Dementia impacts a person’s ability to complete day-to-day activities like familiar tasks at work or at home. What if we could identify these changes in everyday behaviors early enough to identify preclinical Alzheimer’s disease? That’s what Dr. Sayeh Bayat, an assistant professor at the University of Calgary, looked to find out. Dr. Bayat is the lead author of a recent paper highlighting how driving behaviors such as braking, following the speed limit and the number of trips taken could predict preclinical Alzheimer’s disease. Dr. Bayat joined the podcast to share findings from the paper and discuss some of the ways engineering and machine learning can help us discover more about dementia and aging.
Guest: Sayeh Bayat, PhD, assistant professor, Department of Geomatics Engineering, University of Calgary
1:05 - What led you to study this intersection of engineering and aging?
3:23 - What inspired you to study the topic of driving and aging?
5:30 - Who was involved in the study, and how long were these participants monitored?
7:01 - What did you find?
7:50 - Can you explain machine learning?
11:10 - Different health and life factors can impact driving. Is that something you’re looking to control for in future studies?
14:59 - How do you account for people who are just bad drivers without any cognitive change?
15:48 - What is the direction for your research in the future?
Learn more about Dr. Sayeh Bayat’s study in the New York Times article, “Seeking Early Signals of Dementia in Driving and Credit Scores” and in the BBC article, “How your driving might reveal early signs of Alzheimer’s”.
Find a free PDF of Dr. Bayat’s paper, “GPS driving: a digital biomarker for preclinical Alzheimer disease,” through the National Library of Medicine.
Intro: I’m Dr. Nathaniel Chin, and you’re listening to Dementia Matters, a podcast about Alzheimer's disease. Dementia Matters is a production of the Wisconsin Alzheimer's Disease Research Center. Our goal is to educate listeners on the latest news in Alzheimer's disease research and caregiver strategies. Thanks for joining us.
Chin: Welcome back to Dementia Matters. I'm here with Dr. Sayeh Bayat, an assistant professor in the department of geomatics engineering at the University of Calgary where she studies artificial intelligence, smart cities, health monitoring, mobility and driving, and healthy aging. Dr. Bayat was an author of a recent paper where she and her colleagues found driving behavior - people's actions behind the wheel of a car - could be used to identify people with the earliest stages of Alzheimer's disease. Dr. Bayat. Thank you for joining us on dimension matters to discuss this very exciting work. Now to begin with what led you to steer your career to this intersection of engineering and aging.
Bayat: Great. Thank you so much for having me.
Chin: Now to begin with, what led you to steer your career to this intersection of engineering and aging?
Bayat: So my story starts back in my undergrad days where I was going through engineering school. So I was training to become an aerospace engineer. All my internships, research focus, and summer jobs were in that field. But unfortunately when it came for me to graduate, the aerospace job market in Canada was taking a major hit so there weren't really many job opportunities. And at that point I had to decide what I wanted to do and I decided to go back and do graduate work. And one thing that I learned through my internships and through my experiences was that I wanted to apply my engineering skills in a different area that allowed me to interact with people more and work on real life problems that I can come up with solutions for. So this time I decided to go back to do my graduate work in the biomedical engineering area. It was really through my graduate studies that I got to learn more about the implications of this aging population that we're seeing for the societies. And I got to explore the ways in which engineering can play a role and we can develop tools and solutions so that care systems can be delivered in new ways and we can have people be independent and live in their communities for a longer time.
Chin: Well and thank you for sharing that story. You know, I will say in having interviewed and interacted with engineers, you certainly see the world in a different way - and I mean that in a good way. And so having that perspective, and that training, that skill as you said, is so critical in a field like aging and Alzheimer's disease where we need as much help as we can get in pushing our goals. So while I'm sorry to hear that your aerospace engineering career ended prematurely, I'm certainly glad you're here in the aging space, which leads me to your paper. And so this recent paper was published in the summer of 2021 - it's just fascinating so before we get to the results. Can you tell us what inspired you to study the topic of driving and aging?
Bayat: Yeah, so I have been interested in understanding the ways people interact with their environment through sensors, mobile technologies, for a long time now. And this is because I really believe that these interactions can tell us a lot about people's health and people's well-being. So my PhD thesis broadly was investigating whether mobility patterns from GPS data can be used to explain, influence, and predict image. But - so I was really interested in this general area, but this specific paper and project was the result of years of research that has been conducted by the Drives research team at Washington university in St Louis and they've been looking to understand the links between driving behaviors in naturalistic settings and early signs of Alzheimer's disease for a number of years now. And I was fortunate enough to be able to connect with Dr. Rowe and Dr. Babelau at a conference back in 2019 and that was really the start of this engaging and interdisciplinary project that led to this paper. And when I say interdisciplinary, our team members actually had expertise from Alzheimer's disease to transportation and mobility, naturalistic driving, blood-based biomarkers, so it was this complimentary set of skill sets that led to this work.
Chin: It sounds like a wonderful team and I'm going to ask you about naturalistic driving in a few questions down the road here. But one of the things that just is so profound is this idea of predicting dementia without actually doing any cognitive testing. And so that I think for our audience, just so they're aware - I mean this is a really new field and emerging field and it's so critical and driving is such an interesting aspect of it. And so like, at this point I'd really like to hear from you who was in your study and how long were these individuals monitored for?
Bayat: Yeah, so we looked at about a hundred people and this was over-65 each group. And what we did was over a course of a year in 2019 - so before the pandemic - we monitored the driving patterns of this cohort using GPS devices. And these GPS devices were installed into the onboard diagnostic port of each of our participants' vehicles. So our system really allowed us to monitor the car's performance and collect information from the sensors in the cars.
Chin: So this is incredible and I just want our audience to appreciate that - I mean this study took a long time. I mean, you did this prior to the COVID-19 pandemic and then the paper came out in 2021. But really, you followed people for over a year and they allowed you to to study their driving behaviors which I’m not sure how many people would really want that, so these very loyal participants. The other thing is, did you study people who came from the WashU research center, similar to the one here in Madison, Wisconsin, or the general public?
Bayat: Our pool of participants came from the WashU research center as well as separately through our advertisements in the community.
Chin: So it's a wonderful combination and so now at this point can you share with us, what did you find in this study?
Bayat: Yeah, definitely. So what we found was, using machine learning, we can identify very subtle patterns in driving that may be associated with preclinical Alzheimer's disease. And our model actually achieved very high accuracy and sensitivity and performance. But I also want to take a step back and explain what preclinical Alzheimer's disease is. It's a stage that happens up to 20 years before clinical diagnosis of Alzheimer's disease and it's when we have early brain brain changes with very subtle cognitive changes in the population.
Chin: And while you're explaining some of the terms, can you also explain machine learning too?
Bayat: So machine learning is a method of data analysis that really automates a lot of these steps that we take. It's really a branch of computer science that learns from the data and can identify patterns in the data and make decisions in a very autonomous manner.
Chin: And so before you tell us some of these subtle changes that you can notice, can you explain for our audience - you knew who had preclinical Alzheimer's disease and who did not and then you were able to study driving patterns that were different among those two groups. Is that right?
Bayat: Yes, so we had cerebrospinal fluid biomarkers that allowed us to know who in our cohort had preclinical Alzheimer's disease, but it's also important to note that our participants weren’t aware of this information and that didn't influence their driving patterns.
Chin: And so then you were able to look at the driving patterns within each of those groups, preclinical Alzheimer's disease and not, and determine if there was a significant difference.
Chin: All right, well then with that in mind, what did you find were the differentiating factors?
Bayat: So we measured driving performance, which can be how often you accelerate or brake aggressively, whether you exceed or fall below the speed limit of the roads, or whether you make abrupt changes or moves during your driving. But we also looked at metrics that explain driving space and these were, for example, the number of trips that you make, the average distance that you travel, or the number of unique destinations that you visit in your excursions. So a combination of these factors and metrics were shown to be important. One that was really important and I thought was interesting was jerk, which is the rate of change of acceleration. So it's really measuring how abruptly you're driving. Other metrics that were important were the number of trips that were made at night and the typical distance that you traveled to. So these were other metrics.
Chin: And so when you say jerking, that's accelerating, that's not hitting the brake more often?
Bayat: And so that really includes both, so whether it's acceleration - abrupt acceleration or deceleration in the form of braking.
Chin: Now in your paper you also show that driving more slowly and logging less total miles are factors too. Is that right?
Bayat: Yeah, those were the two other measures that we computed and we saw that they are correlated really with preclinical Alzheimer's disease.
Chin: And I'm glad you said correlated. So it's not that you're saying one is causing the other, but there seems to be a relationship to a driving behavior and performance and having these preclinical - meaning, you know, technically no symptoms or at least cognitive impairment - in the individuals. So the tricky thing for me is that, you know, a lot of different health and life factors can impact driving and I imagine it's really hard to control for those variables such as vision changes, back injuries, new medications, anything that could happen over the course of the year that you're studying these individuals. Is that something that you're looking to, in future studies, to somehow control for these kinds of factors?
Bayat: Absolutely. So that's a really important point. What we saw in this paper was that the results suggest that there may be some changes and differences in driving patterns of these two groups, but really we need to test the validity of our models in a larger population, more diverse population, and account for these variables that you mentioned before we can implement these solutions in real settings.
Chin: And you know, when people think about Alzheimer's disease they’re usually just thinking about memory change - I'm forgetful or I have a short-term memory issue. But this study would suggest, possibly, that cognitive - other cognitive functions like a person's ability to pay attention, to multitask, or even just their visual spatial abilities could actually be impacted even earlier and therefore impact driving. And I'm not saying you're stating that's the case, but that's possible. And so the reason I asked earlier about where the participants came from is that at a research center you probably have access to at least some of these participants’ cognitive test scores because that's the benefit of following people over time. I'm wondering if you're going to look at cognitive scores - and even if it's not an impairment but just cognitive change - and some of these driving patterns. Is that in the future?
Bayat: Yes, definitely. We're going to look further into these but I want to also mention that we looked at the cognitive rating measure assessment that was conducted with our participants and this is a measure that identifies the overall severity of dementia. It has six different areas that it accounts for; for example, memory and orientation. And all of the participants that were included in this study had a clinical dementia rating score of zero, meaning they had no cognitive impairment by this test.
Chin: And all rights, these were, you know, average of - regular people from the community.
Bayat: Yeah, exactly.
Chin: Wonderful. Okay, well then now I wanted to get at that comment you made about naturalistic. So how is driving assessment using a naturalistic environment better or worse than traditional, onsite Department of Transportation driving assessments? And if you could start by explaining naturalistic to us too.
Bayat: Yeah, so naturalistic driving is when we can really look at and monitor driving patterns of individuals in the settings that they usually drive, whether it's the road that you take to visit your friend's home, your daughter's home, the road that you take to drive to work, so we can really look at your driving behaviors in those settings. I would say the problem with the traditional road tests that we have is that it's really a measure of driving performance under a controlled condition and at a specific site and at a specific time, so that can really influence your driving behaviors. There are often really good measures of driving performance but they're not measures of daily driving behavior, if that makes sense and. Because of these differences, I would say the field of driving research is really shifting toward looking at naturalistic outcomes and this is now easier because we have access to mobile technologies, variable technologies, and we can collect these huge data sets.
Chin: Now I mentioned this interview to a teammate of mine who asked this really great question that I'm hoping you can answer, which is how do you account for people that are just bad drivers at baseline? They don't have any changes in their brain but they were just not good drivers to begin with.
Bayat: Yeah, so this paper in this particular project was a cross-sectional study where we compared two groups in this time period. I think what would be really interesting as a next step would be to look at people's driving patterns longitudinally to be able to compare each person's driving performance to their prior driving performance and patterns, and to see if you can identify changes longitudinally. But yeah, that's something that I think it would be important to account for later on but that we did not account for in this particular paper.
Chin: And so to end, I was hoping you could share with us what you're currently working on. What is the direction of this research or your research lab in particular?
Bayat: So I'm hoping to continue this line of research and we want to expand this work in different sites with larger populations so we can really test whether these models are - can be validated in different settings and with different drivers. And as you might imagine, driving is something that really depends on the weather. It can depend on the location that you're driving, whether you live in a rural area or urban setting, that can influence your driving pattern. So there are many different variables that we have to look into and account for in our future studies that I'm hoping to be able to look into it.
Chin: Well I hope so too. And with that, thank you Dr. Bayat for being on Dementia Matters. We do hope to have you on in the future when you have more of your data.
Bayat: Thank you so much for having me. It was great to talk to you
Outro: Thanks for listening to Dementia Matters. Be sure to follow us on Apple Podcasts, Spotify, Google Podcasts, or wherever you get your podcasts to be notified about upcoming episodes. You can also listen to our show by asking your smart speaker to play the Dementia Matters podcast. And please rate us on your favorite podcast app -- it helps other people find our show and lets us know how we are doing.
Dementia Matters is brought to you by the Wisconsin Alzheimer's Disease Research Center. The Wisconsin Alzheimer's Disease Research Center combines academic, clinical, and research expertise from the University of Wisconsin School of Medicine and Public Health and the Geriatric Research Education and Clinical Center of the William S. Middleton Memorial Veterans Hospital in Madison, Wisconsin. It receives funding from private university, state, and national sources, including a grant from the National Institutes of Health for Alzheimer's Disease Centers.
This episode of Dementia Matters was produced by Rebecca Wasieleski and edited by Caoilfhinn Rauwerdink. Our musical jingle is "Cases to Rest" by Blue Dot Sessions.
To learn more about the Wisconsin Alzheimer's Disease Research Center and Dementia Matters, check out our website at adrc.wisc.edu. You can also follow our Facebook page at Wisconsin Alzheimer’s Disease Research Center and our Twitter @wisconsinadrc. If you have any questions or comments, email us at email@example.com. Thanks for listening.