New Matter: Inside the Minds of SLAS Scientists
New Matter: Inside the Minds of SLAS Scientists
Vagus Nerve Stimulation for Language Acquisition with AI | Tracy Centanni, M.S., Ph.D.
The vagus nerve plays a crucial role in the autonomic nervous system as it is responsible for many vital functions like adjusting heart rate, breathing and digestion among many others. Current research on the vagus nerve is looking at how stimulating the nerve can treat auditory and motor impairments and diseases.
Our guest for this episode is University of Florida Associate Professor Tracy Centanni, M.S., Ph.D., to explain how vagus nerve stimulation intersects with screening and AI.
Key Learning Points:
- The inspiration for vagus nerve stimulation as an intervention for language intervention
- Other conditions vagus nerve stimulation could potentially treat
- Screening methods for auditory issues
- The challenges of training AI models for personalized medicine for conditions like dyslexia
Full Transcript Available on Buzzsprout
About Tracy Centanni, M.S., Ph.D.
Centanni is the Director of the Genetics of Auditory and visual Perception and Plasticity (GAPP) Lab in the College of Public Health and Health Professions at the University of Florida. Her research program focuses on the genetic and neural correlates of reading acquisition, factors that drive reading disorders (e.g., dyslexia), and neural plasticity including using non-invasive vagus nerve stimulation to improve reading and language.
She received her Ph.D. in 2013 from the University of Texas at Dallas, using rat models to probe the influence of dyslexia-candidate genes on auditory perception. She then completed postdoctoral training at the MGH Institute of Health Professions and at MIT, with the goal of learning translational and clinically relevant research methods.
Take our Survey
We want to hear from YOU to ensure that we keep providing valuable content that reflects what our listeners and the SLAS community are interested in! Take our brief survey
Stay connected with SLAS
About SLAS
SLAS (Society for Laboratory Automation and Screening) is an international professional society of academic, industry and government life sciences researchers and the developers and providers of laboratory automation technology. The SLAS mission is to bring together researchers in academia, industry and government to advance life sciences discovery and technology via education, knowledge exchange and global community building.
Upcoming SLAS Events:
SLAS 2024 Data Sciences and AI Symposium
- November 12-13, 2024
- Cambridge, MA, USA
SLAS2025 International Conference and Exhibition
- January 25-29, 2025
- San Diego, CA, USA
SLAS Europe 2025 Conference and Exhibition
- 20-22 May 2025
- Hamburg, Germany
Hannah Rosen:
Hello everyone and welcome to New Matter, the SLAS podcast where we interview life science luminaries. I'm your host, Hannah Rosen, and joining me today is Tracy Centanni, associate professor at the University of Florida and also just so happens to be my sister. Tracy is here to talk to us about her research on vagus nerve stimulation and the intersection of AI with healthcare and diagnostics. Welcome to the podcast, Tracy.
Tracy Centanni:
Thanks for having me.
Hannah Rosen:
It's our pleasure, very exciting. So, to start off, can you just provide us with a little bit of your professional background and your expertise?
Tracy Centanni:
Yeah, absolutely. As you mentioned, I'm an associate Professor at the University of Florida. And so, our research focuses largely on the neural and genetic correlates of auditory and visual perception. So. Focusing largely on how the brain processes language and reading, and the plasticity that happens during that process. So, in terms of my background, I have a bachelors degree in psychology from Penn State University, I have a masters and a PhD in cognition and neuroscience from the University of Texas at Dallas, and I was able to complete some postdoctoral work in Boston at Mass General Hospital and MIT, and then started my lab originally at Texas Christian University in Fort Worth, TX, and I was there for about 6 years, and literally just moved to UFC a couple months ago.
Hannah Rosen:
Awesome. Exciting. And I'm excited because, you know, honestly, I'm reflecting and you talking about that, we don't have very many academics on the podcast, so it's gonna be kind of a fun little detour for our listeners, that perspective.
Tracy Centanni:
Gotta represent the academic bench research side.
Hannah Rosen:
That's right. So can you tell us a little bit about, you know, where did you get the idea for using vagus nerve stimulation as an intervention for a language acquisition?
Tracy Centanni:
Absolutely. So I did my PhD as I mentioned in Dallas, TX and the lab where I was doing my PhD work were really pioneers in the use of vagus nerve stimulation and specifically for treatments for diseases like tinnitus, or conditions like tinnitus, as well as post stroke motor impairment. And so that lab was doing a lot of preclinical work in animal models, and I got to watch that preclinical work be translated into human populations, so I was there when they did the first clinical study in 10 patients with tinnitus. And so I've really got to understand how it works and some various impacts of vagus nerve stimulation on neuroplasticity and changes in the brain. And so, at the same time I was working on speech coding, so trying to understand how the brain encodes the sounds of language, how does your brain tell apart AD from AB, and watching this vagus nerve stimulation happening in a nearby team. And so as I started to get interested in conditions like dyslexia and autism and language impairment. Of course, the light bulb starts to go off well, wonder if this technology could be useful for these other kinds of populations.
Hannah Rosen:
Wow. So can you go onto a little bit, because, you know, I, and I'm sure a lot of our listeners are probably familiar with, they've heard of vagus nerve simulation, they know the vagus nerve in the body, but, you know, as you're talking, I'm sitting here thinking, OK, so how does this work? How does vagus nerve simulation influence all these different conditions? And, you know, in your research do you know how vagus nerve stimulation helps to improve language acquisition?
Tracy Centanni:
So yes and no. It's a really great question. So, the vagus nerve, of course in general terms is involved in fight or flight. So, when you are that the class example I use is, pretend that you're a gazelle on the Savannah having your gazelle lunch and a lion jumps out at you from behind a tree. So, fight or flight kicks in, which of course is regulated by the vagus nerve, allowing you to escape that threat and survive. But what also happens as part of that fight or flight response is a lifelong emotional memory of that experience. So of course, for your survival, it's very important that you know that this tree could be hiding a lion, you know, this could potentially be a dangerous place, and it only takes one exposure for that emotional memory to be formed. And so, we know from preclinical work in animal models that vagus nerve stimulation, if we put an electrode around the vagus, we are driving the release of certain neurotransmitters, these chemical signals in the brain that do lots of different things, but in this case three key ones are norepinephrine, acetylcholine, and serotonin. And so when you release these neurotransmitters at the same time as some really scary event, or in our case, laboratory based training programs, what we think we're doing is strengthening synapses, so, connections between different brain regions. There's still a lot that we don't know about exactly how it works and under what conditions it works, and that’s some of the research that we're doing in the lab as well as trying to understand under what conditions it works and what is actually happening in the brain during that training.
Hannah Rosen:
Hmm, that's interesting and I have so many follow up questions, but it's just occurring to me, you know, you're talking about this and, you know, anybody who has any basic knowledge of the function of nerves knows that they're one way, right? So if the vagus nerve is coming out of the brain to the rest of the body, how then does vagus nerve stimulation affect the brain itself? Because obviously, you know, the signal's not going up the nerve into the brain. So, do you know how these impacts in the brain are happening?
Tracy Centanni:
Yeah, that's a fantastic question. So the vagus nerve is actually a bit unique in this way. The vagus nerve actually has both afferent and efferent fibers, so there are pathways going both towards the brain and from the brain towards the body. So, we won't necessarily get into a dissection talk on the vagus nerve here, but when we think about classic vagus nerve stimulation, which is invasive, there are actually two branches of the vagus nerve in the neck behind the trachea, basically right next to the carotid artery. And the left branch of the vagus is primarily afferent, so signals coming from the body towards the brain, and that's the branch that we stimulate for vagus nerve stimulation. And then the right branch of the vagus is mostly efferent, so, signals coming from the brain going to the body, helping to control a lot of that fight or flight digestion, all kinds of body systems that are connected to the brain.
Hannah Rosen:
Got it. OK, that makes a lot of sense. So, then I'm wondering then, you know, cause you discussed that it's so tied into fight or flight and these emotional memories. So, is it then when you're doing vagus nerve simulation, are you then tying, so for example, if you're doing language acquisition and you're trying to, you know, maybe help somebody learn a new language through this vagus nerve simulation, is it, do you think it's working because it is tying this factual information to an emotion? Because I've heard that, you know, if you tie different senses together or like an emotion to a fact, you'll remember it better. And so, is that kind of what's happening here?
Tracy Centanni:
Yeah, so not in this particular case, because the vagus nerve has that bidirectional anatomy. What we can really do is isolate the afferent branch and release those neurotransmitters without invoking any kinds of emotional components. So, even though the vagus nerve does have synapses into the limbic system, the emotional processing centers. We are not specifically activating those brain regions unless we choose to provide an emotional stimulus at the same time. So, for example, there are some labs working with models of PTSD trying to use vagus nerve stimulation to help improve that, and in those cases they're providing emotional stimuli by definition. But the vagus nerve stimulation itself doesn't create an emotional response on its own.
Hannah Rosen:
Very interesting. So, you know, you had mentioned that traditional agus nerve stimulation is very invasive. So, can you kind of discuss the differences in invasive versus noninvasive vagus nerve stimulation?
Tracy Centanni:
Definitely. So again, vagus nerve stimulation began with an invasive procedure. So, the cervical branch of the vagus runs through the neck, and in animal models, of course, this is a a small surgery, in human patients, it's an outpatient surgery, it's about $30,000. It's a relatively simple, it's a little cuff electrode that gets implanted around, again, the left branch of the vagus, and then there's a subdermal battery and Bluetooth patch that usually gets implanted in the chest, and that allows the clinician to send information or commands to the stimulator to control, you know, all the different parameters for the stimulation. Of course, when you're talking about conditions like epilepsy, depression, tinnitus, stroke, all the conditions that it's currently FDA approved for, those kinds of conditions are severe enough to really justify a surgery. However, when we're talking about things like language learning and reading learning in our lab, we study dyslexia quite a bit. I still have not met a parent that is going to sign off on a $30,000 outpatient procedure to implant a couple electrodes in their kid. So, we had been searching for a non-invasive way of getting the same bang for our buck, again without the need for surgery. And so luckily for us, there is a branch of the vagus that runs through the outer ear called the auricular branch. And so, we're able to stimulate the vagus nerve non-invasively through the ear, just using a little surface electrode. And there's been a number of research studies now showing that we can drive activation in the same brain regions as the invasive form, but again without the need for surgery.
Hannah Rosen:
Wow, could this be possible to use as a treatment for some of these other conditions as well that have previously required the invasive surgery, or is it too soon to say?
Tracy Centanni:
No, actually the very first proof of concept of the non-invasive form was a replication of the post stroke motor treatment. So basically, patients who have had a stroke, they're experiencing difficulties relearning some of the fine motor movements, things like tying their shoe or turning the knob on the door. You can get the same improvement with non-invasive vagus nerve stimulation as you can with invasive.
Hannah Rosen:
Wow, so do you think that there might be a future where the non-invasive method might ultimately replace this invasive method for other conditions, aside from dyslexia where it may be the non-invasive is a requirement, but do you think that it could replace the invasive methods for some of these other conditions like stroke or epilepsy?
Tracy Centanni:
I think so, but I think there's certainly a limit on that. I know that there are some groups providing this technology, the non-invasive form, for epilepsy and depression. I believe those trials are going on if they're not already out. Certainly for other conditions as well, we have a colleague and collaborator doing work with opioid withdrawal using a non-invasive stimulator. There are certainly some conditions though that I think will probably always require the cervical stimulation, and those are conditions that involve body systems, so things like irritable bowel, where you need to be stimulating that efferent branch going from the brain to the body. But if you're trying to stimulate up towards the brain, I think the non-invasive form should be able to take over a lot of that space.
Hannah Rosen:
Interesting. Are there any conditions that you can think of that vagus nerve stimulation isn't currently being used for, but that you can see the potential where maybe some researchers should think about or consider pursuing vagus nerve stimulation as a potential treatment?
Tracy Centanni:
Yeah, I think that it really, vagus nerve stimulation has exploded in the last several years in terms of researchers trying to tackle it for various conditions. So, there are clinical trials going on, I mentioned IBS, there are clinical trials for Afib, atrial fibrillation. There are clinical trials, as I mentioned, for PTSD and some other conditions. I have a colleague at UFC doing some work with Alzheimer's disease. The vagus nerve has a big role with inflammation and so, you know, any condition where you have, for example, chronic arthritis potentially. I think the sky is the limit really, because the vagus nerve impacts so many body systems. Certainly, there are going to be limitations, but I would say anybody that's looking for alternative treatments for any kind of condition, you know, at least take a look at vagus nerve stimulation and see whether it's applicable. But yeah, I think that it's going to be present in a lot of different subspecialties moving forward.
Hannah Rosen:
Wow. So, can you tell us a little bit about, you know, in your work when you are using this non-invasive vagus nerve stimulation for interventions with language acquisition, can you just tell us a little bit about what you've been doing in that area? I understand you are working both with patients with dyslexia, this is an intervention for that, and then also just potentially people who want to learn a second language or even a third language.
Tracy Centanni:
Yeah. So, we have been doing a few studies. So, two of our prior studies are currently published. So we have one study looking at letter sound learning, so learning to read in a new alphabet. And we are currently moving that work into dyslexia trials, so we can't say for sure yet whether or not it's going to work well, but we are testing that right now. And then we've also published our work looking at reading comprehension. So, we've all had this experience where we're reading a textbook or an article, and you read a paragraph and you know you've read every single word, but you get to the end and you have no idea what you just read. And so, what we found is that vagus nerve stimulation improves memory for read passages. So that's pretty interesting. And then we are working on publishing our language learning work. So, we've got young adults learning a novel language, which of course we know is much harder to do in adulthood. In this country, we teach foreign languages in high school, which is just way too late for the brain to really grasp that language and learn it fluently. And so, we found that after a single session of training, we can improve retention of novel words up to a week later. We haven't tested it beyond a week, but we can at least show benefit a week later. So, we're now moving into some other areas of interest, you know, language, other reading, other populations and also adding in some neural imaging. So, looking to see what's happening in the brain during that training.
Hannah Rosen:
Yeah, wow, that would be really interesting to see what those images look like. So, when you're doing these studies, you know, and you're taking individuals and you're using the vagus nerve stimulation, are you seeing that there are any individuals that this therapy just doesn't seem to work? You know, maybe have some people who are improving their memory and then, are there some people where it's just, you don't see any effect at all?
Tracy Centanni:
Yeah, there's no silver bullet, unfortunately. And so there certainly are some individuals that respond better than others and we do have a a line of research in the lab to probe those differences and figure out what is causing the differences. Are there, for example, some genetic markers that block efficacy? Aad we have one line of research looking at that. Are there medications on board? So, for example, if we're targeting these neurotransmitters, these chemical signals in the brain, lots of medications influence those same signals, and so is it the case that a participant who is taking Adderall, for example, which impacts norepinephrine, does that block vagus nerve stimulation or not? So we are currently doing a lot of that work behind the scenes to hopefully create a set of guidelines for future researchers, but also for patients and clinicians to think about when they're when they're considering vagus nerve stimulation as a course of treatment.
Hannah Rosen:
Are there any other challenges with vagus nerve stimulation that might prevent this from becoming more mainstream?
Tracy Centanni:
I think really the only other challenges would be other contraindications. So, you know, anything, for example, like a pacemaker or existing heart conditions. The risk for cardiac side effects is incredibly low. However, because as we talked about before, vagus nerve is fight or flight, so you are impacting heart rate, you're impacting heart rate variability depending on the type of stimulation you do. And so, it's certainly possible that patients that have a pacemaker or have a known heart condition might not be the best candidates for vagus nerve stimulation. But in terms of other cautions or other contraindications, this is a lot of work that still needs to be done to figure out exactly what populations this works well for, and which populations it doesn’t.
Hannah Rosen:
So, can you talk a little bit about some of the current methods for screening for conditions that you're looking at in these auditory processing areas, you know, such as dyslexia, or maybe some other auditory issues, and what is the process like currently for matching patients with an ideal intervention?
Tracy Centanni:
So most of the current screening methods are behavioral. So, if we think about, for example, dyslexia, the vast majority of the time we're bringing kids into a clinic, they're doing an hour to two hour long behavioral battery and then a clinician who's been trained in identifying dyslexia will use those scores to identify them. In terms of sound, there are some additional procedures, things like hearing screening. We all remember in elementary school, the red and blue headphones where you raise your hand if you hear the tones. So, there's some additional techniques there. But dyslexia, especially because it's such a heterogeneous population, there's really no one-size-fits-all approach to an intervention. So even though we have a pretty standard set of assessment scores that would qualify a child as having dyslexia, there is no single behavioral intervention that can help everybody. So right now, the people who are doing the interventions are largely clinicians that are based in schools, so speech language pathologists or reading specialists. And most of these schools will have a single intervention option. So, if a kid is diagnosed with dyslexia, they just get whatever the school has available, regardless of whether or not that's the best intervention for them. So, it's certainly not ideal.
Hannah Rosen:
So what are some of the different types of interventions currently available?
Tracy Centanni:
So in the context of dyslexia, there are many, many different options, but the two major approaches that are validated by research would be ones that focus on phonological awareness, so the tiny little building blocks of speech. So, for example, if I say the word break and I say the word break without a B sound, the B in break would be a phoneme, and so your ability to remove that phoneme and say the word rake demonstrates some of your phonological awareness. And that helps with things like spelling and decoding and reading brand new words that you've never seen before. And then some other interventions focus on the speed piece. So, what we call automaticity. So, your ability to look at a letter and just automatically know what sound it makes, which is very important. When you become a fluent reader, if you are a fluent reader, you can look at a word and just pretty much know what it is without having to sound out every single letter. And so that's a function of automaticity. And then of course, there are interventions that combine the two use a lot of movement to help the kids make these connections and things like that. But those are kind of the two major skills that are emphasized.
Hannah Rosen:
It's interesting, when it comes to dyslexia and the causes of dyslexia, are there any genotypes that are associated with dyslexia that we know of?
Tracy Centanni:
Well, we know dyslexia runs in families. The heritability rates about 40-ish percent, and those numbers can vary a little bit depending on what part of the world you're looking at. But unfortunately, unlike lots of other diseases and conditions, there's no single genetic profile for dyslexia. There are a bunch of genes that are associated with dyslexia, so individuals that have reading impairment tend to have variance in this certain set of genes, but they're not consistent across individuals. So, for example, you could have two kids that both qualify for a dyslexia diagnosis, but their genetic profiles are completely not overlapping, so it makes it really difficult for us to suggest any kind of genetic testing just because there's so much variation.
Hannah Rosen:
Do you think it's possible that if there are potentially multiple different causes, or the symptoms of dyslexia that could then be related to what is going to be the most effective intervention for an individual with dyslexia. If, you know, you have two individuals that have completely different genetic profiles but are displaying the same difficulties in reading, could that then mean that one intervention will work better for the one individual than it would for the other?
Tracy Centanni:
Absolutely. And that's really where the field is going, you know, for decades this one-size-fits-all approach was dominant, and in the research side there was a lot of, I don't want to say infighting, but maybe a little controversy about what is the single core mechanism, what is the reason for kids to have dyslexia. And it's only been in the last you know maybe 10, 15, 20 years that researchers are starting to embrace this heterogeneity idea that as we started digging into the genetics of dyslexia, we started to see these different subgroups pop out. If this gene is abnormal in a particular individual, that might mean they have this profile behaviorally. And vice versa for a different gene. And so a lot of the work that we're doing, and others are doing, is focused on identifying some of those different subgroups and hopefully matching them with interventions that target the specific deficits that they have.
Hannah Rosen:
So, shifting gears slightly, what is the role that you believe that AI can play in this whole process?
Tracy Centanni:
Yeah. So, I'm going to say that I don't have a perfect answer here, and the reason is because this whole AI thing is still relatively new in our world. We've been hearing about AI, everyone's been hearing about AI in a lot of different contexts. And there's lots of funding now coming from places like NIH and NSF trying to encourage us to include AI into healthcare and in places like that. And so we've relatively recently been starting to think about this. I think that because there's so much heterogeneity in the population, there's so many sources of information we can pull from, genetic data we can pull from, neural imaging data we can pull from, behavioral data. We've already seen in some statistical modeling that the combination of behavioral data and neural imaging data improves your ability to characterize or to diagnose individuals. So, if you add in the genetics, this is very likely that that's just so much information that, you know, a human brain looking at those patterns of data, we're not going to be really great at doing that kind of analysis on the fly. But if we had a program that could pull in some of the genetics data, some of the behavioral data, some of the neural imaging data, and if we have large enough samples, which we do, there's lots of researchers, you know, thousands to tens of thousands of fMRI scans on dyslexia alone, for example, then ideally you would have the power of computation to not only identify subgroups that maybe we've missed just in looking at our small samples that we get in a single lab, but also in grouping those patterns and matching them potentially, or giving us suggestions of matching to different interventions that we can then test.
Hannah Rosen:
So, because there is so much variability in how this condition is presented, you know, what are going to be some of the challenges in training an AI model for personalized medicine like dyslexia when there isn't just one specific genotype or phenotype?
Tracy Centanni:
Yeah, I think the biggest challenge is going to be figuring out which variables are informative and which ones aren't. You ask any clinician, any reading specialist, any teacher for that matter, what are the main issues in dyslexia, they're going to tell you phonological awareness and rapid naming, those are the two big ones. And so yes, there are others, there are some kids that have difficulties with working memory or some that have difficulties with vision, you know, there are a handful of other phenotypes within the mix, but I think it's very likely that there are, for example, multiple genes that create the same phenotype. So, for the model, the model's gonna have to figure out which factors are helpful in putting a kid in a particular category and which ones are noise. And those variables might differ from kid to kid. You might have for example, two genes that we know interact with each other that in isolation mean one thing, but together means something else and it's just going to take lots and lots of data to figure that out. And I'll put in a quick plug here for a consortium that I'm a member of, it's the genetics of language consortium, and it's a worldwide team of language researchers that are focused on the genetics of language, and the big push right now is trying to standardize a lot of the measures that we use so that we can get these cohorts of thousands to tens of thousands to hundreds of thousands to millions of people in order to really dissect what these different genes are doing, which is a limitation right now in language science. You know, we don't have the sample sizes to be able to feed into AI yet, but hopefully, you know, with this push towards improving AI and bringing it into our world, we'll be able to get those sample sizes.
Hannah Rosen:
Yeah, yeah, it seems to be a common thread as we talk more and more about this concept of the lab of the future, which is going to require, you know, these larger data sets and the integration and communication between different labs and different equipments and all sorts of things is the standardization is really going to be crucial to to being able to move forward with these technologies.
Tracy Centanni:
It is. And when you think about language especially, there are so many challenges with that because of how many languages we speak and different ways of testing language around the world. If you're screening for a particular enzyme in a sample, that enzyme doesn't change necessarily regardless of where you live, or the test doesn't change. The results might change, but the test itself, you know, an EKG is an EKG, it doesn't matter where you are. But for language, the way I test someone's English is fundamentally different from the way someone would test their German. And so, figuring out how to come up with variables that are consistent, or ways of testing where we can combine samples across languages is it a huge challenge, and one that I think we need to embrace. And, you know, you mentioned me being a lone academic in the series, I'll just say one of the challenges that we have in academia, in approaching this challenge is the fact that we are in some ways very siloed in the way that academia creates the lab structure and the promotional structure and things like that, it's great to have large teams, but in a lot of ways we incentivize labs kind of doing things in their own bubble, so that's potentially one challenge as well, thinking about creating these measures that cross labs and countries and so on.
Hannah Rosen:
Yeah, I mean, that's amazing. That's a dimension I hadn't even thought about. You know, obviously it's got to be hard enough to get the same standard between different labs that are working on the same language but then yeah, when you add in the fact that different languages almost inherently require different standards just because different languages function so different. OK. That's, gosh, it seems like perhaps this integration of AI may be a little bit further off than I initially thought.
Tracy Centanni:
It is possible. You know, we're hopeful. I think there are ways that we can start small with AI. You know, things that we can do within the US, for example, and kids that are learning to read in English. There's lots of kids in the US that we could pull from just to start to integrate it, start to train some models and just see where we are. A lot of the, you know, machine learning, types of approaches, neural networks, those are not new in our world, but in terms of these "big” data sets that require what we're now calling AI and these adaptive algorithms, you know, that will take more data. And so, if we can create some more robust consortiums in the US to at least get us started while we're kind of figuring out how do we do cross language work, I think that's a good place to start because you gotta start somewhere.
Hannah Rosen:
Yeah, I'm now starting to think that my next question may be a little bit too forward thinking given where we are with AI, but I'm gonna ask it anyway, because I'm curious what your thoughts are. You know, I'm wondering what, because a condition like dyslexia, you know it's more complex almost, I think, in the treatments, because you have all these different layers for the intervention of, you have not only the patient themselves, the child, but typically you have the parents or guardians, you have what interventions are available to them through the school. And so I'm curious, you know, when we do eventually get to the point where these AI models become more robust and are able to put out these predictions and recommendations for effective interventions. How do you envision that transition going from using AI in the preclinical and just kind of research driven space into a more clinical setting where we're actually using these models to potentially diagnose and predict treatments for patients?
Tracy Centanni:
Well, I think the good news is that a lot of the clinic is moving in a digital direction anyway, so for example, there are a number of states now that are developing computer based adaptive screeners for dyslexia. So at least for me, I'm going to throw back, picture 5th grade, go to the computer lab when it was so exciting that the whole class get’s to go play Oregon Trail. The same kind of thing, but with a reading measure. So, you have kids at the computer or an iPad and they're doing this adaptive screening and based on their performance, the app or the program determines where you fall on your score range. So, you basically output an estimate of this kid's reading ability, areas where they might struggle, risk for dyslexia, and so because those systems already exist, if we can take what we're doing on the basic research side, thinking about how the genes and the brain patterns influence behavior, then that's more information to feed into those adaptive screeners that may provide just a better, not only output of their behavioral skill set, but also a suggestion of what intervention might be best now in terms of actually giving that intervention. That's a policy problem. I hate that that's the answer but it is. There are lots of states that, in the last 10 to 15 years, that have done a lot of really great work adding dyslexia legislation, putting it on the books as a diagnosable condition, providing funding for schools, for teachers and clinicians to get training in various reading areas. But it's not enough, you know, SLPs and reading specialists are just so overwhelmed, their caseloads are unbelievable and often they're working out of supply closets with minimal resources and so, you know, in terms of implementing customized intervention, that solution is going to have to come from policy and funding. And so, once we have some information on the research side about, you know, diagnostics and optimizing intervention choice proof that this works, then the next step will be going into legislative bodies and pushing for this kind of change because it's not something that clinicians have the bandwidth to take on by themselves.
Hannah Rosen:
Yeah, that's so interesting because it seems like, you know, because of the unique nature of this type, you know, typically on the podcast we talk a lot to people who are doing oncology research, and treatments for those sorts of diseases where it is very much something is wrong, you go to the clinic, and you go through the medical system. But because this is often treated as a behavioral issue and the interventions are mostly in the educational system and not necessarily in the clinic, it seems like there's an advantage in that, you know, you don't have to necessarily go through the clinic for diagnosis. You know, maybe they're doing screening in the classroom just as part of the typical curriculum, which is great. And so, you have all this opportunities to identify this condition early on, but then at the same time you're not going through the same options in terms of intervention and treatment where, you know, if you're in a doctor's office, they may say, OK, well here are the different options we have for treating your cancer and, you know, this is what your insurance will cover, and then you can kind of make your decisions based off that. Where in this case, it seems a lot more just like, here's what the school can do for you, even though this is a medical issue, right?
Tracy Centanni:
Yeah. And in some ways, it's worse than that because a lot of places, the schools themselves, aren't actually doing the diagnostics. So a lot of times, if a teacher, for example, has a concern about a child or a parent has a concern, they will go seek out a speech language pathology clinic, which is often not covered by insurance, which can be several hundred dollars for a screening and a diagnosis that the parents are paying for out of pocket. And then the clinic can either provide services, which again is a cost, or they can provide the diagnosis to the parent, which they then give to the school for an education plan. So, there's sometimes a disconnect between the clinic itself and what the clinic might recommend, versus what the school can offer.
Hannah Rosen:
Why is that? Why is there such a disconnect with this being not included typically in health insurance when it is a medical issue as opposed to, you know, if I, you know, pull a muscle or, you know, if I physically hurt myself, you know, a lot of times the insurance will cover physical therapy to help train that and, you know, get it back into proper working order. Why is it different when it comes to matters of the brain?
Tracy Centanni:
Yeah, it well, so certainly not all measures of the brain, but in this case it comes back to legislation. So, and this is state specific, some states do not have, to the best of my knowledge the last time I looked, a dyslexia category, medical insurance options. We think about historically, dyslexia, as you mentioned before, was often seen as a behavioral issue. A lot of schools, even today, still think that kids with dyslexia are somehow intellectually challenged, which we know is not the case. And so there really is this, for lack of a better word, again, disconnect between what we know about dyslexia, the biological basis of it, and the lifelong consequences of it. Kids with dyslexia are significantly more likely to end up in the justice system because of that, they have lower academic and vocational outcomes, self-esteem takes a huge hit. So, this is not just a “minor” struggle in reading, this is a really big deal. It has economic impact and of course it has personal impact on these individuals and so really it should be part of health care, but it takes a lot of data and a lot of convincing to get the legislation to adopt that and to add it into the books. And I think there's just probably a lot more work that needs to be done to fully integrate it into that medical system.
Hannah Rosen:
I wonder, you know, especially because there is this disconnected, so many of these interventions and treatments are more of behavioral side, and it seems like it can be so removed from the clinic, all of these treatments that are being done may not necessarily be undergoing the same study requirements to prove their efficacy as something that's maybe undergoing FDA approval. And so how could this potentially impact the ability to create an AI model to recommend interventions when you're then having to pull treatments that have potentially gone through FDA approval and some that have not.
Tracy Centanni:
That's a great point I think. We're lucky in this sense because on the research side there is quite a rigorous clinical trial process. So, lots of researchers doing work with randomized control trials for reading intervention, really studying its efficacy. Does it work, does it not work, long term follow ups, and even though those kinds of interventions are not going through FDA, they're not going through FDA because of behavioral intervention. You know, the risk of a side effect from a behavioral intervention, either you learn it or you don't. But we're not really doing anything medically complicated where you would worry about patient safety for example. So even though you may end up with a mix of treatments, so, for example, vagus nerve stimulation will absolutely go through FDA, it already has for a number of conditions. And so if we get to the point where we think vagus nerve stimulation is a really solid treatment for things like reading or language in either typical individuals or those with a disorder or a disease that impacts language reading, those treatments, those protocols would of course go through the FDA, but I don't think you would necessarily need to segregate them in the context of AI, but you may be able to provide some options for a clinician using the software to basically say here's a parent that has said, I don't want X, and you could maybe filter through. So, if a parent says I don't want stimulation, or let's say we have a child that has unfortunately a heart condition or is taking a medication that we find out contraindicates VNS, they may be able to put that in as an option, to say this kid can't have these kinds of treatment, and then hopefully the AI would say, OK, here's your next best option, removing these options, here's where you should go. But I don't think having a mix of FDA versus non-FDA, it shouldn't generate problems with the AI. It's really about efficacy and at the end of the day, if we can get the same bang for our buck with the behavioral intervention and we don't need to do stimulation, that is always better.
Hannah Rosen:
Tracy, thank you so much for joining me today. This has been a really, really fun conversation, both because it is just a little bit different than a lot of the conversations we've been having lately, and so that's always fun, but also just been fun to have my sister on the podcast and get to hear more about what you do every day. Thank you so much for joining us and, really looking forward to seeing where this research goes in the future.
Tracy Centanni:
Absolutely. Thanks again for having me and for letting me share my work with a a different audience.
