Latest / The SOM Podcast / SOM MINI CPD: AI and Audiometry in Occupational Health with Trevor Chapman

Transcript

SOM MINI CPD: AI and Audiometry in Occupational Health with Trevor Chapman · The SOM Podcast

• 0:00
  Speaker 1: This podcast is sponsored by Orchid Live. Hello and welcome
• 0:18
  to the SOM podcast. I'm Steve Randall, and this episode
• 0:21
  is one of a monthly series to help you as
• 0:24
  an occupational health professional with your CPD, as we invite
• 0:28
  specialists to share an overview of some of the topics
• 0:32
  you've told us you'd like to find out more about.
• 0:34
  Speaker 2: So my name is Trevor Chapman. I'm an audiologist and
• 0:38
  founder of a company called Global Hearing Solutions.
• 0:41
  Speaker 1: Well, Trevor, welcome to the SOM podcast. It seems that
• 0:43
  pretty much everything we talk about these days in our
• 0:46
  lives and in our work has some kind of AI connection.
• 0:50
  And we're going to be exploring that a little bit
• 0:52
  in this mini CPD episode, and particularly how AI AI
• 0:57
  and hearing tests kind of work together and how it
• 1:00
  can be done in a compliant way. So where are
• 1:03
  we at the moment on that?
• 1:04
  Speaker 2: Thanks, Steve. Thank you for the opportunity to join the podcast.
• 1:07
  AI in healthcare is growing at a rapid pace. and
• 1:12
  we're now starting to see it being used in hearing screening.
• 1:16
  So we're using a machine learning algorithm that was developed
• 1:20
  and validated in academia, and we've now put it to
• 1:22
  good use into a hearing screening solution. What it allows
• 1:26
  us to do is to test hearing in a fraction
• 1:30
  of the time And as accurately as what we could
• 1:34
  achieve in a fully purposed audiology clinic. So it's quicker
• 1:38
  and extremely accurate.
• 1:40
  Speaker 1: And the benefits presumably are economy of scale for a start,
• 1:43
  being able to do things much quicker means that you can.
• 1:47
  see more patients do more tests in the same amount
• 1:50
  of time, for example.
• 1:52
  Speaker 2: Yeah, that's correct. One of the other benefits that we see,
• 1:55
  and this has come out of some of the work
• 1:57
  we've been doing with some of our defence sector partners,
• 2:00
  are things like consistency and accuracy. So having a hearing
• 2:05
  screening test being done in the south of the country
• 2:09
  or in the north of the country, regardless of nurse, technician, audiologist, audiometrist, doctor,
• 2:16
  et cetera, et cetera, having a test that is performed
• 2:19
  the same way and returning results in a consistent and
• 2:24
  accurate manner. So that means that you're not having any inconsistencies.
• 2:28
  You're not having people who are trained in a certain
• 2:31
  way doing it differently to another person who is trained
• 2:33
  in a slightly different way. I'm not saying those results
• 2:36
  are inaccurate or not correct. What I'm saying is having
• 2:40
  a solution that will do it consistently and accurately each
• 2:45
  time makes the reliability so much better. We have a
• 2:50
  military standard hard case, which incorporates audiometric headphones. So the
• 2:55
  same headphones you would find in audiometry clinic. It has
• 2:59
  a tablet built in and we also use a video otoscope,
• 3:02
  which allows our patients or our test takers to to
• 3:05
  take pictures inside of the ear canal so we can
• 3:08
  check to see whether there's any wax accumulation or any
• 3:11
  foreign bodies or indeed any outer ear pathology. I often
• 3:16
  joke with the defense sector saying our kid can survive
• 3:19
  everything unless you run it over in a tank. Wow.
• 3:22
  Speaker 1: I mean, it's impressive stuff. There are so many questions, though,
• 3:26
  with anything relating to AI. First of all, and let's
• 3:29
  address the elephant in the room, if this is a
• 3:32
  more efficient technology, does it mean that fewer people are
• 3:36
  needed to carry out the hearing tests that are required
• 3:40
  across everybody who needs those? So is this going to
• 3:42
  have an impact on jobs, for example?
• 3:45
  Speaker 2: So we get that question asked a lot, Steve. And
• 3:48
  what I'd love to say is, whilst the technology is
• 3:51
  able to test somebody's hearing in a quicker period, what
• 3:55
  that means is the technician or the nurse that's conducting
• 3:59
  the appointments or the time on site, what that allows
• 4:03
  them to do is to use that additional time to
• 4:06
  either complete additional questionnaires or perform other health screening systems
• 4:11
  or criteria that the employer or the occupational health provider
• 4:14
  wants to deploy. So we're not here to look at
• 4:17
  removing staff or reducing the numbers of staff. What we're
• 4:21
  looking to do is provide staff the ability to provide
• 4:24
  a test that is accurate and consistent, freeing up their
• 4:27
  time to perform other roles. What artificial intelligence is good
• 4:32
  at doing is doing tests and analyzing data. What it's
• 4:36
  not good at doing is all the stuff that we
• 4:39
  need to do when we're with our clients. So asking
• 4:42
  correct questions, writing down those answers, checking eyes, checking ears,
• 4:48
  checking lung functions, et cetera, et cetera. So there's still
• 4:51
  always going to be a need for a human to
• 4:53
  be involved in the process.
• 4:55
  Speaker 1: And we also know certainly that AI is, doesn't always
• 4:59
  get things right, you know, and human oversight is clearly
• 5:02
  very important in that regard.
• 5:04
  Speaker 2: Yeah, that's absolutely correct. So whilst our AI has been
• 5:07
  validated in academia, we also have a logic that sits
• 5:11
  on top of the AI that will highlight any red
• 5:15
  flags based on the health and safety executive's criteria. So
• 5:19
  things like category threes and category fours, noise-induced hearing losses,
• 5:23
  whether it's unilateral or asymmetric, So we can do instantaneous
• 5:27
  red flag categorization. I say instantaneous, but it's 0.43 seconds.
• 5:33
  So that will allow us to identify to either the
• 5:36
  nurse or technician on site or to the person responsible
• 5:40
  for running the health surveillance program looking at the hearing
• 5:43
  test result. We can show them all the red flags
• 5:46
  and then you'll need a person involved there to determine
• 5:49
  what is going to be the next onward step. Is
• 5:52
  it a referral to occupational health physician? Is it going
• 5:55
  to be need for a retest, et cetera? So there's
• 5:58
  always going to be a human involved in looking over
• 6:02
  the results and looking over the red flags. In a way,
• 6:05
  AI is a catch-all phrase, artificial intelligence. But AI is
• 6:12
  typically made up of three different things. There's artificial intelligence,
• 6:16
  there's machine learning algorithms, and then there's deep neural networks.
• 6:20
  And what our system is doing, it's working on machine
• 6:23
  learning algorithms. So it's very fine art in what it
• 6:27
  can do. It isn't a chatbot. It isn't a deep
• 6:31
  neural network that can make gazillions of calculations per second
• 6:34
  and try and infiltrate another system. What we have is
• 6:37
  a system that is designed to check for hearing in
• 6:41
  a very fine art parameter, which has been validated in
• 6:45
  academia to be shown to be accurate and correct. And
• 6:48
  is this.
• 6:49
  Speaker 1: an evolving thing? Will it be updated with new data sets,
• 6:53
  for example, over time?
• 6:55
  Speaker 2: Absolutely correct. And so one of the great things we
• 6:58
  have is the machine learning algorithm which runs, the more
• 7:02
  we train it, the better it becomes, which is great.
• 7:04
  But what we have on top of that is the
• 7:06
  logic that we use. So for example, one of the
• 7:09
  great tools that we've developed is temporary threshold shift testing.
• 7:14
  And so we call it TTS, a temporary threshold shift.
• 7:17
  And I think every person possibly listening to this or
• 7:20
  indeed walking on the planet would have experienced a temporary
• 7:23
  threshold shift at some point in their life. You can
• 7:26
  probably remember back to when you went to a noisy
• 7:28
  concert or to a noisy party. and you came out
• 7:31
  of the party and you felt like you had cotton
• 7:32
  wool in your ears. Now that is a temporary threshold shift.
• 7:35
  That's your body's natural way of trying to protect itself
• 7:39
  from loud noise. After a few hours, that resolves itself
• 7:44
  and your hearing returns back to normal. So the way
• 7:47
  we can use this in the example of occupational health
• 7:51
  is imagine if you have a noisy piece of equipment
• 7:55
  and you want to check to see whether the headphones
• 7:58
  being worn by a subsection of staff who are using
• 8:02
  that machine, whether they're wearing their PPE correctly. So you
• 8:06
  can do a number of tests before they're exposed to
• 8:08
  noise on the machine, and you can do a number
• 8:11
  of tests after they've been exposed to noise on the machine.
• 8:14
  And what we tend to do is we'd run a
• 8:16
  test every four hours after exposure. And you can track
• 8:20
  the recovery of hearing loss as a result of the
• 8:23
  exposure to the noise. And what that allows us to
• 8:25
  do is to check and see whether the PPE is adequate,
• 8:29
  whether it's rated correctly, whether it's worn correctly, for example.
• 8:32
  But I can give you another example that's used in defense.
• 8:35
  Let's say manufacturer A creates a platform, a helicopter, a
• 8:41
  fast jet, a tank, whatever it would be. And as
• 8:43
  a result of time in the cockpit, the hearing drops
• 8:48
  of the pilot, for example, we can measure that hearing
• 8:51
  drop over a period of time and either go back
• 8:54
  and ask the manufacturer of the helmet to create greater
• 8:58
  noise protection, or we can ask for the rotors to
• 9:01
  be made a bit more quieter, or we can increase
• 9:03
  the insulation in the cockpit. Or alternatively, if hearing drops
• 9:07
  after a two or three hour time in the cockpit seat,
• 9:10
  for example, we can increase radio traffic between the ground
• 9:15
  and the pilots to make sure instructions are carried out correctly.
• 9:18
  So being able to use the technology to develop something
• 9:21
  like temporary threshold shift testing makes audiology or hearing testing
• 9:26
  more proactive not reactive.
• 9:30
  Speaker 1: Is that testing done currently, but in a more long
• 9:34
  and drawn out way, or is it just not possible currently?
• 9:37
  Speaker 2: So currently we don't see this being used as a
• 9:40
  regular occurrence. So let's say, for example, you work in
• 9:44
  a factory and there's been an explosion, a compressor has
• 9:47
  blown up or something else has happened and there's been
• 9:49
  an accident. You'll get your first responders come onto site
• 9:52
  and they'll check for burns, they'll check for puncture wounds,
• 9:55
  they'll check for spills, but nobody's checking for hearing damage.
• 9:59
  So imagine having a kit that you can quickly grab
• 10:01
  off the shelf and run down to the area, take
• 10:04
  the person affected by the incident and check their hearing
• 10:07
  and monitor their hearing over a period of time to
• 10:10
  determine whether or not there's been any acoustic trauma, whether
• 10:13
  there's going to be any potential for permanent hearing loss, etc.
• 10:16
  Speaker 1: etc.
• 10:18
  Speaker 2: Hearing loss is an invisible disability. You can't see if
• 10:21
  somebody has a hearing loss. And so it goes underdiagnosed, underreported, undermonitored,
• 10:27
  underscreened at every step of the way.
• 10:30
  Speaker 1: Could this be in hearing aids? Would it be something
• 10:32
  that could actually be put into that technology so that
• 10:35
  hearing aids become accessible? truly adaptive? I know they are
• 10:38
  to some degree already, but this would power that.
• 10:41
  Speaker 2: So I am a clinical audiologist. So I also own
• 10:45
  an independent audiology clinic. So I'm a practicing audiologist at
• 10:48
  the same time. And we're now using hearing instruments that
• 10:52
  have deep neural networks in them. So they have proper
• 10:55
  AI built into them. They do about 7.7 billion, that's
• 10:59
  with a B, calculations per second. And And they're very
• 11:01
  good at removing background noise. They're very good at understanding
• 11:04
  what is a sudden sharp onset of noise. And yeah,
• 11:08
  we could absolutely see hearing instruments evolving in the future
• 11:11
  to test for things like temporary threshold shift or to
• 11:15
  test for other elements as well. The ear canal is
• 11:19
  a great place to check other things that is going
• 11:22
  on in the body. So, for example, things like heart
• 11:24
  rate and blood pressure. And there are some instruments out
• 11:27
  there at the moment that have accelerometers built into them
• 11:30
  that if you'd suffer from a fall, they would alert
• 11:33
  a loved one, a neighbor or a caregiver. What we
• 11:36
  want to do is to help people age healthily. And
• 11:40
  one of the things that happens when we get older
• 11:43
  is our hearing deteriorates. Now, whether that's as a result
• 11:46
  of age or whether that's a result of noise-induced hearing
• 11:49
  loss or for whatever reason, their hearing loss has a
• 11:52
  negative impact on our healthy aging. It also has a
• 11:56
  negative impact in the workplace. So people with untreated hearing
• 11:59
  loss in the workplace don't perform as well as their
• 12:01
  counterparts that have normal hearing. People with untreated hearing loss
• 12:05
  in the workplace feel like they can't perform as well
• 12:07
  or they don't have the same opportunities because of their
• 12:09
  hearing loss. Treating hearing loss is relatively easy. And for us,
• 12:14
  the biggest issue we face still is stigma. So if
• 12:17
  I had a friend who had hearing loss, and they
• 12:20
  were struggling to hear, I'd be the first one to
• 12:22
  support them in going to get themselves sorted out. But
• 12:25
  that person themselves, whilst they might want to support me
• 12:28
  with my hearing loss, if I had to offer them
• 12:31
  a solution to treat their hearing loss, they would probably go, no,
• 12:34
  not quite for me. I'm not quite old enough yet.
• 12:36
  I don't want to wear hearing aids. And at the moment,
• 12:38
  hearing aids are so advanced. They've got Bluetooth. I wear
• 12:42
  hearing aids. I have hearing loss. So they stream my
• 12:44
  phone calls. They stream my social media. They do everything
• 12:48
  on my phone. They're hands-free and they're pretty cool. And
• 12:52
  they connect me back to my world. They connect me
• 12:54
  back to my friends, my family. I don't need to
• 12:56
  say pardon all the time. And if we make hearing
• 12:59
  screening more relevant in the workplace. So not just for
• 13:03
  occupational health, but also for employee wellness programs. If we
• 13:07
  make hearing screening normal, we're going to be doing the
• 13:10
  same thing that optics did in glasses 40 years ago
• 13:14
  when it was uncool to wear glasses. If we make
• 13:17
  hearing screening normal, we can make treating hearing loss normal,
• 13:21
  and therefore we can make aging healthily with better hearing normal.
• 13:26
  Speaker 1: Yeah, I think that's fantastic. And it's amazing, really, that,
• 13:29
  and you highlighted the fact that people wear glasses and even,
• 13:33
  you know, people choose to wear glasses even when they
• 13:35
  don't need glasses. They'll have clear glass or they'll wear,
• 13:38
  you know, things like these meta AI glasses that are
• 13:40
  coming out and things like that. You know, people choose
• 13:42
  to do that. And yet there is still a stigma
• 13:44
  around hearing loss.
• 13:45
  Speaker 2: It's baffling. Stigma is, it's annoying as a clinician. I
• 13:50
  understand where it comes from, from the hearing impaired person
• 13:54
  that we're looking at, but it's also unnecessary because helping
• 13:58
  yourself hear better for longer means you're going to age
• 14:01
  more healthy because the comorbidities or the other health conditions
• 14:05
  that are impacted by hearing loss are vast. So things
• 14:09
  like social isolation, withdrawing from situations because you struggle to hear.
• 14:15
  That leads to things like depression. People with untreated hearing
• 14:18
  loss are more likely to suffer from a fall. There
• 14:21
  are discussions around increased cognitive decline or onset of dementia
• 14:26
  as a result of untreated hearing loss. So this is
• 14:29
  where my passion as an audiologist and somebody who loves
• 14:32
  the flappy bits on the sides of people's heads is
• 14:35
  utilizing a machine learning algorithm and putting it into a
• 14:39
  piece of technology that we can put into the workplace,
• 14:43
  into schools, wherever we can, that makes hearing screening normal,
• 14:50
  but also having it done in such a way that
• 14:52
  is compliant and It's accurate. It's repeatable. It looks for
• 14:56
  additional elements, not just the hearing loss, but also helps
• 15:00
  to identify referable criteria or referable conditions. And we can
• 15:05
  use it proactively instead of something that is reactive and
• 15:09
  has a stigma attached to it.
• 15:11
  Speaker 1: Great. Thanks very much for being on the SOM podcast.
• 15:14
  Speaker 2: Thank you, Steve. Appreciate the opportunity.
• 15:25
  Speaker 1: And work with both in-house OH teams and OH providers.
• 15:28
  You can find out more at orchidlive.com.