#434 Vocal biomarkers: Detecting the unspoken with Kang Hsu from Canary Speech

0:00Welcome to the Health Tech Podcast. Here we talk about everything healthcare and technology. and I'm your host James Summery. Hey everybody, delighted this week to be joined by Kang Xu who is Chief Medical Officer at Canary Speech as well as having one of, I would say like one of the coolest career paths of any clinician that's been on here to be honest in terms of everything that you've pretty much done tech from every side of every table in healthcare as a clinician and you've got that chief medical officer position it's the dream for so many people actually going through uh going especially just medicine in general now that are looking at technology and looking at ways to work the way up so it's going to be awesome talking to you about your career path because you've worked at some amazing places as well um out in the us and so yeah really looking forward to getting into that with you man and obviously yeah to talk about canary speech and everything that you're doing there but welcome how are you sir i am great and thanks so much for having me on the air so you're very welcome man you're very welcome as i say really comfortable to get into it so whereabouts are you speaking to us from today kang i'm uh talking to you from las vegas nevada vegas are you based in vegas i am based in vegas and i was uh i'm here currently just for the cs show as well yeah that's on at the minute isn't it i was looking at some of the um i was looking at some of the technology a lot of the b2c technology i was actually looking at so i saw like smart scales and some like some hellish led masks from l 'oreal that they were you know trying to say is like health tech the the line between like consumer and and health tech and clinical is like getting quite blurred isn't it what you what are you seeing at ces what's going on i mean there are some crazy televisions from samsung that i saw come out um samsung's definitely trying to reach out into the health health space now oh interesting yeah like definitely samsung why are their tvs oh well actually tvs and just because samsung is in so many homes right so many products in the homes they're they're using these products to listen or watch what you're doing in the sense that maybe they can help detect falls they can help if you're always forgetting your password or typing mistyping maybe that's an early sign of a cognitive disorder so they're uh talked about their brain health uh something they're going to be i think releasing in the oh wow upcoming months or years so a lot of that stuff lots of wearables lots of wearable stuff that we saw and a lot of robotics what's the trend looking like with wearables then are they getting bigger smaller more far-reaching are they in interstitial fluid or blood or like what are they what are they doing not ambitious a floor of blood yet that i saw but definitely um you know the rings are getting smaller and doing more okay um there were some talks about just even what was like a little drop of blood be able to uh look for a concussion um lots of different uh different things that we saw out there it was pretty cool and again lots of robotic stuff lots of robots doing you know there's a ping pong robot there was a robot robot you could kind of box with which seemed interesting so yeah that would be quite interesting when it just decides it wants to win that's exactly what i thought when i was watching that that's like how much do you really trust whoever's behind the algorithm or the ai oh dear that's fascinating how did you feel as a i guess as a clinician that's involved in tech going through that do you do you feel optimistic as you go around looking at the technology i i mean the health part definitely optimistic right because we're now bringing bringing um things to to the patients so where the people could actually use things at home so they they're going to have a lot more data at their own hands and of course if they're willing to share some of that data with their healthcare providers, it's going to give the providers a much larger data set to work with.

4:08Yeah. And then of course, having AI decipher some of that data for the clinicians is helpful just because if you give me a year of your data, I'm not going to be able to look through that while I'm talking to you. But if I can summarize that, it helps me and it also helps the patient to just know. i think if a patient knows that they're not sleeping well uh they might actually start doing something with that right if their watcher ring tells them or you know if the watcher ring tells them they may have sleep apnea it's something that may not have been noticed but now it can be caught sooner right so that something can be done yeah it's interesting isn't it that layer between all of that wearables wearables data the patient reported whether it's outcome stuff or whether it's just you know my fitness pal and throwing your food diary in there there's all of that as well there's just there's all of that data and there's the layer between that and the clinician or that and the self and going well yeah an ai layer in there to decipher it and give us decent insights that actually get acted on don't make us additionally anxious is the right level for us as individuals and it's personalized it seems like well chat gpt health right that was what yesterday that announcement day before like is is yeah i mean that's literally what they've just talked about isn't it you know throwing all of your well throwing everything in there throwing the patient you know the the letters from the hospitals the like all of your data like this that the other throwing it all in there and chat gpt is now going to be that isn't it's going to literally be the the companion and it was quite interesting actually when um because they released their report didn't they the sort of day before that as well like earlier in the week and they were saying you know in the us five percent of all chats uh from for health and all this sort of stuff what was so interesting there for me was the um it was at the end when they were looking at like the recommendations and they just fired some shots at the fda being like you will regulate us like you will yeah we would urge the fda to consider how they're regulating things like this full stops or something like that and it was like okay they are they are fully going after medical device status and they're not afraid to tell the fda exactly how their process should be i thought that was fascinating oh what was your what was your thought when you saw the chat gbt stuff i mean i think chat gbt is being used quite a bit obviously from people just trying to ask it's being used anyway is this really an announcement question mark yeah yeah i mean it's being used either way right it's being used it's being used by clinicians too right i mean it helps it helps us too but it definitely helps folks just trying to figure out what might be happening or what might they what might they do that might improve their health i mean obviously be better coming from a physician but it's better than not doing anything that you were doing right so which is a huge part of their argument they make a they make an incredible accessibility argument when they start saying the seven out of ten of the chats are out of hours 600 000 people live in underserved areas or something along those lines that are using it underserved rural areas and and the video that they had at the top of their press release that they put on their website or that was what the case studies were it was people in a cabin somewhere in the middle of nowhere that have a chronic disease have a flare-up of that chronic disease and then all of a sudden are in loads of trouble apart from the fact that they can now converse with this thing that knows all of their data and everything it's it's just it's like are we in a are we in a dystopia or are we in a utopia that's the question that still remains doesn't it it's like we're at this fork in the road of all of this stuff who knows where we actually are um but anyway man i know i'm skipping ahead but great to have you and actually really want to really want to talk to you about your career first of all to kind of orientate everybody into into like who you are and and what you've done so I mean why don't you tell it you know starts at UCLA but might even start before then it might start with you as an entrepreneur as a child or I don't know whether health tech was talked about around the dinner table or entrepreneurship or whatever but like I don't know man like how did you grow up and and what what what do you think set you up to be the health tech leader that you are now it's very funny you know when I was growing up my my parents wanted me to be a physician that that's all they talked to me about and i think uh when i was probably laying in the crib my i think they were whispering become a doctor become a doctor but uh you know both my parents were engineers okay it's kind of funny they wanted me to go the opposite way oh interesting they thought it'd be a safer career to be a physician and you know do good etc but i always had interest in tech you know at least video games of course that's growing yeah but um yeah so i i think just always being asked or told to be a physician kind of led me towards science so i had an interest in science and ended up going to medicine um that's kind of how i started out and you know became a i became a hospitalist so uh internal medicine where you work in the hospital take care of patients right who are ill that was at kaiser right i did some moonlighting at kaiser but i actually started my actual job was actually at a st joseph's health so i said joseph's hospital in orange county in california nice so i was hospice there for a while and it was right at that time where you know we're still on paper right we're scribbling our notes scribbling orders i probably have one of the the messiest handwriting of anyone in the hospital so well that would be that that would be a heck of a competition for like who actually which which clinician has the worst handwriting i mean wouldn't make a great reality tv show but it'd be it'd be hotly contested that that prize definitely so you know no one could really read what i was writing and around that time maybe 2010 or so you know the hospital was going to start implementing or bringing in this electronic health record so um they wanted doctors to help and no one wanted to help off right because everyone was like this is gonna just make everything worse worse for us um so basically when was the i think it was right around 2010 i would say is when my hospital started going into this so i was you know our group was asked which one of you is going to help us do this medical record and all the physicians in my group said not me so i kind of got involved and told i would say to uh be the guy or be the physician who helped at least from our group and that ended up probably being the best thing that ever happened i started thinking why would i want to type orders or type notes into a something that you'd have to read when i could write much faster but it turns out of course it's probably a lot faster to type right and you won't type the wrong medication that someone couldn't read because they can obviously read what you type yes or later you would be clicking clicking medications right so that's kind of how i got started um so i kind of helped our hospital go up at that time on meditech so as that happened we had a chief medical informatics officer at the health system level that you know i never really knew but um they kind of then reached out to me to start helping do the same thing in our hospitals and you know texas and in northern california and so forth so that's what kind of brought me into informatics um without knowing i was going to do that amazing so that's kind of how it all started yeah and so from there then did you keep practicing medicine fully or was part of your job becoming yeah this i don't know what was it would even been called then like informaticist or it guy or like health it yeah i was just like the it doctor is what they kind of called me but yeah i was still a full-time then but as i was doing more and more i ended up doing part-time so a hospital schedule affords you like a week on week off kind of schedule so i was able to do a lot on my weeks off.

12:29So when I went part-time, I would do one week of clinical practice, right? Seven days in a row. And then I'd have three weeks to do the informatics work. So that gave me a little more time. So I ended up reaching out to the electronic health record company, right? That we were using and then kind of cold called them and said, Hey, would you ever want a physician to help? So they ended up hiring me there. So I did a lot of sales and implementation of their product. just with various health systems across the country. So what happened next then? Because I've seen, so now you're chief medical officer of Canary Speech.

13:06So a startup doing very cool voice biomarker stuff, which we'll definitely get into. But how did you plot the path then from what you were doing as the clinical IT guy, the medical IT guy, to then was it an interest in the startup world? were you exploring that stuff like how do you plot the path between the two no so the the funny thing is uh after doing a lot of this work i ended up becoming the chief medical informatics officer for orange county california uh then a little bit of northern california so we uh the saint joseph health um health system ended up kind of being merged into providence health system yeah and Providence is much larger, right?

13:54So they have about 51 hospitals. So as I continued doing this work, I ended up becoming the acute chief medical informatics officer. So for all of our hospitals and for informatics related to lab and pharmacy, we had like a physician success team where we would help physicians use electronic, you know, electric, sorry, use technology, right? Because a lot of physicians don't want to interact with technology unless they have to. So when I was doing all that, we also had a grand round. So we had an innovation grand rounds where I worked. And not a lot of people were helping, I would say, find speakers.

14:34So I think actually in 2005 at HIMSS, I ran into Henry O 'Connell, the co-founder and CEO of Canary Speech. And he was in the Microsoft booth showing off their technology. Nice. So when I talked to him, I was fascinated by what he was doing. And since we had no speakers to do our grand rounds in the next month or two, I invited him to talk. So when he spoke at our grand rounds, lots of doctors, they don't always ask questions during the presentation, but they're texting me. What the heck is this? I've never heard of this. This is real. Ask this guy questions for us. So after he presented at the Grand Rounds, I was constantly emailing him questions, send me references about the research you've done, what other areas are you guys working in, et cetera.

15:27So as we talked more and more, one day I just kind of asked him, is there a chief medical officer in your company I can talk to so I don't keep bothering you since you're a CEO? So he had mentioned just at that time they were literally about to start hiring or looking at hiring a chief medical officer. So I said I would actually be very interested in that position. Wow. That's actually what led me. It's just a series of coincidences led to where I am now. Yeah, a series of fortunate events, but you make your own luck, right? And this is the thing, you know, you put yourself in the right place at the right time for that sort of stuff to happen by, you know, going to that booth and going to that event and all the rest of it.

16:06And you had a genuine interest again to speak and, you know, all that sort of stuff is great. I think what I want to ask you, though, so Providence, right? When you're leading, for want of a better phrase, you're sort of leading the adoption of health tech in the whole area. That's a heck of a job. That's a heck of a role. And just looking here at some of the stuff that you did, partnering with Microsoft on AI in healthcare and leading an EHR implementation for 16 hospitals, striving to use technology to achieve the quadruple aim so you're you're looking you know you're looking for loads of technology you're helping bring it in you're helping those sites and organizations adopt it it's you know it's a big role there's like a lot there's a lot to do there but but i'm interested in are you optimistic about the adoption of tech in healthcare after doing that job because it's interesting to me that you know i imagine a job like that you it has a lifespan for how for how you know probably tiring it is and all the things that you see and it goes in cycles and that sort of stuff and then you've now stepped out to be on the other side of the table as well which is quite interesting so yeah are you optimistic or pessimistic based on what you've seen like is health care because the big one's ai right like and all this sort of stuff.

17:36And you said, you know, partnering Microsoft on AI and healthcare is one of the things that you did in that job. So, yeah, given your role there, given everything that you know, what do you think about what's going on now? And are you optimistic? I would say I was probably carefully optimistic, but now I'm a lot more optimistic. I think different technologies were probably harder to get folks to adopt. You know, you always have your innovators, early adopters. who are going to want to try it, but it's the other clinicians that take more time. I mean, if you can prove to them that what you're asking them to try out helps patient care, has research that shows that it actually is effective, has some kind of ROI attached to it, they still may not use it, right?

18:22But if it doesn't burden them in any way to use the product and it have the other three factors, then most of the time they will use it. But AI was different. I didn't think AI would be adopted by the clinical space that quickly. Clinical stuff happens very slowly. I think health systems are probably one of the slowest conglomerations to ever adopt technology. But AI and our health system, when Chachibiti came out, we had a Chachibiti kind of product already built in within months, I would say. Really? under inside our walls secure that people could use um so that it was already there you know we adopted it was doing more we didn't want it to really be used for patient care but we were already allowing you you could use it to you know help help write your reviews for the your direct reports yeah okay you could have it help you uh generate your emails and stuff like that but we were pretty quick to do that we um you know we adopted microsoft copilot so that you know i i don't know um i i definitely get lots of emails back then where it would be like hey can you take a look at this and they'd be you know one year of emails between 20 people so with the ability to kind of at least summarize it would give me a chance right rather than having to read yeah i don't know how many emails like that so we were pretty quick to do that which was nice yeah um then we you know we started of course with the uh ai scribe ambient technology right so we were a microsoft shop so getting doctors to adopt that took a little more time but the ones who started what was the key to getting that right yeah i think uh when it first came out uh the product had the physician talk to the patient and then somebody would be drafting the note for you so you wouldn't get the note kind of in real time that's obviously changed now right so now now when you're recording and talking to the patient when you're walking out of the room the notes are already ready for you to review so i think once that step happened there was a lot more optimism in using the product so um it definitely helps our physicians save time right get some time back and actually spend more quality time with the patient you did a lot clearly of implementations of various things in that time and oversaw a lot of technology coming in.

20:46You said that the key is not necessarily the innovators' early adopters, it's the early majority and getting the early majority in and that being the key to a good adoption. Broadly, what are the principles to getting that early majority? You said making sure that they're not burdened, which might seem like an obvious point, but can you like flesh this out a little bit reason i ask there'll be lots of people listening that are on a lot of different sides of this table there'll be early stage entrepreneurs thinking about a new technology there'll be relatively established startups that are on the scaling part where you know adoption implementation and speeding that up or making sure they get a decent decent utilization of it is key to them getting the contract after the pilot Or similarly, there'll be people in organizations that might need to coach their workforce because the organization wants something to work, but they just need it to be utilized to work.

21:51So there's like adoption being such a key part, you doing so much of it. You know that the early majority is the key here. So what principles do you have for everybody here to making that work? I would say if you have a product, no matter how great it is, if you're expecting that you need a physician to open a separate window or click some extra buttons to do something, it doesn't seem a lot to the person that's trying to sell the technology. But to the person who's using it where they have 10 to 15 minutes to see a patient, these extra clicks add up, right? I mean, the whole thing for us is click reduction.

22:36how do we let doctors click less so they can do more that's actually a work stream click reduction is a work stream it literally is something we work on right how we decrease the clicks for physicians right or and nurses of course but i think that is probably one of the biggest things is if it could just be seamless within the workflow i'm doing anyway it would help a lot to get people to adopt the technology. Again, it could do great things, but there's so much burden on the physicians and nurses right now. Anything actually you're asking, because that's one technology. And they're like, well, we also want you to use this, so can you do these three things to get that working?

23:18And then can you use something else where you have to turn on a whole separate device? Each of those products wouldn't be as much of a burden, But now you're using five products a day that will help improve your technology. So if everything is streamlined, seamless, all integrated into one thing that you can use, where you can use these three products without having to log into something else. I think that helps a lot to get the early majority. And of course, it has to work well. It has to actually improve patient care. It has to actually help you save time or be more efficient, you know, or actually obviously not miss something, right?

23:55If it's helping you do those things, it's going to help. A lot of people find that the innovators, early adopters are very forthcoming with their sort of want and desire to help and for it to work. The early majority, however, are rightfully more, they're more cynical because they don't have the same joy from technology. They do the job and just want their job to be better. they don't feel any sort of emotional attachment to the tech or any you know they don't think it's cool or any of that sort of stuff they just want their life to be better so what how do you beat that cynicism especially with a new technology that might not necessarily have the evidence yeah i think part of it is actually relationship building so i know a lot of the docs that i work with and i know a lot of the grumpier physicians i didn't want to say it but that's the word i was looking for i mean even during the pilot phase you know when you ask for volunteers we're going to get all the people who want to try it out yeah this is the thing they're already excited about it so we actually would try to get obviously someone who wants to use it someone who's tech savvy but then we also get the folks who wouldn't really want to use it at all to try to pilot it because they're going to give you different feedback than the doc who's going to love it anyway yeah um so So we would always try to get a good mix of people in our pilots to do it so that you can actually get all forms of feedback, right?

25:26Not that like, oh, this is great. We need to turn this on. We want the folks who are going to tell me like, I don't like this at all. I think this is horrible. But if you did this, maybe it'd be better, right? Or again, if you could make it just log in when I turned on the HR, then maybe I'd use it. An innovator is not going to tell you that because they just want to try out the product they might tell you but they're gonna like it either way as long as it does something cool do you have any stories of things going like really badly wrong that you managed to learn from like any sort of disasters any like dumps to fire like implementations without naming like actual products yeah i mean we've over the years we've had different products that probably didn't go the way we thought they would um but instead of trying to go all the way through we would actually take a hard look at it right in our committees and actually just stop the project um i'm not gonna name which ones those are but there have there have been there have been a handful where you know it's not going to do what what was promised so yeah no i understood man so So let's talk about Canary Speech.

26:35So how, what's it like jumping onto the other side of the table then? Because all of a sudden now you've got to start convincing probably your old colleagues or your previous counterparts in various different regions and going on the other side of the table. Yeah, what was it like making a switch? It was, I mean, obviously nerve wracking at first, but very exciting. I think, you know, I put a lot of thought before I moved into this role, but I like the technology. It happens to be seamless. It doesn't seem to burden any physician on what we're asking them to do to actually help improve patient care.

27:13So it fits your rules of what you would have adopted. Yeah, I made sure it fit what I would want before I actually took on the role. But yeah, it's been great. It's definitely different, right? I came from having lots of direct reports and a lot slower moving kind of health system, right? Or any health system to something that's fast moving where we're looking for the slow, right? The slow health systems to try something pretty novel. So there's been some excitement. I mean, obviously, a lot of places are going to want to wait and see, right? But then we have people already early adopting, using it with a good outcome.

27:52so it's been great so far i have an amazing team and an amazing ceo that i'm working with so nice let's talk about what it is so um i know that you sort of oversee the whole clinical direction there which we can i definitely want to dig into what that actually is and what that role means and um all that sort of stuff but let's let's talk about what it actually is first so what is a voice biomarker and what do you guys do with them so yeah um i would say the common misconception is people think we're using technology to decipher what what people are saying so actually canary speech is looking at vocal biomarkers so one step below actual words right so we don't look at what you're saying we're actually looking at how you say it so how your vocal cords vibrate the jitter, the tone of your speech, not actually what you say.

28:47So basically, these vocal bond markers are subconscious, right? I don't actually know how to control them technically, because I just, I'm just speaking words. So what makes it fascinating is we're looking objectively at that speech, not subjectively, right? A lot of times a patient can tell you something, it may not actually be what they mean, right? You know, we have a lot of screening tools where patients answer questions but we're needing them to answer them honestly or actually they may not even know right you know a lot of surveys will say sometimes most of the time all the time within however many days have you had these symptoms i mean imagine a patient who's already ill trying to answer that so we're looking objectively at how you're saying something and we're basically when you're talking we need about a 40 second voice sample and we're looking at around 2 590 vocal elements every 10 milliseconds so if you imagine hold on 2 590 vocal elements what's a vocal element so like just different uh vibration jitter tone those are all these different kinds of wow so we're only at 25 so 2 ,590 of them every 10 milliseconds so if you imagine like a minute of speech that's about it's over 15 million vocal elements we're looking at but if you're looking at the word someone says to you in a minute maybe 300 words a minute so the amount of data we're able to accumulate to do our analysis is a lot larger than if you looked at the words which again are subjective do you know what's really interesting that just came to my mind then i wonder how many of those are perceivable by humans especially a human that knows that person incredibly well i think i'm finding this fascinating at the minute that as we go through as we go through time and our use of technology gets better and better that we are and I've spoken to a few people on the podcast doing this now we're turning what was previously unmeasurable into the measurable and I think that yeah is a fascinating thing for me because medicine being so being practiced so much on like vibes quite often you know medical school exams you'll do oscies I don't know what they're called in the US but like the stations where you have to diagnose someone in five minutes each time like and your first thing is always to look at your overall impression do they look sick or not what's around the bed like how they walk it like and you and was you know that that's a few really crude type things of like are they flushed are they this are they that are they do they look gray like all there's all that sort of stuff is there you know we'll talk about speech and is there pressure of speech are they are they this are they that like there are these quite crude ones but i've named about 10 there which was about all that you would do you're talking about two and a half thousand of them just in audible speech that is it's so fascinating to me that we're starting to put metrics around what we think is well a human has to do that because they just know them and it's it's it's interesting to me that that's where this is going um i'm sure there are things that will remain just human because humans can only do them um but that is a fact that's a really fascinating metric for me yeah and then i think the even more intriguing thing is these vocal biomarkers can be used for clinical decision support to help detect certain disease conditions and that's important because when you have 10-15 minutes to talk to a patient in your clinic you're usually addressing maybe why they've come in for the day, you know, or you're monitoring them for their diabetes and try to see what you're going to do with their medications, et cetera.

32:53So while you're talking to them about that, if you imagine now we're listening, right, and getting 40 seconds of that speech from the patient and then telling you this patient might have depression or this patient might have mild cognitive impairment, this is something you weren't even looking for because you're taking care of a different issue. We're almost taking like a different history than the history you're taking at the same time. So if you can detect certain diseases earlier, you can, of course, intervene earlier, treat earlier, maybe prevent morbidity and hopefully mortality. That's, I think, the amazing part about this technology.

33:34And it's non-invasive, right? Non-invasive, it's non-intrusive. And if you're using, a lot of people are now using, right? Generative AI or ambient scribe technology you're literally we're just listening in right we're on top of that software anyway yeah listening anyway and then giving it to you without you doing anything else so my mind's racing with all the different places that this can and could be used and should be used and in part i'm like god is my is my uh is my health privacy at risk because i've got 400 hours of podcasts and people are going to be able to put that through and figure out when I was depressed or not I don't know or struggling but that is what it is but who who's using this and when are they using it so what I mean by that is are you guys thinking of this as okay there's a lot of telemedicine going on at the minute is this something that just runs in the back of telemedicine is it and then I'm like well actually are someone's already interacting with the healthcare system is that too late are you looking to pick things up even earlier is there a is there a thought where this should be used more often is this something that should be just attached to the phone and runs in the background of your iphone so that it's always on and listening and can nudge you and and help you and all that sorts of stuff and where does this become a lifestyle wellness or kind of always on preventative uh device i guess versus is this an uh something that you have to interact with when you interact with the healthcare system so just do you see what i'm getting at like where my question is there i haven't asked it very well at all but you know what i'm saying like where actually you where is this probably you may have answered your own question lots of lots of uh forms this can be used that i think i'll name a couple you know the the classic one would be primary care clinic.

35:31Primary care clinic where the clinician is using an ambient scribe. So they're going to turn on the scribe to record. We're piggybacked on. So we're currently integrated with Microsoft Dragon Copilot. So as you're seeing the patient, we have diarization so we can tell if it's the patient or you're speaking. And once we've picked up the amount of time sample we need for the voice. It shoots up, of course, encrypted into the cloud. We do our analysis. We shoot it back down encrypted, and you get a result on your phone, the same device you're doing your ambient scrap technology on before you walk out of the room.

36:07So it basically would give you a nudge, you know, maybe we've detected mild cognitive impairment or something. Then it's, of course, the clinician's decision to either maybe talk to the patient a little more about that, or if they're very busy that day, they can maybe repeat the canary speech screen on the next visit. If it's still showing the same thing, then maybe they ask a little bit more about memory problems, right? Or they conduct a MOCA, a Montreal cognitive assessment to see if there might be something there. And depending on their findings, then they may have to refer to a neurologist.

36:46That's definitely one bread and butter use case. another use case you also mentioned telehealth right so we're actually uh in zoom so if you're doing telehealth you can actually yeah so we're we're integrated with zoom you could just turn us on and again it's just listening in the background during your during your call with the patient uh you mentioned security and privacy issues so obviously you know just like with ambient scrub technology you'll be consenting the patient that you're going to be doing this but um but yeah in in a telehealth call it it's seamless right you don't again there's no burden on a physician to to use the software it's just running if you want it to be running um another space is uh wearables so um where we have a prototype in the samsung watch currently but you can imagine either someone goes home they're getting remote patient monitoring right the the watch is checking your pulse maybe uh you have other stuff like a blood pressure cuff But now we're also listening to vocal biomarkers and we can longitudinally monitor your progression, right?

37:54Your improvement or worsening of symptoms and different disease conditions. That's another great play for the technology. We also have a very novel thing we're doing now. And, you know, there's a lot of, I'm not sure about in the UK, but in America, there's a lot of violence in the workplace towards our, especially towards our nurses and other clinical caregivers, right? So not always the fault of the patient, right? The patient's ill. they're sometimes delirious, et cetera, but we actually can be in a virtual nursing equipment. So one company such as Teladoc, we could be in their device and we can be listening in while the patient's, you know, during their hospitalization.

38:32And as we pick up their speech or, you know, family or a visitor, we can detect possible levels of aggression. So we can alert the staff, right, that there's a chance that someone in this room may be aggressive. So Instead of sometimes where a nurse goes in not knowing and they're attacked, and then we put a warning in the EHR like risk of violence, we can now know before you're going into the room that something might happen. Obviously, every hospital has a protocol on what to do if there may be a risk of aggression. This happens every day in our hospitals where someone's attacked verbally, physically, etc.

39:09That's another use case we have. Those are the more clinical ones. You could also be in a contact center. You could be when you contact, when someone contacts you after someone's dismissed from the hospital, you could use it there. You could use it to screen patients for research. You could quickly screen to see if you might want to do more screening. Instead of spending like 10, 15 minutes doing a screen, like a typical screen, you could do a 40-minute screen on a phone call. And then, oh, this is a patient that might have something that we want to study for a drug study or whatever it might be.

39:46I think those are the most common use cases, and there's a lot more. We're working with a South American police department where they want to make sure their police don't have a behavioral health issue, right? So it's not easy to screen 600 police officers in the normal way, but if you use canary speech, you can screen them quickly. and then the ones that may have depression or anxiety or something we can then refer them to get care airlines for their pilots and there's also yeah crazy yeah definitely yep a lot of questions there do you need a baseline uh no we don't need a baseline because we already have studied the conditions we're looking at we've already done research studies on right so we've already looked at people with or without and then we can tell which vocal bond markers you've got ranges yeah okay yeah yeah that makes sense um one criticism that might be thrown at you is how sensitive versus specific is this and are you then just flooding the system with more referrals for more stuff so yeah we can go into that so um for parkinson's disease our sensitive specificity is around 97 okay so there won't be too much of it but of course there's going to be false but for that yeah i was gonna say for that obviously there's clear there's a clear physiological link between that disorder and certain vocal things happening like that you can draw a very clear line too so i'd i'd understandably you know i'd expect that to be you know high and the highest but there's obviously that is that is the highest right um so alzheimer's so alzheimer's disease is around 91 for both which is still very high um and that's for early detection this is early yeah so i mean you probably don't need canary species to detect you know and stage alzheimer's you would probably know so this is for early alzheimer's where you may not even notice that the patient has these symptoms yeah right but we can tell from the voice so if you technically if you did a scan on this patient you might already be seeing amyloid plaques right um for uh myocardial impairments around 88 percent and then for our behavioral health ones it's lower uh for you know let's say major depression it's around 70s right but the reason is because our our usual screening methodology which would be maybe phq9 that runs in the 70s so we can't be much better than whatever the gold standard already was because that's what we're testing against yeah sure i also think it's probably worth noting as well that in that scenario thinking about depression anxiety for example it's night and day how easy it is for someone or to get someone to do the phq9 score ring versus just taking the recording from the consultation because one of those requires a very kind of upfront you know discussion and asking and all those sorts of things that that's that's a that's a very proactive step that you need to take in order to then collect that and they the patient then needs to be very confronted by all that to then actually do it whereas on the other hand this is a this is a technology running in the background that's alerting the clinician to hey this might be an issue who can then behave accordingly and they will know the patient in order to handle that as sensitively as possible that's my read on it is that is that fair yeah totally fair and we've already had stories from physicians using a technology where you know when you ask a patient about their mood they're usually going to say i'm fine but they may not be fine right so we've had instances where the patient is telling the physician, you know, they're feeling great.

43:48They don't have any mood problems. You know, there's nothing going on in terms of that. But then the canary speech is returning a possible depression. So then when you kind of dig deeper and talk to the patient, they open up and they're actually suffering from depression. This isn't, it would be very common that this might happen. Just again, like you said, one, are you going to do a PHQ-9 every time you see a patient? you probably are but a patient doesn't get depressed once a year right they could be depressed anytime they're seeing you but you're not screening every time you're talking to them i think as well with with well with continuity of care where it's at it's it's not as high as it once was with the amount that certain clinicians will see certain patients and get to know them but it it you know it's still relatively common for continuity of care to be there and also So I think we rightly trust our human intuition quite a lot.

44:48But it's also incredibly validating and very nice to then also have a metric to point to. So if we are thinking, hmm, I wonder if, and then you're looking at Canary's Beach, which says, yeah, this is worth looking at. all of a sudden then you're you're more inclined to take that step knowing that there might be there might be a difficulty in that conversation with the patients you perhaps don't want to say it if it's not going to be an issue but actually if the metric is telling you and your intuition was telling you you're then far more likely to then go down that route I say so it's kind of empowering in that sense I guess for clinicians that are trying to back up what they're trying to what they perhaps might be thinking I imagine though that you can easily make the flip argument of like you know you might get stopped or told based on something that's incorrect but at the same time like we're just trying to hear it's validating and and actually i guess helping broadly yeah no i agree and again you if you don't feel it's something you should pursue you don't have to but it's letting you know good point you know maybe there's something going on maybe we should address it yeah or maybe at least delve a little deeper just to make sure right and rule it out yeah so what do you pick up then you've mentioned cognitive decline you've mentioned mild cognitive impairment you've mentioned alzheimer's you've mentioned uh other things like parkinson's disease we've mentioned anxiety depression how do you classify the different types of things that you do pick up and how far reaching is that yeah so currently behavioral health we're doing depression and anxiety um we're also starting to work on ptsd um yep and then in the neurologic space it's uh parkinson's disease and huntington's disease so obviously we know someone's going to have hunting disease right because on a little dominant but we can tell when you're going to go from pre-manifest to manifest sooner than you might notice and then cognitive you mentioned it's mild cognitive impairment and alzheimer's disease but we're also starting to study a lot of new diseases since i've come on well we have multiple sclerosis a study we're doing that we should have data in the coming months because you know we're trying to look at what are the diseases that might be harder to diagnose yeah also treatable yeah that's a great one that we're gonna probably have a great part the venn diagram for you isn't it because that's then you adding broader value within the healthcare system and being as useful as possible rather than you know the state of you know telling people the obvious type stuff i guess which would be the other end of the extreme another exciting one is um adhd right um parents often right parents might think their children have it but the way to get seen and evaluated could be months just depending on where you where you are but if you use another technology that's saying that this patient may have it you could actually then expedite these patients to be evaluated quicker right so we're studying that uh we're looking at pain right pain is interesting because patients uh so i'm actually also a hospice physician so a lot of patients will tell you they don't have pain, right?

48:07But they've got metastatic pancreatic cancer and they probably are in quite a bit of pain. So if you had a more objective measure of pain, it might help you treat pain better, right? So we're looking at that. We're going to start to look at lung cancer, right? One of the less screenable cancers, right? You know, breast cancer, colon cancer, we've got pretty good screening methods for that. But lung cancer is a whole different thing, right? So what if there were vocal biomarkers for lung cancer right so that's something we're going to explore and then we're also talking to a couple of places about copd and chf maybe not so much trying to diagnose it but what about longitudinally where maybe your hemodynamic pressures change i'm gonna say and your vocal biomarkers change with them yeah just supporting the management of it like uh right i can't put a hemodynamic monitor and every patient has a heart barrier of course but i could look at their voice so if there is a correlation we can maybe tell they're going to decompensate before the physical signs happen we can treat early yeah and again as long as you're beating the current the current precedent right so it's like what do they get weighed in some mechanism for retention of fluid and like these other things so it's like yeah it yeah it can either help build a picture with that stuff or it can actually be more sensitive more specific than that stuff which i think is the the aim then isn't it yeah and then one other one that we're looking at is suicidal ideation again it's a subjective thing sometimes right so if you could objectively see that in a patient who's coming to present with that you might be better able to treat them it's fascinating and i imagine for for some people listening it's sort of unbelievable that those types of things can be picked up when you're thinking of something like chf or congestive heart failure for those that might not be that medical but like heart failure or lung cancer what of those 2 590 different things is changing that affects the voice in in conditions like that that perhaps like for me now i'm struggling to draw the line between the two maybe maybe some edema in the in the larynx maybe or like yeah yeah literally was i say right now i I mean, so your lungs and your diaphragm obviously help provide the airflow for speech.

50:26And then, of course, the larynx and vocal cords are going to help generate the sound right through vibration. So if there's more edema, there's probably there could be some biomarkers for that, right? Yeah. So that's some of it. Even other diseases like maybe vocal cord cancers, there are definitely going to be vocal biomarkers for that. right so so we're trying to look at ones that should have vocal biomarkers but also will help clinically right as you're speaking there as well one thing that occurs to me is you're learning quite a lot as a company and you as the chief medical officer overseeing clinical direction it must be quite an exciting place from a research perspective because you're sort of strengthening your own hypotheses and conclusions there because over time as you learn more and and you get more data you're getting more sensitive and more specific in these areas you're able to pick up more diseases and and that sort of stuff do you have like specific work streams around research and publication uh we have an entire research team actually working on that that i obviously collaborate with so that makes it sound like i planted the question i didn't but Yeah, so yeah, we have a whole team doing this.

51:46We've got folks working on research. We have people writing the protocols, right? We work with different either academic health centers or even just clinics where we're trying to start research, going through IRBs, etc. you know we have a roadmap of the condition some of them which we just listed that we are targeting currently and we're somewhat opportunistic too right we will have sometimes an academic institution reach out saying hey what about this and that's actually how the suicidal ideation and actually a schizophrenia study we're looking at is they're asking is there a possibility we could do something like this so schizophrenia is a perfect one where i was giving a talk at a conference and someone who was working on schizophrenia thinking about vocal biomarkers approached us so wow um a little bit opportunistic and some targeted of course so yeah of course of course um my final question on the clinical stuff is like what's the what's the extent what's the clinical extent to which vocal biomarkers can be useful because you mentioned a lot of stuff stuff i imagine it's not beyond the realms of possibility that a change in blood pressure if significant enough could affect the vocal cords at some point you'll stop being able to speak so there'll be some sliding scale there but i you know i wonder if the minutiae of blood pressure can be tested in this stuff or particularly if you've got a lot of data on the same person's voice and all that kind of thing and heart rate and like i honestly don't know but it is is there a worldview where it replaces monitoring and i know that sounds a bit weird but like in that thought experiment is is that a possibility or indeed like what's the what does the future look like here in terms of the realm of possibility i mean i think there definitely is is an area that could help out in even for blood pressure vital signs we there are a lot of facial diagnostics actually i saw that at cs2 right where oh nice they look at your face for about 25 seconds and they can tell your pulse and your blood pressure i don't know exactly yeah just subtle changes in color the pulse i can see them getting right yeah of course so there's some of that maybe combined with voice maybe combined with other monitors right maybe that can help i think another space is just wellness in general you know we also have a wellness side to this right where you could talk to your watch or right and you can get a wellness score from your mood right or your energy levels uh and if you merge that with sleep and your heart rate and all that you get a little more powerful wellness score so that's something that do you know i was just i was just thinking as well kangla you know i've just got um i've never been a wearables guy whatsoever but i've got a one-year-old and it's awakened this don't die instinct in me um for him at least uh so well you know for me for him kind of thing so i've got um i've just got the ultra human ring so the sort of aura ring without the subscription that one yeah and you know i'm just thinking as i sit here now you know what like if if if this just clocks that i'm speaking and my hands relatively near my face like this there's no reason why this couldn't just you know whack up a quick recording and and do the processing on it and give me as part of all the little modules i've got in my app one of those modules on the assumption of mood and then it correlates that between everything else and all that sort of stuff like that seems like not a far stretch whatsoever really yeah that's literally what we're uh working out with uh samsung so yeah you know could a watch or ring you know not a lot of the rings have a a speaker in them yeah but if they did you could totally do that right i mean even a ring with the speaker could help with sleep apnea right if you think about it yes not everyone's gonna wear their watch to sleep the ring is probably less less uh invasive annoying right if you're trying to sleep with the ring so yeah it's honestly it's it's it's absolutely fascinating um i just want to finish by talking about your role so overseeing clinical direction the chief chief medical officer can mean lots of different things in lots of different companies what does it what does it mean for you what are you actually doing here what are you overseeing and what are you uh what is that what What does a typical day or week or month look like for you?

56:07You know, again, a lot of it, a lot of, a lot of it on research, right? Trying to find partners that will work with us on different conditions. There's a lot of, a lot of the UI, right? So again, I'm a physician, so I would probably see the product being used differently in terms of how is it going to be viewed in the HR? How, how will it, you know, merge with the different companies we're collaborating with, trying to give more clinical direction there, again, so that when it comes out to our physicians, they're not having to click five buttons or something like that, right? Trying to make it as unburdening to the clinicians who are going to use it as possible.

56:47Of course, working with health systems that I have relationship with, you know, seeing what else do they want to see in this product? You know, how else can we make it better? Lots of meetings, of course, day-to-day and going to conferences and just kind of seeking out what other people are doing and what people think about it i think that's some of the work i'm doing um a lot of it is also just brainstorming on what other conditions can we do and now kind of leading us more towards the longitudinal part right okay we can screen for a disease but can we monitor that disease as the patient lives with it so that's actually one of the key things i wanted to focus on for 2026.

57:26amazing um i normally end this kang by asking uh what's one thing you're excited about i know you're you've told me everything that you're excited about here i think frankly the whole thing's exciting um so i'm going to switch this up slightly i know you're a foodie what's the what's what's your favorite restaurant and why my favorite restaurant like in the world yes in the entire world if you've got one in london i'll specifically take that one but yeah let's go for the world i mean in london i like restaurant uh gordon ramsay and oh interesting okay yeah and heston those are probably my two uh favorite places in london very nice in the world god that rolls a hard question there uh maybe a dish futar in spain and there's actually a restaurant in new york called uh cesar those are probably two of my favorite favorite restaurants of all time appreciate you can i'm gonna check those out when uh where in spain is that was that one oh uh district are is in uh barcelona oh perfect what a not too far from you beautiful city yeah that's a good thing about europe on this side of the world not not and nothing's that far away it's uh it's quite pleasant to up on a flight flight to places but um kang listen absolute pleasure man like this has been awesome um i think giving people an education from both sides of the table has just been fantastic i love all your your principles around adoption and i love the fact you're not able to do that on the other side of the table i think it sounds absolutely fascinating i think voice biomarker stuff is it's clearly a frontier of medicine right now which is like it makes it one of the most exciting places to be in the sense that all the research and work that you are doing we are learning new things it's so nice to be in one of those areas of medicine because there's that there's the microbi the gut microbiome the breast microbiome there's these different places where we're just learning so much new information and yet it's still in medicine that you and i practiced and learned all those years ago but yeah it's it's fascinating to apply the technology the way that you guys have and i really appreciate you coming on if people want to learn more from you um they want to get in touch or anything along those lines what is the best way for them to do so uh simplest way they could just reach out to me my my email is kang at canary speech.com and our website is canary speech.com and they can you know interact with the website or totally feel free to just email me beautiful enjoy man see you later yeah thanks so much hey everyone thanks for listening and making it all the way to the end of this episode remember to subscribe rate us and leave a review and you can head to the description of this episode to follow me on all of my social media so you don't miss out on any of the latest health tech content Thank you.