Challenges for Mobile Diagnostics: Mobile MRI as a Case Study

Challenges for Mobile Diagnostics: Mobile MRI as a Case Study

Show Video

[DAVID SIMON]: Okay, hello, everyone, and  welcome. Thank you for attending Challenges   for Mobile Diagnostics: Mobile MRI as a Case  Study. My name is David Simon; I'm a research   fellow at The Petrie-Flom Center for Health  Law Policy, Biotechnology, and Bioethics at   Harvard Law School. The Petrie-Flom Center is  delighted to bring you this event as part of our  

research project diagnosing in the home the  ethical legal and regulatory challenges and   opportunities of digital home health. Before we  get to the discussion a few housekeeping matters,   we welcome audience questions for our  speakers, so please submit your questions.   Throughout this entire event we will be pulling  those questions to pose to the panelists.   You may be asking: how do I submit questions? The  best way to do it is to use the Zoom Q&A feature,   which if you scroll towards the bottom of your  zoom window you'll see a little button marked Q&A;   just type your question in there. I'll be  checking it very frequently. You're also   welcome to join the conversation or submit a  question on Twitter using the #privatemris;   again that's hashtag #private M as in mike, R is  in roger, I is in iris, s as in sam. If you do   submit a question there, Petrie-Fom staff will be  monitoring it and we'll pull it into the zoom Q&A   feature. Ways that you should not try to submit  questions: the raise your hand feature on Zoom;  

we will not be checking that feature. We've also  turned off the chat function so that our panelists   can really focus on what's going on, so that's not  going to be a venue for you to submit questions.   If you're interested in this event and interested  in other health policy, bioethics, biotechnology   related topics, we strongly encourage you to sign  up for the Petrie-Flom center newsletter to read   the blog the Bill of Health which features some  really cutting edge commentary by legal scholars.

Before I introduce our panelists, I want  to thank some of our Petrie-Flom staff   who helped make this event possible: Laura  Chong and Chloe Reichel. Thank you very much.   Our panelists today are Dr. Damien Fair, professor  at the Institute of Child Development, department   of pediatrics at the University of Minnesota  medical school, and he's also the Redleaf   Endowed Director for the Masonic Institute for the  Developing Brain. We also have Dr. Francis Shen,   professor of law, Mcknight presidential  fellow at the University of Minnesota   Law School. I'd just like to say a few words  to introduce the topic before we get going.   Technological progress is moving at a rapid clip,  of course the COVID-19 pandemic illustrates this   with various entities developing vaccines at  breakneck speed, but the rapid development of   medical technologies is not limited to vaccines.  Companies have been working to develop various   diagnostic tools that can be deployed in  new settings using new technologies with new   capabilities that were unimaginable or at least  highly speculative only a decade or two ago.  

For example the butterfly IQ plus is a highly  mobile ultrasound device that can be used with   simply an iOS or android device like  a smartphone or a tablet. Another   example is the EKO a digital stethoscope that  enables physicians to make cardiac assessments   that are recorded, playable, and analyzable on  an iOS or android device, and it can also be   used as a single-lead electrocardiogram. Both  technologies make certain diagnostic tools,   like ultrasounds and ECGs, more portable and  accessible potentially allowing diagnostic   assessments outside the traditional confines  of the hospital, clinic or physician's office. Technological advances, however, also have  limits, and moving technology outside of the   clinic poses legal, ethical and social challenges  for clinicians and researchers alike; for example,   if testing new technology requires introducing and  explaining the technology in a foreign language   how can we ensure that research participants  understand both the technology and what they   are agreeing to? What ethical obligations do  researchers have to communicate incidental   findings, that is, findings they might that  might occur through their research to their   research subjects? Do they have an ethical  obligation to provide care to their research   subjects or to connect them with care? Today  we have experts who will help us understand   these issues and many more using a different  technology, one that historically has been   much less portable than the two that I  mentioned, magnetic resonance imaging or MRI.   With that, I'd like to turn it over to our  panelists starting first with Dr. Fair.

[DAMIEN FAIR]: Alright, thank you, and thank you  for having me. I'm happy to come and give you a   little perspective from my space on brain  imaging in the era of big data. I am what   they call a cognitive neuroscientist, and  I do a lot of imaging trying to understand   the basics of brain development with non-invasive  techniques like MRI, and I think that we've   learned you know quite a bit over the last decade  or so. which as I was just noting earlier. is kind   of changing the landscape about how we utilize MRI  for both our research and for our various types of   clinical applications. Okay so there's two main  bullets here for what I'll describe today, one is   I'm going to give you a brief history of cognitive  neuroscience and functional MRI, which i'm going   to utilize as an exemplar here for where the space  is going with regard to using techniques like MRI   and the reason why we can see the use of  these technologies in the home and other   places making much more access accessible in  the in the near future, and then I'll talk about   some the idea of big data which is a big topic not  right now, where we have many many subjects and   studies and what that's taught us about where  we're going. All right so the the actual term   cognitive neurosciences is now about 50 years old,  and it was meant to describe the intersection of   the biology of the brain with various types  of concepts of the mind and psychology.  

There was a huge boost in this space,  in this field, with the advent of   PET imaging, and also what I'll talk about here,  functional MRI which essentially is measuring   the intricate nature of how neurons in the brain  kind of relate to metabolism and blood flow which   allows us to see activity in the brain related to  neurons without actually ever even touching it.   Now when this technology first came out and this  is kind of how I started my career there was the   potential of these technologies and how it could  transform our mental health and clinical practice   and research was was enormous; in fact I found  this old paper around the time when I was just   starting graduate school which was talking about  how fMRI can be used — here's just a little quote:   "Discussions of the future of fMRI have conjured  up visions of mind reading devices used everywhere   from the front door at the airport terminal  to back room corporate personnel offices.   At least one neuromarketing research firm is  already trying to use fMRI to probe what consumers   really think about their clients' product." So the  potential at least for for early on was very high,  

but while the techniques continue to be  valuable characterizing activation patterns, its   clinical utility has probably been relegated to  pre-surgical planning and and even then it's not   really widely used. Well in 1995 this guy by the  name Brock Biswald looked at different types of   activity in the brain; instead of activity related  to when you're actually doing something — so   you know when I'm like pressing my finger or not  and looking at where the activity in the brain   lies — he was looking at spontaneous activity when  the brain's at rest, not doing anything at all,   and what he's able to find is the spontaneous  brain activity um when you're not actually doing   any tasks can largely mimic the network in the  structure of the brain when you're actually doing   a specific task. It's very important because  it allowed for the use of these techniques for   people to be sitting in the scanner and looking  at the function without them doing anything at all   even being anesthetized, sleeping, anything  of that nature, so there's many fundamental   properties of organizations that have come  out of these techniques over the years   and the thoughts around the clinical and applied  revolution of what this is called functional   connectivity MRI was probably even greater,  maybe rivaled the traditional MRI at the time.   Still we haven't really gotten there. Well along  the way, as we kind of expanded our understanding   of brain organization and development using  these types of techniques and research,   the data sets were also increasing in sample  size or the amount of data collected per subject,   so now we're starting to do studies instead  of traditional studies of 50 or 100 people   now they're rivaling thousands of people, so 1200  people, 1500 people, 10,000 people a new study   out of the UK, called UK Biobank, 100 000 people  with the MRI scans, but at the same time as this   started to grow, we started seeing signs of  reproducibility failures where you might see   something in one study in one MRI which you don't  see in another study from a different institution.  

Lots of papers highlighting some big data  challenges and some findings that suggest   that maybe how good we thought we were doing at  identifying some of these characters in the brain   were not actually real and not really replicable.  So just to conclude this first part here,   the field continues to evolve as data collected  on the broader populations at a very fast rate;   however, the arrival of these and these very large  data sets are potentially highlighting finally   some of the parameters and the context — which was  required to get more reliable kind of data from   from various types of studies and in the clinical  ending to use more broadly in the clinical   sciences more effectively, and in the second I'm  just going to give you an example of what I mean   and just talk about this idea  of big data many subjects,   so this actually comes this content actually comes  from this paper that we just recently published   it's in press in Nature right now; it's a really  big deal for how we think — how we utilize,   contextualized data from MRI, and it's called  Toward reproducible brain wide association   studies, and it's just a very basic question,  right, and the question was: does the reliance   on on typical neural energy sample sizes in our  research studies provide an explanation of why,   when we do large studies relating to  clinical outcomes and things like that,   don't really replicate and if so why is that? I'm  not going to go through all the data, but I'm just   going to show you a quick little example which  revolves around this idea of sampling variability   which, from all your very basic statistical  classes, you've definitely gone over this which   just measures the effect size estimate and how  it varies between different sample samples from   the population, and one of my colleagues says  you know it's objectively boring boring and   rarely considered, but it's really extremely  important how we think about our findings,   so here's just the example: my question was what's  the relationship between height and age as as kids   grow from from nine to 10 to 11 years old? Now I  can go into a population — this is actually real   data — and I can grab a sample of 25 people,  25 kids, and I'll see there's a relationship   between the height and age — maybe it's a  correlation around 0.85 or something like that,   but I could go back into another population,  or just try it again and grab another 25 kids,   and I might get a sample that is highly  variable and where the correlation   across height and age is actually zero which  gives me a completely different result.   I can do this subsampling and sample 25, sampling  25, sampling 25, over and over and over again,   and I'll get a distribution of that  relationship, and what you'll see,   again this is actually real data from what's  called the ABCD study of adolescents, is that you   see of that small sample there's lots of variation  of those samples that you can get out this is   sampling variability we can get correlations of  height versus age order approach one or you might   even get something that actually negative  relationship is a completely wrong answer,   and then I can redo that thing, I can redo that  same exercise that samples of 40, 50, 100, 500,   9000 and look at those distributions, and what  you'll see is that is that you need lots and   lots of participants to be able to get the true  relationship which is approximately 0.5 or so  

to be able to get where the variability of what  you get is much lower. And why is this important? Well it's because if I don't do that; I don't have  a large enough sample, and I sample just a small,   small group of people I can get findings from  one research institution or one clinical research   problem that shows me one answer — this is real  data of brain imaging — that you might have a   positive relationship between some type of some  cognitive ability or psychopathology or somebody   else another institution will get something that  could be opposite which is even negative it means   the sample size need to be much larger than we  ever had imagined, and they suggest that these   that that these consortium level data for many  types of questions that we are trying to utilize   for clinical applications need to have  thousands and thousands of thousands of   people to be able to get there, nearly 2000  people for some of the highest effects. Now what this kind of reminds us of  actually is a place that genetics was   between 10 and 20 years ago where they were  identifying that lots of findings in genetics   you know and how it relates to specific types  of mental health disorders and other types of   diseases were not really replicating either,  and they also found the same type of issue, that   in order to get replicable findings you need  to have depending on the size of the effect   you need to have thousands of participants  yet all the studies were had had much fewer   participants in them, which is probably related  to some of the reliability issues that we have,   and this is a quote actually from this one  of these original papers just 10 years ago   which highlights how in this new  era of big data and small effects   a recalibration of views about what groundbreaking  findings is actually important and necessary.   So of course the genetic world didn't sit  there on their hands they came up with these   these ideas to be able to to kind of  leverage very large sample sizes to   identify the relationships with the specific  genes in various types of complex behaviors.   Now we're not going to go through all  these slides, in part because of timing,   but I'll just point out that what was required  is you would do what's called these polygenic   risk scores where you take a bunch of findings  in the genes, relate them to your outcome,   and do that over thousands tens of thousands,  20 thousands, hundreds of thousands of people   and then combine all those small effects to  identify in the target population how they   might be at risk for certain problem or a certain  issue or certain mental health disorder things   like that. Well in in the neuroimaging world the  exact same thing is now beginning to be applied  

where you can take those types of small effects or  differences, apply them to these very very large   samples, and then once you have all those types of  risks across the brain instead of your genes, you   can identify really specific relationships or risk  factors related to complex behaviors like, in this   case um in this case, ADHD, but again the point is  that to do this correctly you need many subjects,   thousands and thousands and  thousands of people, to do this.   Now one of the other lessons learned  from this time was that is that if you   don't do this right and you know the special  sauce of kind of what needs to be required   to be able to utilize some of this information in  the most optimal way for certain types of issues   that, when you're generating those big sample  sizes that serve as the base of the research,   is that you need to be maximally inclusive, so  here's just an example of some of these polygenic   risk scores in genetics and how most of those base  samples were based on folks from European descent   and the predictive accuracy, how well they work  in folks who are not part of that initial base,   seems to works extremely well if you also have  the same same cultural and genetic background   but not so well if you're from any different  kind of cultural group. Highlighting that,   this is an important lesson that we that to learn  is that if we're if we're going to utilize some of   these new techniques and this new understanding of  how to maximize the efficacy and the applicability   of rMRI and other MRI techniques then as we start  generating these large data sets you need to be   maximally inclusive of folks from various types of  backgrounds. So the potential for non-invasive MRI   to improve our understanding of brain function  and clinical outcomes of brain based disorders   is really, at this stage, higher than it's  ever been before, but it's going to require   larger samples than previously managed to really  realize that potential. Technologies that make MRI   or similar non-invasive neural imaging more  accessible and a broad use will undoubtedly be   part of this future; that's a bit of what Francis  is going to talk about and why we're here, because   now MRI is being made accessible, even in the  home, using various types of mobile technologies,   but while the work can put us on a more  solid footing with regard to the fundamental   findings in brain organization in clinical  applications, its growth and accessibility   outside of our universities and hospitals,  which has been the primary target of where   we collect these types of data will require a  keen eye to maximize representation of the types   of data collected but also an infrastructure to  promote its ethical use. So I'll in there — not  

exactly sure how we want to take questions  but just thank you for having me. I'm glad we   get to have this discussion here's just a  bunch of the people involved with some of   the data I just showed in the lab and also  just, of course, there's lots of funding   that we that we get to conduct a lot of the  work. I'll stop there and and hand it over. [SIMON]: Thanks so much, Dr. Shen.

[FRANCIS SHEN[: Damon, if you stop your  screen sharing, I'll share mine and   we'll get going here. Well, as Damien  said it's really nice to be here,   thanks to Petrie-Flom, to David,  to Chloe, Laura, everyone for   having this program. I'm gonna talk about what  Damien mentioned at the very end, the advent of   more portable brain imaging. I have provocatively  titled my presentation brain scans for everyone,   but I want to talk about the ethical, legal, and  social implications and the equity, diversity and   inclusion challenges that accompany this move  towards more inclusive and more pervasive brain   scanning, and I'll take you on a little tour  of work that we're doing including work that   a working group is doing involving Damien as well.  Just no disclosures; I want to acknowledge funding  

from NIH and others and particularly want to  acknowledge my colleagues Frances Lawrenz,   Susan Wolf, Mike Garwood and the NIH for the grant  work that is fueling most the presentation today;   we have a grant on highly portable and  cloud-enabled neuroimaging research, confronting   ethics challenges in field research with new  populations, and we have an awesome working group,   all these folks here, so anything I say today  I've learned from them, but what I say they are   my own thoughts and should not be subscribed  to the entire group, and if you get interested   in this work, we've got a really cool website including his bibliography,   and we put all of our work on there for everyone  for free. All right, so I want to cover three   things today, the first is kind of to set the  stage and talk about the developments in more   mobile and portable MRA, sort of but the potential  even for an at-home almost concierge service MRI,   and then I really want to focus the bulk of  my conversation around this identifying ELSI   and EDI challenges; there are many, and I'll talk  about some of them and then finally just a couple   words at the end towards solutions and invite you  to be a part of that that conversation as well.   All right, well let's start with the fun stuff  which is the emergence of these new technologies.  

I think it can be summarized in three, again  provocative, but I think grain of truth headlines,   one is brain scans for everyone including  vulnerable populations. The article that Damien   references, a great article and from the abstract,  I was just looking over again, here's what they   say: brain behavioral phenotype association  stabilize and become more reproducible with   sample sizes of n greater or equal to 2000. So  for anyone who knows a brain imager, here's a fun   game find that brain imager friend and ask what's  the largest sample size of subjects in your study   your MRI study; they're not going to be anywhere  even in the ballpark of 2000s, which tells you for   research purposes you need more brain scans and  of course for clinical purposes access, so brain   scans for everyone. Well in order to get that you  aren't going to have radiologists everywhere; you  

need to reduce the barriers of entry, and the idea  is that anyone can scan, homage to a great movie   a Ratatouille anyone can cook, but here anyone  can scan potentially, and I'll talk about this,   those who may not have the requisite training,  that is, just because you can get behind the car   and press the pedal and drive doesn't necessarily  mean you should be driving, and finally   in order to produce more brain scans, and if  anyone can do it, boy, you can really take the   brain scanner places it's never been before, so  brain scans everywhere. These technologies on this   screen they come from an article that we published  last year in Neuroimage, with permission,   and I'm not going to talk about any one particular  technology. I'm not going to get into the details;   I just want to highlight that there are a suite  of new technologies — these are only some of   them — different ones being developed; we'll talk  a little bit, and you see some images of this is   called the Hyperfine a company has a device. Mike  Garwood and colleagues Tommy Vaughn are developing   a different looking device; Larry Wald and  MGH have a device. There are others pictured   here as well, so there are many different types of  devices. From a law and regulatory point of view,   rather than key on any one particular piece of  technology because we don't know who will become   market dominant or what new technologies will  show up, the thought is to find the key features,   the common features, that define this suite and  to prepare and anticipate them. So let me just  

give you some from the headlines, so this is not  sci-fi; this is Twitter and YouTube. One of the   most interesting things for this conversation  today is the work of professor Shaun DeonI at   Brown and the Gates Foundation, and this is  from his twitter feed, advanced baby imaging.   They've live streamed, and put on YouTube, the  first home-based MR; they have an MR in a van,   and these are screenshots from their video which  you can go on and watch, and this is the picture   from like walking out of the house, and here's  the research team right there on your front lawn,   and this is from what they want to do — this  is the first time they ever did it — and they   want to do more they want to do MRI house calls,  right. Imagine DoorDash showing up except instead   of Chipotle coming out of the van, you can go in  and get an MR scan, so this is really happening.   It's also happening worldwide; this is just one  example from the head of the Hyperfine group,   and just to give you a sense of how different  this is, this was a tweet on the receiving end;   they said, we needed to construct a whole building  for our first MRI; this one fits in a cupboard.  

This is really different. Hospitals in the US are  doing this, too; portable MR opens up a world of   possibility; this is for our colleagues actually  here at University of Minnesota M Health Fairview;   this was one in Canada: new portable MR has the  potential to change the future of health care.   And just from two weeks ago, there's a story of  a group in Tennessee that is exploring the use of   this device in an ambulance. This stuff is really  happening, and there are a lot of possibilities,   new ones, some that we're exploring. Damien and  I put in a grant — the initial one wasn't funded,   but I think we'll find funding — we wanted to  create a Minnesota mobile MRI lab. Now this  

is, if you notice similarities, it's because for  purposes of the grant this doesn't actually exist   but we sort of labeled and said you know, what if  we took Professor Deoni's, and sort of modified it   and did here, and our thought was to  address inequities and access both to   research and then eventually care for MR,  and we wanted to to do that. The big point   is that to date magnetic resonance research and  a lot of other types of neuroimaging research   are geographically constrained, even our  vocabulary use, you have to go to the   research facility. Tomorrow's research and  clinical practice is field based and potentially   home based. Tou really could go anywhere and as  David said at the outset MR can be understood   as one of multiple technologies that are moving  outside the hospital and into our everyday lives;   there are a lot of reasons to be excited  about these technologies. For instance for  

consumers brain scans are in demand and at great  convenience, can increase access — Damien talked   a little bit about that — you can monitor  participants in more real-life environments   and potentially for some technology to be  combined and have real-life interventions,   maybe more objective data, and you  can scan with much greater frequency.   Typically we say have you ever had an MR not do  you get your monthly MR right and so these things   could all change and they're exciting; they're  exciting both for clinical and research purposes,   but there are a lot of ethical, legal, social  implications and a lot of equity and diversity   and inclusion challenges, and this has been  the work of our grant and the work that I do um   with with colleagues, and I want to talk about  that. We had a first grant a couple years ago   with Mike Garwood, Gil Gonzalez, MGH with  Susan Wolf, and some of the core issues that   we identified are the following: Informed consent,  again you're out there, you're you're not in the   hospital setting, privacy issues , I'll talk  about those in a little bit, this is really big,   establishing capacity to interpret and communicate  data to remote participants. You can bring the   scanner to the home; you can't bring, and so  somewhat misleading, you're typically not going   to have Professor Deoni and his entire crew there;  you're going to have maybe just the tech. There's   going to be extensive reliance on machine learning  and artificial intelligence; we don't have maybe   time to talk about that fully today, but it's  important, and I want to flag it because of that   there's potential bias and interpretive algorithms  especially in diverse populations. If you can take  

this scanner and you begin scanning in populations  you've never stand in before that's great,   but because you've never scanned with those  populations before, because you haven't had   diverse and large sample pools, because we're  only now beginning to enter the era of big data   neuroscience, that neuroimaging that Damien talked  about, what do you do in the interim? Can we trust   the data and the algorithm we currently have?  Return of results is a major issue because what if   you're out there, you're remote, and you find this  brain scan, and there's something problematic,   some structural abnormality? Again you've brought  the brain scanner out far away, but you haven't   brought the hospital, you haven't brought your  entourage of expertise, how do you handle that?   And of course access to data. I'll just briefly  say that, if this was the traditional model,   with everything pretty self-contained within a  research facility or a hospital system, the new   model is one in which the research facility is  left behind. The scanner is out scanning; the   data is being sent via the cloud; algorithms are  analyzing it; radiologists might look at it, but   that radiologist is not local and raises a lot of  questions, when you've got a more geographically   dispersed and culturally diverse set of  participants, you have less immediate access to   medical facilities, so if there's a major problem  in an MR scan in the facilities at Harvard or MGH,   University of Minnesota, the hospital is right  over there; there's a pathway; you can just   get them over, but if you're hours away what do  you do? I mentioned the greater reliance on AI.   There are also movements to utilize these  technologies internationally, in remote and   resource limited international settings. In  the interest of time, I won't go into detail;  

I just want to flag that, with another grant  from NIH and again a number of colleagues,   we put a piece out in her image last year,  and I just want to highlight a couple of the   take-home points that would be relevant for  direct-to-consumer and home use as well:   one is that there's a real concern around the  therapeutic misconception, which if you're not   familiar with that term, means for someone  doing just research or just wellness, there   is a misconception potential that it has  clinical value or that the data derived from it   can provides a brain health assessment, but  that's often not the case, so there's a disconnect   between what the consumer or the research  participant expects or thinks they're getting   and what they're actually getting; this could be  especially problematic with very powerful brain   data. We've got to ensure safety; I'll talk about  that in a bit. The privacy issues are pronounced   as well; there are structural abnormalities that  could be of great interest to insurance companies   for instance and how do we handle the flow of  data, who gets it, who gets access to it, and   its interpretation? I flagged AI, and of course,  this incidental finance matters as well. Two of   our take-home points here with this idea that,  in developing guidance, ought to be looking for   local partnerships and sustain local engagement  and creating sustainable value now, this thing   about local communities. When you transport those  ideas to direct to consumer setting, it means you   got to think a lot about those consumers,  especially those in vulnerable populations,   and here are some concerns, so kind of the flip  side of the promise. So one of the promises was   this is awesome for consumers, suddenly  it's like click a button on your app   and order a brain scan for later in the  week, what could be what could be better?   Well here's some perils: who's actually showing  up to administer this brain scan? One of the   values and the great promise of this technology  is you don't have to go to years and years   of training. If you know how to operate an iPad  and you know how to potentially position someone   in this device maybe you could acquire brain data;  on one hand, that's great; on the other hand,   that raises concerned about the standards required  for those operating the equipment. Then there's  

communication; it's one thing to get the data.  Damien, I'm sure, will tell you that that data   unless you have an expert to analyze it and then  an expert to interpret isn't going to be — I   wouldn't know what to do with it, a consumer  won't know what to do with it; there's heavy   reliance on this machinery of interpretation, and  the language of interpretation, and because we've   never had to do that outside the hospital or  limited research setting, we don't have like a   language to use. Genetics is a good example where  there's an entire field of genetic counseling; you   don't understand what those 23andme results mean;  there's like an opportunity to understand; we   don't have that parallel setup, something that I'm  arguing we ought to have but we don't have it yet. And then is there a plan for handling those  incidental findings? Oh great for me to have   a brain scan — not so great because you found  some tumor that I didn't know existed. As it  

problematic? Can I live with it? Do I have the  money to do further follow-up? Am I now living the   next three weeks or three years in fear? Suddenly  that on-demand convenient brain scan doesn't seem   so convenient. Just one other set of concerns:  it could increase access in many important ways,   especially to remote and marginalized populations,  but it could not, so how will the technology   actually be used. Will it fulfill its potential  promise will those go to issues I know NIH cares   about for instance? Who's the workforce using  this? Who are the intended consumers? Is this   being marketed for a fancy brain club like only  the high end could get this additional technology?   How do we recruit and retain diverse populations?  I can tell you because I've looked at this that   at present there's no neural imager training  for field-based research because it's never   been done before, like you've never had to take  your machine or never had the opportunity to take   your machine into the field, and these are things  that we're trying to think about and again access.   So putting all this stuff together, let me think  about solutions; just for a couple minutes.  

One thing to say at the outset, and I should  probably said at the beginning is that   none of the technologies that I've mentioned, that  I'm aware of, is intended as a replacement for   fixed traditional MR. There are a lot of  reasons for that; there are a number of things   that these technologies which rely on lower  fields and produce images of different quality,   maybe sufficiently high quality, but different  quality, there's some things you'll just never do,   but there are also some things that they'll be  able to do that fixed scanners have never done   and that takes us back to where it began. There  are going to be new markets for MR, again this   these technologies at least in my view are  less a replacement and more a supplement and   a complement and an expansion of imaging. It  really is imaging for potentially everyone,   potentially anyone can scan, but should they? And  brain scans could be done everywhere but will they   and should they? And to me the biggest question  right now is in this space there are — because   it's so new um there are no real standards, and  my question is will we collectively, the relevant   groups — these are professional organizations,  researchers, clinicians, developers, regulators,   patients, research participants, consumers —  develop high and meaningful standards to guide   this world of brain imaging. There is a future in  which we don't: there's a future in which brain  

scans run wild; you intersect that with business  and profit motives and you have a very problematic   world. There's also a problematic world in which  inclusivity and equity are thrown to the side   and, in particular, that the development of  these standards are not developed along with   a diverse set of stakeholders. The work that we're  doing in the grant, the work that we're doing in   other grants, is trying to address this, and let  me just say a couple words about that in closing.   MR is one of a suite of new technologies that  are going — consumer and digital technologies   with Benjamin Silverman and others at McLean  hospital; we've got a program at coming at   Radcliffe institute called intimate data: ensuring  equity as psychiatry embraces boundless data   and AI, and these are the sorts of things that  you know we're thinking about, I think ought to   be thought about, issues of racial, gender,  socioeconomic equity and justice and bias   and in the grant that I flag at the beginning,  this is our charge. We're right in the middle   we're kind of starting to develop our first set  of consensus guidance in the second of four years   and our goal which we'll accomplish is to generate  evidence-based consensus recommendations for the   ethical conduct of research using these new  technologies, using them in new and diverse   field settings so I'm confident we'll contribute  to these standards and I think that the sort of   pessimistic features I mentioned won't come  to bear, and I think law has a role to play   so since this is in a law school, our event,  let me close final slide about the rule of law.  

The clinical path I think is straightforward.  I mean it's bumpy and be convoluted but   straightforward, right. The idea is let's get  better measures of brain behavior and brain   function . Let's do that in the real world; that  gives us more individualized it's intimate data,  

and it's delivered to your door. Why do we want  to do that? Improve care, improve brain health,   development of novel cures for brain diseases  and disorders, but this just doesn't happen   without the law; on one hand at this early  stage where there's so much research needed,   we need to regulate the research and we need to  promote the research and promote it in ways that   adhere to our values and as it moves from the  research to the applied we're now boy actually   we do have co-shared service; we do have this  proliferation of brain data; we need new policies   new standards guidelines training implementation  and new laws and the work that Petrie-Flom Center   does at these intersections you know is right at  the heart of it. I'll stop there and stop sharing   my screen and thanks to everyone for questions  that I'm sure will come great thank you so much. [SIMON]: I think I'll start off with a  question that picks up on something you   mentioned towards the end of your talk which  is the market for these technologies, and one   question that came to mind was are there  other uses for the mobile MR technology   currently being explored other than imaging  the brain, for example, imaging the knee, or   some organ that's maybe less complex? Is there  research being done and how is that research   being carried out? Is it similar to  the kinds you've been undertaking? [SHEN]: Yeah I can say just a brief word on that,  so the answer is, yes, there are, and — caveat on   this: it's not my expertise, but I do track  it a bit. In fact I just was tracking a um mr   imaging company that is promoting a technology  that will allow for assessment of body health   in particular, a lot of things around fat in  the body and other thing. So this is happening   and their marketing plan, this is one example  but I think it's illustrative of ideas,   is that it will aid clinicians and patients  in decision making, and so without naming the   company by name, the idea was that they  I think they even said like color coded,   easy to understand output for you to then guide  your patient about she or he or they should do so   absolutely this is happening, and MR — Damien may  talk about this more broadly — is used robustly   on other other body parts; in fact most of  us if we've had an MR been like on the knee   or you know something something like that  so yeah that's that's happening as well.

[SIMON]: Damien, I don't  know if you had a comment. [FAIR]: Yeah, no, I was just going to say that  a lot of the early technologies on this front   have been driven a lot around the brain, but  it's coming to review, and it's not just the   MR technology that there's a lot of heavy  investments in — it's the the associated   infrastructure. Once it's collected where did  the imaging go? Like there's these are massive   in size and there's little tiny hard drives and  machines that go somewhere but then how do you   get it there and then; is it protected you know  because this is all protected health information?   So there are several companies — lots  of investment around the services about   you know storing, grabbing,  holding, viewing, analyzing,   processing like that service part of it is also  being heavily invested in as well, on all fronts. [SIMON]: Again going on this issue of  market access how do you see standard   setting organizations and stakeholders coming  together or working together to ensure there are   parameters that everyone can agree on and  follow to make sure the imaging we get   is of good quality and will be  used for the right purposes? [SHEN]: Damien, do you want to say something?  I have lots of thoughts from that but yeah — [FAIR]: Obviously I'll let  you start because that's the — [SHEN]: Okay, sure, yeah, so I guess several  things today, David — it's a great question.   So one thing that just has to happen is some of  the research, and this is starting to happen,   and so you know a basic type of research  is how does using the low field scanner,   one of the portable scanners, line up  to using the traditional fix scanner,   and it's a pretty straightforward set  of research studies that you run, and   these are happening across some hospitals and more  many more of these will happen; when the relevant   societies the neurologist radiologist will come  to a conclusion about whether or not they can   convincingly or credibly do the same things with  the new technology that they could do with the   old, and they'll determine we can use it for this  but not for that. I think that has — it takes time  

and it's complicated it's not easy but at least  it's regularly done when any new technology shows   up. I think that unique to brain imaging will be —  maybe it's involved because — let me rephrase that   one of the things that is especially important  in brain imaging and probably also especially   important in a few other places like genetics  is that the nature of the data collected   has implications that can run quite deep because  I mean try to save it right the brain is the organ   that puts it all together that really defines  who we are and so scanning the brain in my view   is different than scanning the knee and different  than scanning the heart not on the the technical   side and for the heart you'd have to also figure  out you know under the knee does this new device   do the same thing as traditional mr does? For the  brain I think you have to think carefully about   what would an influx of brain imaging  data do and especially do in communities   and amongst populations that have never had  this data before and with clinician populations   that have never utilized this sort of data  before and on that front I think the sort   of standard setting is more complicated because  you have more stakeholders you have more actors,   actors who you know haven't maybe done this  before, and that's you know part what we're trying   to do on the grant is this really diverse set of  professions and stakeholders, and I think if I was   going to start anywhere however I'd start with the  standards around who gets to use the technology   and what are the requirements, the training  requirements whether you're a graduate student   or you're an undergrad, right? This is  the technology that undergrad psychology   department could purchase or even a high school  could potentially purpose. Indeed the grant   that Damien and I would like to fund is we  would like to take this into middle schools   because it could be a wonderful teaching  device but we wouldn't do that at the part   of the grant is to think about okay how would  you train, what does someone have to know   in order to use this technology? Both about how  to use it, like which button to press, but much   more importantly about what to do like Damien just  saying about understanding how data is processed   understanding what an image is which a graphical  creation of the statistical maps, so there's   a lot of that work has to be done, and I hope  that we're you know a part of making it happen.

[SIMON]:   Okay I wanted to shift a little  bit based on something you said   about data quality this came up in I think  both of your talks, specifically the role   of AI and machine learning in  running these technologies and   you mentioned that — Dr Shen, you mentioned that  it's important to have a representative population   from which to draw data because otherwise you can  get data sets that don't track the characteristics   in the relevant populations and this was something  that happened in the in the genetic testing   context, and there was a question about this in  the Q&A also that related to, well, how do you   figure out which population maybe  to test against even if you get   all the representative data, and if we  start testing based on ancestry, are we   moving towards or away or somewhere different  from the kind of race-based medicine that   at one time was a popular way to — or at least  one way to to figure out what treatment to use? [FAIR]: That is a very complex question of course.  The call to you know an entirely abandon race from   medical research endeavors started several decades  ago. In fact the AMA, I believe just last year   just recognized race as social non-biological  construct for the first for the first time and   really you know the way that I've been thinking  that I've been thinking about this quite a bit is   making sure that how you use race or really what  which you oftentimes is used as a surrogate for   what you really should be measuring is like  some of the structural inequities that exist   in our society with regard to socioeconomic  status and things like that which is like the   the combination of which is kind of being  read out as a race thing even though it's   not really race it's not really the biology —  that you have to be really careful about what   your what the question what you're actually  asking what are you trying to figure out,   because if you're trying to figure out concepts  or answers related to the social inequities then   races is something you might want to look at, but  if you're actually you're trying to understand or   develop develop new therapies that are based  on the biology of the brain or the knee then   it's probably that's when you then you're  likely going backwards. I think that the,   in your terms, I think there is like a slow  movement to kind of recognize what this   what this actual difference is, but it's still  extremely complex, and there's several new um   papers and views out that are really quite good —  I should look them up and maybe put them in chat,   more recently, describing a more detailed way to  think about these this um these particular issues   in our research and how to move forward making  sure that when we're developing new therapies   that utilize these types of technologies that  we're inclusive enough such that you're not   biasing anything to even potentially be harmful  for one group in our society versus another. [SHEN]: If i could um piggyback on that Damien,  and I realized we should talk about this because   I think there are some really concrete questions  around this issue that are just now emerging   because of big data and more diverse data sets,  and I'll give you one of them. I'm presuming,   in most of the data sets, Damien, that you and  your colleagues are working in fact I know;   NIH requires data sharing at the end. Once you've  done it you put up your data set, and there's  

going to be you know for each participant all  sorts of information and you probably include age;   I don't know if you include age the participant in  the shared data, maybe you do, maybe you don't — [FAIR]: You can't you include  birthday but you can include the age — [SHEN]: Okay, which is I presume  because it's thought to be relevant;   you don't include hair color hair color completely  irrelevant I presume, but you include age and some   other things maybe like height, weight, like for  some of the developmental stuff for others — who   knows whatever you can connect I'm considering.  Should that data set include, if you had it, you   could have a measure of participant self-reported  race? Do you include gender in the data set? [FAIR]: In fact most of the demographic tables   there is some ethnicity and  race is actually collected. [SHEN]: It's collected but okay — [FAIR]: And shared. [SHEN]: So as that gets shared more  and more you got these big data sets,   what if someone comes along, not you,  someone comes along and starts doing   some of these studies that would seem to  take us back and draws inferences that you   can't do that but like the media really  — so it's like a really I think — [FAIR]: That is not theoretical; that is happening  today you know, so like the ABC. This is just  

happening in the ABCD study where you know  this is a big national sample of 10000 people,   and we've had to develop groups  to kind of read back some of some   of exactly what you're talking about you  know misuse of some of the information — [SHEN]: Right so imagine that on a grand scale  where you now have even more massive imaging,   and you're doing it at home, and you've  got you know the do-it-yourself — it's   a great question because on one  hand you want more inclusive data   sets for lots of reasons as Damien mentioned, on  the other hand you have to think carefully about   what does that mean for the practice of research  and then clinical care and clinical use,   and to my mind we're in a moment of flux, in  some ways a good reflection that people are   talking about it meaningfully and carefully, but I  will say is I think many people on this call know   there is — and this is not unique to brain  science there are other areas of science   as well — but there is a very sordid history  of brain data and race in the United States;   I mean just just horrible, and I'm always  concerned that we replicate that inadvertently. [SIMON]: There's also, just about that point,  gender-based misuse of data as well dating   back a long time. I wanted to ask a question  about — this is more of a technical question   but involves kind of legal and policy questions  as well also about AI, so assume that we get   a representative data set and we start running  a machine learning algorithm, some kind of AI   ,and what kind of tools do we have to double check  that after we've run the algorithm for a year or   two that it's still producing accurate  data. Do we have to update data sets or   if it's using data as it comes in, can that  potentially bias the algorithm if it's getting   fed data from different kinds of people? Those  those kinds of questions is more of like how   do we know that it's functioning properly, what  kind of tools do we need to use to make sure these   mobile MRIs continue to operate accurately that  we maybe didn't need in the traditional setting? [FAIR]: That's another really great question.  There are — usually in these models that are  

being built there's often kind of  the user which tries to monitor   success and changes and biases  and things like that, but   as we've seen particularly when we start getting  this smoothness of the commercial space, that   there's all sorts of conflicts and things you have  to you have to consider. One of the big pushes in   today's world is related to Francis's comments  earlier about how we share data is giving access   an open in an open science framework such that  there have lots of eyeballs these are the data   that's used to develop some of these algorithms  and that allow people to test the veracity of them   outside of the proprietary user? So I think  that you know like further infrastructure   to maximize the access to the information and the  utilization of that is something that can help   avoid some misuses and drifting of some of  these models to things that we don't want. [SHEN]: Yeah, and I would just  add that you know the systems   can operate remotely because the data flies  back to mission control where the proprietary AI   analyzes and spits out then the image that shows  up on whether it's iPad or phone or or the remote   location, and I think that transparency will  be important, but I think that transparency is   going to be in tension with IP and protection  of intellectual property; you've just invested   and your investors and your company have just  given you millions hundred millions of dollars,   you can't make that open source, so that's  not unique to brain imaging. That's where   brain imaging will — we can learn lessons  from other areas of law, but we haven't had   to confront that in brain imaging before and  just as there was ask for a chat, there are   many different pieces are neuroscience and race,  but there's great new work by Oliver Rollins.   I'm putting a one link to an article and then  I'll put his book link as well, great, great book [SIMON]: Great, thanks. So we have about  four minutes left and usually we like to   finish about a minute or two early, so what  I'd like to ask each of you to do is comment on   the following question or questions,  which is relating to this project   what keeps you up at night when thinking about  mobile brain imaging, both from you know the   positive perspective and also from the perspective  maybe having some concerns that make you worry.

[SHEN]: Well, I can go first. The thing  that keeps me up at night is the advent of   snake oil, the sales people um who show up  and are going to start direct advertising   and running late night commercials and be on radio  and are going to snicker people into thinking that   this imaging is giving them information that  it's not. I think that's, and then taking a whole   bunch of money and becoming rich while doing it, I  think that's a real big problem and very possible.   What I'm really excited about is that this is  an amazing technology; it's only getting better;   it's not perfect; it's not a  substitute for lots of things,   but it can be a real contributor to our advancing  understanding and then um of significant brain   health concerns and improving brain health  and improving mental health along the way,   so I think there's tremendous potential  here, and I hope we avoid the perils. [FAIR]: I think I just have to echo that by far  the thing you worry about the most is folks trying   to apply these new technologies to conditions and  things they can't actually assist with or answer.   I mean it already happens. I mean even without  this widespread accessibility and with it it's a  

it's a big fear that we've got to — we definitely  want to be you know ahead of the head of the the   game here to assist with avoiding some of the  the pitfalls of that accessibility as well,   but the possibilities are really amazingly high  at this stage in the game and like I was saying   earlier you know MRI, it hasn't been around in the  scheme of things it hasn't been around that long;   we recognize it for things that we use  for particularly in the brain you know   stroke and for tumors and things like that but  really the space and the potential for it to   expand into into um functions that don't  necessarily have a structural signature is   just enormous, and if you think today's world  about of understanding and characterizing really   complex disorders involved in all of mental  health and various neurologic conditions,   even in the even in you know in musculoskeletal  conditions and you know it's just the   the potential is extremely high and I think that  over the last you know several decades that we're   kind of finally there to have the right context  and all the special sauce of the technology,   funding, the understanding of all the  things you did wrong, all that's kind of   coming together all at one point so the potential  the next decade I think will be something   on this front that we that we  certainly haven't seen in the past;   the potential is just way high so that  part is extremely. extremely exciting. [SIMON]: Well, great, thank you, thank you  both for really interesting presentations   and a lively Q&A and we hope  that everyone has enjoyed this   webcast. Thank you for joining us  and we hope to see you next time.

2022-02-07 13:51

Show Video

Other news