you’re asking quite a detailed question and it is quite specific. But the key to it again, I’ll just repeat, is just once you have the raw data, then subsequently you can make sense of it. And it’s about having enough of it that you can actually do something useful with it, like you say, which is for example, if perhaps you have an interest in say what you might call stress, however you define it, then for example, let’s say if we have a standardized measure of galvanic skin response, I suppose down the line then I guess stress is highly correlated with skin moisture. So you could say moisture—a galvanic skin response sensor is what you need, a standardized device like that. So of course it would be more complicated than that. It will inevitably be more complicated than that. And there are serious confounders in this kind of data. But you need to know what they are. And the way to know what they are is to have the context, to have the rest of the data available. So for example, a simple one is, if you’re really trying to measure heart rate, it’s no use if the device isn’t on your wrist. I mean it’s obvious, right. Well how do you know that. What if you have a proximity sensor. And that tells you whether or not you're wearing it. Another example. so you could pick up, you can measure gait very accurately with smartphones. How would you do it? Well it’s really got to be in your pocket. You can’t measure gait if it’s not in your pocket, if it’s sitting on the table. But how do you know that, well you know it because it has a proximity sensor. And a proximity sensor pretty much tells you whether it’s in your pocket or not. Now of course you’re going to say, yeah but it may be in your handbag or something like that instead of being in your pocket. True. But you then look for the inertial signs of it being mechanically coupled to your hip versus not being. And then you know actually, and a lot of information about how momentum is being transferred around is measured by this accelerometry device. So it’s quite specific. It’ll see. And the answers depend quite specially on what you want to measure. But suffice to say that, I think, the full suite of sensors that are now available in this consumer devices pretty much answer most of those questions for us. Of course they’re never going to be 100% accurate. That’s the nature of this. But no device is ever 100% accurate. But we’re moving quite quickly towards the point where a lot of these devices are essentially like, let’s say—vital signs monitoring is almost as good as a 12-lead ECG in an ICU unit. So at what point do you need the clinical device?
I don’t know if that actually answered your question. Slightly.
AUDIENCE MEMBER: Just something we were kind of asking ourselves. How do you actually use the telephones and smartphones to predict Parkinson’s? I mean are you asking people to read a passage? Is it specific questions? And what other things are you measuring? Is it gait? Is it activity? Is it stress? Is it heart rate?
So there’s a couple of questions packed into that. And the first one is, you asked how would you measure Parkinson’s using telephone. The key to the answer to that is a word, dysphonia, it’s a technical word for voice impairment, which is to do with fluctuations in motor control, fine motor control of your vocal organs. And so you can set up very simple tests which probe those symptoms. And then you’ve got the voice recording, so you have full recording of their vocal behaviour, and then you can apply various algorithms in order to extract important clinimetrics of that impairment.
AUDIENCE MEMBER: Michael J. Fox and Ozzy Osbourne, very distinctive voices, but both have Parkinson’s.
Yeah, that’s right. Not everyone has the same kinds of symptoms. There’s actually a whole range of symptoms. This comes back to what I was saying about heterogeneity of Parkinson’s, it’s a very complex disease. And it may not be indeed one disease in fact, maybe several different diseases. But the way to understand that—well there are various different ways of doing it, and we’ve got quite sophisticated algorithms out there now, I mean things like a high accuracy machine learning algorithms can to a certain extent can map clinimetrics onto outcome measures pretty accurately. So in that framework you don’t really need to know a lot of what’s going on inside this black box, because you can just cross validate it, or you know, externally validate it against different data sets. But then, you know, that doesn’t give you expansion power and in a lot of cases you want more expansion power. And so that means more sophisticated models of the biomechanics of impairment. But the general point here is that without raw data, you won’t be able to do that subsequently. That’s a research effort, this is a huge research effort that needs to happen. If the open data is available, shared freely in the scientific community, then you can bet some smart graduate student will figure out how to. There’s enough of them around.
I was very struck by your use of the term intimate in relation to the move towards smartphones, and I was wondering how you foresee these trends continuing. Do we now go beneath the skin? Or where do you think the technology is going?
I mean it definitely is going beneath the skin, that’s for sure. It’s quite a big barrier to cross, in many ways. But it’s already happening. And what I was looking at more recently, something that caught my eye was 3-D printable biocompatible electronics. So I think this just shows you that this sort of technology—that sort of technology is going to be crap for many many years. Until it’s no longer crap. And then of course it will suddenly be cheap and ubiquitously available. And we’ll all not notice. So I think the truth is that yes, that’s certainly on the way, yeah. Because then of course you’re talking about like real-time proteomics, which I think is going to change a lot of things.
Google brain chip, here we come.
Something to look forward to. Do we have any other questions for Max before we wrap up? Okay, I think that was a suitably brain expanding presentation on which to end these two days, so it’s just left to say thank you very much, Max.
[END AT 47:00]