Latest / Becker’s Healthcare Podcast / Reimagining Neurology Care with AI and Precision Medicine with Kemar Green, DO

Transcript

Reimagining Neurology Care with AI and Precision Medicine with Kemar Green, DO · Becker’s Healthcare Podcast

• 0:00
  This is Scott Becker with the Beckers HealthCare
• 0:03
  podcast. I'm thrilled today to be joined by
• 0:05
  a brilliant physician entrepreneur.
• 0:08
  We're joined today by doctor Kumar Green, and
• 0:11
  doctor Green is also a Johns Hopkins engineering
• 0:15
  graduate as well as being a a a
• 0:17
  physician,
• 0:18
  a brilliant career. He is now the cofounder,
• 0:22
  chief executive officer of a chemical Hazy Eyes.
• 0:25
  We're gonna discuss that. He also went through
• 0:27
  the Techstars program, so or been a mentor
• 0:29
  of the Techstars programs, I should say.
• 0:31
  Doctor Green, tell us a bit about yourself
• 0:33
  and your background and and and about, what
• 0:36
  led you to
• 0:37
  engineering after medical school and and and talk
• 0:40
  a bit about your career and and what
• 0:41
  you're doing today.
• 0:43
  Thanks for having me. It's a it's a
• 0:45
  pleasure to be here.
• 0:46
  So so I'm Kumar. I'm a I'm a
• 0:48
  neurologist,
• 0:49
  with some specialty expertise
• 0:52
  in neurotology
• 0:53
  and neurophthalmology,
• 0:55
  which started all at Johns Hopkins after neurology
• 0:58
  residency. Then I I stayed there as a
• 1:00
  neurology faculty member
• 1:02
  in,
• 1:03
  at Hopkins in the neurology department. And while
• 1:06
  I was there, I did the engineering,
• 1:08
  clinical AI degree.
• 1:10
  My research while I was there sort of
• 1:12
  focused on that, engineering degree where I was
• 1:15
  a clinical AI researcher.
• 1:17
  And, you know, last year or last August,
• 1:20
  actually,
• 1:21
  I made the decision to leave academia
• 1:23
  and build,
• 1:24
  you know, because I I believe,
• 1:27
  the future of medicine
• 1:29
  cannot afford to wait while AI just speeds
• 1:32
  away.
• 1:33
  And so, you know,
• 1:34
  as you know, AI is evolving at an
• 1:36
  extraordinarily
• 1:37
  fast pace. Even yesterday, you know, METSA just
• 1:39
  released their new AI model that's rivaling,
• 1:43
  top LLMs.
• 1:44
  And so, you know, back in August, I
• 1:46
  felt,
• 1:47
  that some of the most important ideas in
• 1:49
  neuroscience,
• 1:50
  like
• 1:51
  digital biomarkers
• 1:52
  or intelligent care delivery system
• 1:55
  risk losing momentum if they stayed trapped
• 1:58
  in slower innovation pathways that sort of live
• 2:00
  in academia.
• 2:02
  And so, you know, what we're building,
• 2:04
  is not just another health care AI tool.
• 2:07
  We're trying to build a foundation for new
• 2:10
  models
• 2:11
  for mainly neurologic care, but eventually this could
• 2:14
  be translated across other sub areas of, of
• 2:17
  medicine.
• 2:18
  And our approach is software first. And, you
• 2:20
  know, since people have different devices, they love
• 2:23
  their devices, whether they're consumer device or clinic
• 2:25
  grade device,
• 2:26
  we are positioning ourselves to be device agnostic.
• 2:30
  So we're building a software first,
• 2:32
  device agnostic system
• 2:34
  that turns any human signal,
• 2:37
  such as your movement, your speech, your behavior,
• 2:40
  any sort of neurophysiologic
• 2:42
  pattern
• 2:43
  into, you know, useful clinical insights that we
• 2:46
  could use to make decisions,
• 2:48
  as clinician.
• 2:50
  The long term vision,
• 2:51
  though, is,
• 2:53
  an autonomous remote neurologic clinic,
• 2:56
  and that kind of, you know, dual
• 2:58
  faces as a copilot or in in the
• 3:01
  in the parlance of today's AI experts, super
• 3:04
  you know, clinician superintelligence
• 3:06
  that could expand the care, which could could
• 3:09
  include something as simple as just triaging
• 3:12
  a patient, screening a patient,
• 3:15
  or or monitoring someone. And eventually, that would
• 3:18
  lead to,
• 3:19
  at least in my mind,
• 3:21
  you know, more precision guided interventions.
• 3:25
  It's so fantastic. And talk about some of
• 3:27
  the trends in health care that you're watching
• 3:29
  for the artificial intelligence
• 3:31
  versus the physician shortage.
• 3:33
  You're you're you're a neurologist.
• 3:35
  We're in an area
• 3:36
  when rising booming amounts of Alzheimer's and dementia
• 3:41
  and other neurology
• 3:43
  neurology treated
• 3:45
  or or discovered illnesses are are booming. How
• 3:48
  do you look at some of those shortages
• 3:50
  of neurologists versus dementia and Alzheimer's? What are
• 3:53
  the trends you're asking in terms of AI
• 3:54
  and where it might have impact on some
• 3:55
  of these things?
• 3:57
  Yeah. So that's a that's a that's a
• 3:58
  great question. So Mhmm. I'm watching a lot
• 4:00
  of things because I'm kind of all over
• 4:01
  the place. But, you know, I think we're
• 4:03
  at internal with the most important transition in
• 4:05
  modern health care. You know? The movement from
• 4:07
  reactive,
• 4:08
  episodic, and one size fits all sort of
• 4:11
  care.
• 4:12
  You mentioned, you know, dementia care. It's gonna
• 4:14
  be different for everyone in different,
• 4:16
  in in different areas of the world, with
• 4:18
  different demographics,
• 4:19
  with different backgrounds. So care has to be
• 4:22
  personalized. And so we're shifting towards that sort
• 4:25
  of era. And, you know, with with all
• 4:27
  these tools that we have at home, even
• 4:29
  our our smart devices,
• 4:31
  we're able to provide continuous and precise care.
• 4:34
  And so, you know, as I as I
• 4:35
  mentioned, like, before, AI is forcing us to
• 4:37
  accelerate the way we kinda, like, make that
• 4:39
  transition. And so
• 4:41
  one of the the biggest trend I'm watching
• 4:42
  right now is actually, you know, that shift
• 4:44
  from,
• 4:46
  at least in in the medical AI space
• 4:47
  from it being,
• 4:49
  more generalized.
• 4:51
  You know, you're baking this tool to to
• 4:53
  triage notes to more domain specific,
• 4:57
  AIs in in health care that are have
• 4:59
  higher accountability.
• 5:01
  And, you know, I think the real future,
• 5:04
  in health care at least is not just
• 5:06
  the ability to to generate
• 5:10
  notes from large language models. You know, it's
• 5:13
  the AI that can interpret
• 5:14
  real physiologic change and behavioral change and imaging
• 5:18
  and longitudinal
• 5:19
  data and put that all together to support
• 5:21
  meaningful care. I mean, in neurology, especially,
• 5:24
  at least in, you know, Hawaii before I
• 5:26
  left neurology, one of the biggest issues we
• 5:28
  had was sort of reaching patients with rare
• 5:29
  diseases,
• 5:30
  because we didn't have enough data to be
• 5:33
  able to do that. And so, you know,
• 5:34
  the ability to, like, get these patients from
• 5:37
  wherever they are, by using the devices we
• 5:39
  have to track these things that are essentially
• 5:41
  the neuro exam, you know, will allow us
• 5:43
  to, you know, make more
• 5:46
  measurable,
• 5:47
  impact that are scalable and actionable in in
• 5:49
  neurology.
• 5:50
  And so that's one of the main thing.
• 5:52
  The other issue the other thing that I'm
• 5:53
  kinda watching is synthetic data. You know you
• 5:57
  know, we don't have data, and it's not
• 5:59
  just in medicine. It's kinda all over. There's
• 6:01
  a finite amount of data available to use
• 6:04
  for AI training.
• 6:05
  And, you know, that's kinda like where everything
• 6:07
  is headed now. And so one of the
• 6:09
  most important trends that I I I see
• 6:10
  coming up, not just in in health care,
• 6:12
  but also sort of all over the
• 6:14
  ecosystem is the rise in synthetic data. And
• 6:17
  I believe it is becoming a very strategically
• 6:20
  important tool for model development, obviously,
• 6:23
  but also for privacy and security and trust.
• 6:26
  You know, if, you know, if if we
• 6:28
  use synthetic data responsibly, in my mind, I
• 6:30
  think synthetic data can reduce dependence on directly
• 6:34
  identifying patient information. So this genetic data, for
• 6:37
  example, won't be vulnerable for leakage if we're
• 6:39
  making synthetic genetic information, the trainer models. We
• 6:43
  could create safer environment for training and testing
• 6:45
  and doing research in AI, which has been
• 6:47
  a big barrier in the health care space
• 6:49
  because the community can't share data very well
• 6:51
  because of security issue. And, of course, could
• 6:53
  support collaboration across institutions because as you know,
• 6:56
  you know, building your,
• 6:58
  siloed AI in your lab is not gonna
• 7:00
  translate across,
• 7:02
  the world to help patients and become scalable
• 7:04
  if that's what your intent was. And so
• 7:07
  I think,
• 7:09
  synthetic data is gonna help us bridge sort
• 7:11
  of that gap, and it's becoming a very
• 7:13
  important security layer as well, in my opinion.
• 7:16
  And then, obviously, the buzzword of 2026
• 7:18
  or maybe it was 2025 really, you know,
• 7:21
  is agents. The agents have arrived,
• 7:23
  but nor are they secure. So I've been
• 7:25
  watching very closely,
• 7:27
  I just how people have been approaching
• 7:29
  agent security.
• 7:31
  And, you know, AI systems are no longer
• 7:33
  just passive tools that are
• 7:35
  are are act you know, that are they're
• 7:36
  just making silo of a decision. They're active
• 7:38
  agents
• 7:39
  that retrieve information,
• 7:41
  reason across various environments. And, you know, since
• 7:44
  I'm kinda neurology biased,
• 7:46
  you know, from, like, sensing to clinical notes
• 7:49
  to making clinical decisions,
• 7:51
  they participate in clinical workflow. I'm sure you
• 7:53
  you know, you probably heard about a bridge.
• 7:55
  And, you know, security is a big concern.
• 7:58
  It's no longer a static problem.
• 8:00
  And so,
• 8:01
  it has become,
• 8:03
  an issue now,
• 8:05
  where we have to think about AI as
• 8:07
  being a contextual
• 8:09
  and adaptive and continuous system when when agents
• 8:11
  are involved. And so we will need systems,
• 8:14
  security systems that is,
• 8:16
  that can validate the identity,
• 8:18
  permissions,
• 8:19
  provenance, and intent,
• 8:21
  and risk of an agent since they're operating
• 8:24
  continuously in real time to ensure that we,
• 8:26
  you know, provide the most optimal care to
• 8:29
  patients in health care, which is, I think,
• 8:31
  will be the foundational
• 8:34
  infrastructures
• 8:35
  moving forward to build in, the best AI
• 8:38
  systems.
• 8:39
  I mean, I could go on and on.
• 8:40
  I mean, finally, obviously, you know, as you
• 8:42
  mentioned before that we need to reach patients.
• 8:44
  And so, you know, AI you know, I
• 8:46
  I see which is redesigning
• 8:48
  how human health is measured and how data
• 8:51
  moves,
• 8:52
  and how we sort of,
• 8:54
  like, embed intelligent systems into care with, you
• 8:57
  know, a very well thought out security layer.
• 9:01
  We could personalize medicine and we could reach
• 9:03
  patients wherever they are with whatever conditions they
• 9:05
  have and provide the expertise
• 9:07
  in subspecialty
• 9:09
  or even general,
• 9:11
  neurology or other specialty care to patients anywhere
• 9:14
  in the world or the universe.
• 9:17
  Thank you very, very much. And and talk
• 9:18
  for a second, doctor Green, about the next
• 9:20
  six months. What are you most focused on
• 9:22
  and excited about?
• 9:24
  Yeah. A lot of things. Let me see.
• 9:26
  What could I maybe
• 9:27
  well,
• 9:29
  you know, since I'm building your agent AI,
• 9:31
  and so, like, right now, I'm sort of
• 9:33
  focused on building
• 9:34
  what what I'm calling
• 9:36
  the measurement and intelligence layer for,
• 9:40
  the future of, you know, neurologic care in
• 9:42
  the AI space. But more broadly, the future
• 9:45
  of
• 9:46
  neuro
• 9:47
  I guess you would call it precision neurology.
• 9:49
  Right?
• 9:50
  I I don't think I'd I coined that
• 9:52
  term. But I'm excited about the idea that
• 9:54
  medicine can move beyond just occasional observations
• 9:57
  and,
• 9:58
  you know,
• 10:00
  a lot of speculation in neurology towards more
• 10:02
  continuous intelligent
• 10:03
  understanding of the patient. And we could have
• 10:06
  a patient understand
• 10:08
  in a very different level beyond what we
• 10:10
  were able to do today.
• 10:11
  And not just what a person reports in
• 10:13
  a visit,
• 10:14
  but also their physiology,
• 10:16
  their behavior,
• 10:17
  and the neurology function
• 10:19
  that will reveal itself over time for situations
• 10:22
  where we just don't know on the first
• 10:24
  visit.
• 10:25
  I know. So that's very exciting to me.
• 10:27
  To me you know, one reason why it
• 10:29
  excites me, one reason, is because
• 10:33
  it's the foundation for something so much bigger
• 10:36
  than just, you know, patient diagnostics in my
• 10:39
  mind at least. It creates,
• 10:41
  and this is what I was trying to
• 10:42
  build before a top gain is, the foundation
• 10:45
  for precision neurology,
• 10:47
  personalized medicine, which we're not gonna be able
• 10:49
  to accomplish until we're able to solve all
• 10:50
  those issues, and something new,
• 10:53
  digital neurotherapy where software now becomes
• 10:56
  a treatment tool.
• 10:58
  And, and, you know, with with that, with
• 11:01
  personalized medicine, precision medicine, you could personalize treatment,
• 11:05
  to that individual, which I think is is
• 11:07
  is really, really big. And that's kinda what
• 11:09
  I'm focused on over the next six months.
• 11:11
  Let me ask you one last question. It
• 11:14
  it's is this doctor Green? You've had this
• 11:16
  great career at sort of this sort of
• 11:18
  intersection of technology, engineering, and health care. What
• 11:21
  advice would you have to people, emerging physicians,
• 11:25
  emerging leaders,
• 11:26
  trying to have impactful careers that they'll enjoy
• 11:28
  and thrive with? Any advice that you would
• 11:30
  provide?
• 11:32
  Yeah. I mean I mean, I'm still I'm
• 11:34
  still new to the whole entrepreneur thing, but
• 11:36
  one thing I could say is, you know,
• 11:38
  my biggest advice of this,
• 11:39
  build for the future.
• 11:41
  You know, don't think about what you need
• 11:42
  to build for tomorrow. Think about the new
• 11:45
  systems that are already being implemented fifty years
• 11:48
  from now. I mean, let me use space
• 11:49
  as, like,
• 11:51
  your ceiling, if you will. You know? And,
• 11:53
  you know, but, you know, at the same
• 11:54
  time, even though we're building for the future,
• 11:55
  we're having these big, ambitious ideas, stay anchored
• 11:58
  in reality because, you know,
• 12:00
  reality is is right now. And so, you
• 12:02
  know, in moments
• 12:04
  of, major technological change like right now, and
• 12:06
  a lot of people either become too conservative
• 12:09
  or too intoxicated by the possibility,
• 12:12
  you know, I think
• 12:14
  the future,
• 12:15
  leaders in health care will be neither,
• 12:17
  you know, if they're gonna be great leaders
• 12:20
  at it, in my opinion.
• 12:21
  They have to have the courage to see
• 12:23
  farther than, you know, the others, even if,
• 12:26
  you know, their mentors don't see as far
• 12:28
  as they are seeing in that moment. But,
• 12:30
  again, you know, they still should remain disciplined
• 12:32
  enough
• 12:33
  to build what is true, what is useful,
• 12:36
  and what can easily earn trust,
• 12:39
  in the real world. And in the medical
• 12:41
  world, trust is very hard to to to
• 12:43
  earn. And I I think the most important
• 12:45
  thing is, you know, we're building this system,
• 12:47
  and we're we're not we you know, like,
• 12:49
  for example, I left to build them very
• 12:51
  fast,
• 12:52
  but we still have to,
• 12:55
  not sacrifice,
• 12:57
  scientific rigor for speed, especially in health care.
• 13:00
  You know, that's not a false choice by
• 13:02
  any means. You know, moving too slowly, as
• 13:04
  you know, in this very fast AI
• 13:07
  AI pace
• 13:08
  can deny patients of benefits just
• 13:11
  hamper,
• 13:13
  momentum.
• 13:13
  But moving to carelessly,
• 13:15
  it can harm and destroy trust,
• 13:18
  for our patients,
• 13:20
  for the people who are doing this to
• 13:21
  serve. And so, you know, as future health
• 13:23
  care leaders in the space,
• 13:25
  you know, I think prioritizing
• 13:28
  and kind of balancing speed and rigor
• 13:30
  is is very, very important. And we can
• 13:33
  build systems. Another
• 13:35
  advice, pardon my last one. We can we
• 13:38
  can build systems in silo, especially AI systems.
• 13:41
  We we need teams to come together to
• 13:43
  be able to do this, you know, securely
• 13:46
  with the appropriate data governance.
• 13:49
  We We need the patients involved as well.
• 13:50
  It shouldn't just be billed by clinicians and
• 13:52
  researchers in a lab somewhere. We need patients
• 13:54
  to be at the center of all this
• 13:55
  because if we're redesigning a system that's gonna
• 13:58
  help patients,
• 13:59
  they need to be involved in the design
• 14:02
  from day one.
• 14:04
  Doctor Green, what a pleasure to visit with
• 14:06
  you. What an amazing background. You might be
• 14:08
  the only doctor engineer that I know that
• 14:11
  was also a mix of a biology and
• 14:12
  French student as a college graduate.
• 14:15
  An amazing background. Doctor Kumar Green, what a
• 14:18
  pleasure to visit with you. Thank you so
• 14:19
  much for taking the time with us on
• 14:20
  the Veterans Health Care podcast.
• 14:22
  Well, thank you so much. It was great
• 14:23
  being here.