10.2021-Sense.nano-for-Session 4-specimens-biopsies-Q-A

BRIAN ANTHONY: Praneeth, Micha, Valencia, I send you the virtual and the physical applause. Wonderful presentations. And thank you for joining us on the panel now. So you've told us a little bit about your research. I'd like to kick off the panel with maybe learning a little bit more about you as researchers.

Maybe you could tell the story of the journey that got you to asking the particular set of questions that you were presenting today. Was it serendipitous? Was it strategic? Was it personal? How did you come to ask the things that you presented to us? So let's go in order. Let's go with Praneeth first.

PRANEETH NAMBURI: Oh, thank you, Brian. So I would say it's a mix of personal and serendipity. So my training, PhD training, was in neuroscience. And my main takeaway from my PhD was the better grip we have on the behavior, the better neuroscience we can do, the better we can understand the brain mechanisms.

How is this related to coordination and dancing? The behavior I felt closest to was coordination because of the dance lessons I was taking. And I decided to pin down the behavior and study it more formally. And that's how I arrived on studying coordination and efficient movement, inspired by dancers.

BRIAN ANTHONY: Very good. Thank you, Praneeth. Valencia, what's been your journey?

VALENCIA KOOMSON: Thank you, Brian. And thank you for the invitation to participate in the panel. For me, as a graduate student at the University of Cambridge, I spend a lot of time doing microchip design and thinking about how we could use the computational power of very small scale, micro, nanoscale electrical systems to do really exciting things.

And as I witnessed the explosion in mobile computing devices, in terms of handheld devices and microchips that are used for all types of applications, I started thinking a lot about how it could transform health care, and just really feeling passionate about thinking about how future health care technology could be transformed by making things smaller and cheaper, and bringing imaging tools into at the point of care, at the bedside, in the NICU environment.

I also had a personal experience, where my son spent time in the NICU. And it was fascinating to see the ways in which little, tiny, fragile infants are monitored using these huge machines. And so wanting to look at how we can make imaging tools, biomedical devices smaller, miniaturized, and really fit within the scales that are needed for bedside monitoring.

BRIAN ANTHONY: Very good. Thank you, Valencia. And Micha, what's your journey been?

MICHA FEIGIN-ALMON: Well, it's a combination of personal and professional. So I got to ultrasound imaging due to personal reasons, seeing the limitations of the modality due to assumptions people make and dependence on physician experience. And I started researching into how can I make the medical imaging problem less dependent on subjective experience.

And then I got into functional imaging due to discussions with a rehabilitation physician and learning of the problems of actually assessing how motion works, and also due to a lot of relationships with people with disabilities, and trying to understand how can we get better understanding into how muscle works and how to both assess and drive rehabilitation.

BRIAN ANTHONY: OK, very good. Thank you. So I'll ask another broad question. And it'll mean different things to each one of you. There's a lot of interest and attention in the miniaturization, certainly of devices. But the wearable devices are Fitbits or Apple watches, things that we're wearing not just in the clinical environment, but at home.

How does miniaturization and/or wearableness and/or wirelessness, if those are real world words-- how feasible is it for each of your various research domains? I know it's may not make sense with how do you capture video wirelessly and wearable with Praneeth. But you as think about taking a thing out of the clinical environment and putting it into the home, maybe that's the broader question.

Where do these technologies end up? Does it have to be in the clinic? Does it have to be in the hospital? Or what are the pathways you see where it's more a consumer device, or a home device, or is used by the end user in their own natural environment instead of a clinician for you on you in a medical environment? And maybe you can go in reverse order. Maybe I'll go with Micha.

MICHA FEIGIN-ALMON: OK. So I think the use of deep learning really opens the use of these technologies to wearable devices. And I'm more interested in how it can blend medical analysis to the home. Part of the problem is that when you go to the clinic, you only see a very short snapshot of the general condition.

And I think that this technology can give us access to a much better image, early diagnosis, better diagnosis, understanding how conditions affect the personal life of the patients. And I think this technology really opens the floor to new approaches to diagnostics, and treatment, and actually handling disabilities.

BRIAN ANTHONY: Very good. Maybe I'll go around-- I'm going to go virtual around my table at least. Praneeth?

PRANEETH NAMBURI: OK. So I think I would love to see our work used by physical therapists, sports physiologists, roboticists, and more broadly. I do think how the work maps on the translation that's needed. What we can do in the lab is to identify exactly what needs to be measured, or at least that's what I'm interested in. And once we have that part pinned down, I would love to see targeted sensors developed based on the biomarkers that we identify in the lab.

So for example, if we decided that deformations on the skin are what's most important, then what that would trigger the development of would be something like turning just right a video camera that you can put on the skin to measure skin deformations. We would want to invest resources into turning that into a consumer product. So yeah, that's my two cents on the question. Thank you, Brian.

BRIAN ANTHONY: Great. Valencia?

VALENCIA KOOMSON: OK, thank you. Yes, I think that remote and telemonitoring, telehealth is the wave of the future. And our particular technology where we're trying to look at brain oxygenation, how much oxygen is flowing to various parts of the tissue, particularly for brain development is crucial for clinicians to be able to see how these dynamics evolve as the patient is doing normal tasks and moving around.

And so being able to have a neural monitoring tool that can be used in a setting outside of the hospital is really crucial. One of the applications of this device is looking at cognitive development for children, particularly monitoring neurological disorders. And for premature infants, the risk of brain injury is very high due to the sometimes inadequate supply of oxygen to the brain in those early days of birth.

And so being able to track that over months or even years could have a huge impact or could bring about new insights into cognitive development and looking at the relationships between brain oxygenation and activation, and how that affects that infant over in the very early years. So I think that there's scope to really look at wearability and making these tools more off-the-shelf, just like a blood pressure monitor, a pulse oximeter that you can buy at a local store.

BRIAN ANTHONY: Good, thank you. So I'm going to take some questions that are targeted at each one of you individually now. And I'll go with Valencia first. This is maybe a two-part question. The person was commenting on an article they saw in the New York Times regarding skin color, racial biases in pulse-ox, blood oxygenation systems. And so the question is, are there biases based on skin color, skin pigmentation in the system that you're developing?

VALENCIA KOOMSON: Yes, that's a great question. Definitely in the near-infrared region that we are using, which is between 600 to 900 nanometers, melanin is an absorber in that range. And so what's needed is to-- and this is what we've addressed in our research, is applying advanced signal processing techniques at the front-end, in the front-end of the electrical system, which is the most sensitive part of the complete electrical system, designing sensors that have a wide dynamic range, where we could account for the melanin as an absorber of the light signal at those wavelengths, and being able to design circuitry that has noise characteristics that allow us to pick up very weak optical signals that transfers to the tissue.

So that is an issue. And the silicon microprocessing technology that we're using is directly applicable because it's the same issue with trying to pick up a wireless signal that transfers in through space on your cell phone. It's also transgressing through a noisy environment. And we have done quite well with designing transceivers to be able to pick up those weak radio frequency signals. And that's what we've tried to apply in this particular application with optical signals.

BRIAN ANTHONY: Thank you, Valencia. So I'll put the next question, I think, to Praneeth. The question here is, once you have the motion capture information or the muscle length information, how do you then use that in training? Or how do you give that guidance and feedback to the person that you're observing? Is it just you show them a plot? Or how do you close that loop?

PRANEETH NAMBURI: That's an excellent question. It depends on your setup. So far, in a studio/lab-type setting, we have provided feedback using augmented reality glasses, where we create visualizations of land and just show markers either going away from each other or towards each other after removing the distractions of movement through space itself. So I think combining this with augmented reality is a nice way of providing feedback, where we don't have to turn and-- we don't have to disturb our current motion to look at what's going on in the moment.

And I'll give a second part to that answer. I do think more work is needed in terms of applying machine learning and inference algorithms to understand soft tissue deformations. And if we could quantify those faithfully, perhaps we could provide these metrics using a simple camera system that somebody could install at home.

BRIAN ANTHONY: Very good. Thank you, Praneeth. And then for Micha, the question is, you commented on the training mechanism that you're using simulation. If you want to use this, I guess, on multiple organs other than muscle or body parts, what's your sense in terms of how much training, synthetic training, is necessary? If I add, how realistic do those simulations need to be in order to guide the learning? You're using simulation to try to do a real imaging scenarios. How much simulation do you need to do there? And is it realistic to capture it all over for the entire body?

MICHA FEIGIN-ALMON: It's actually the simulation, whether it was more generic, assuming a generic organ model as opposed to a muscle model. And we assume that whatever we're learning generically is also applicable to muscle. I do expect there to be some bigger issues in muscles because muscles are always a topic. They really depend on the orientation.

And we're really just looking for results the way they only partially translate to real tissue. I do expect that we need to develop-- the next stage in our work is to develop methods to be able to train on real data as opposed to simulation data. So it does look like the simulations are not realistic enough.

And the current research is into developing methods of looking at both collecting and using real data without having actual ground controls to train on because the big difficulty with real data is that we don't actually have the controls. We do not know what the speed of sound is. So going forward, I would say the simulations are not enough. And we really need some transfer learning methods in order to get better results and real data.

BRIAN ANTHONY: OK, very good. Thank you. So the next question, I think, intersects between Valencia and Micha. There's two sets of questions. But one is, is there an optical approach to monitoring of muscle and looking at muscle activation from an oxygenation perspective? And the second question is, what would be, from both of your perspectives, an interesting combination that would combine an optical an ultrasound interrogation of tissue? Is there an opportunity there?

VALENCIA KOOMSON: Oh.

MICHA FEIGIN-ALMON: So what Praneeth is actually doing is looking optically at muscles due to external motion. There is a big issue with doing optical tracking of muscle. If we want to look at deep tissue just with penetration, we can probably do some combined work of doing some of the acoustical methods.

But I would expect it to be difficult to look at internal muscles optically, which is part of the reason we're looking at ultrasound. And I would expect some deformation to happen-- that's really more for Valencia-- to come out of oxygenation. But there's different ways of activating muscles, some of them would would call. And they would generate different oxygenation patterns. So it's an interesting field of research. But I think we need to address different types of motion for that.

VALENCIA KOOMSON: Yes, great. And I agree with Micha. We are just starting to look at muscle hemodynamics. And one of the things that we are trying to do with our instrument is get better depth resolution of our transmitted optical signal so that we can see deeper into the tissue.

And so that that's to be determined, whether we can isolate specific muscles and be able to determine a map, if you will, of the hemodynamic response in a particular area of a tissue and really isolate individual muscles. That's something that's to be determined. We focus primarily on the brain oxygenation up to this point. But I think this is an interesting area to explore.

BRIAN ANTHONY: Thank you. So I wanted to close out the panel with one last Round Robin question. What's your question? So what is the burning question, the connection that you would like to have, the outreach that you would love to have from a clinician or a collaborative scientist? Why do you want people to reach out to you?

What are the things you want to learn from the people that are watching you now that you say, hey, reach out to me, I want to know these things, please help? So what are those burning questions that you would like to ask of the audience and to connect with you? And let's go-- I forget which order I've been going in. Let's go with Praneeth first now. And we'll go then Valencia, then Micha.

PRANEETH NAMBURI: I was actually talking about this to Valencia just now. The burning question I'm after is, is more coordinated movement more efficient from a metabolic perspective? And I was asking her if I could use some of her techniques to monitor metabolism in muscles as we go through different modes of coordination.

VALENCIA KOOMSON: Great, thanks. I'll follow on from you, Praneeth. Yes, I think that photon interaction with tissue is an exciting area. And this is the perfect opportunity, this era where we have off-the-shelf components that are cost effective that we can buy and build unique systems to explore all types of biomarkers and biological chromophores in the body and correlate that with health status, looking at muscle metabolism.

And like Praneeth talked about, I think that this is a great opportunity to do that in this era. So I'm looking forward to finding out from clinicians, what are your pain points? And what would you like to be able to probe or be able to see in the office? Outside of the clinical setting, what would you like to be able to track? And how can we do that using near-infrared spectroscopy and using optical signaling? So I think that those are still open questions, yeah.

BRIAN ANTHONY: Thank you, Valencia. And Micha?

MICHA FEIGIN-ALMON: I think generally when I'm speaking to physicians, my question often is firstly, what is your moonshot? What would you like to know as opposed to what would you think is the way to get there? And I think that often the problem speaking between physicians and engineers is that physicians think about how they know to get to the point. And engineers think about the same thing. And there's always a short circuit between them about what's the best way to approach the problem.

And specifically to my interests, I would really love to know how would understanding motion or what information about dynamic motion can help you in diagnosing patients and interacting with patients? And what would be a good way to compress long term data? If we do that diagnosis at home, how can I compress that information into a small amount of data that you can actually use to help your patients in practice?

BRIAN ANTHONY: Very good. Thank you. And hopefully, you get a flurry of people reaching out to you now. And so Valencia, Praneeth, Micha, thank you very much. Again, the virtual and the physical applause for both the wonderful presentations and the discussion. And I look forward to seeing and hearing great things about the progress that you make in the future.