[00:00:06]
>> So thank you all I guess I'll just start by by kind of introducing the philosophy of a lot of the activities in my lab right now that are kind of ongoing and some of the things I'm going to present today essentially fill in some of the blocks that you can see on this diagram so the concept that my lab is really trying to push recently and really is not just my lab there's many There's many people around the country kind of working on this around the world there's a whole initiatives that are sometimes focused specifically on the electrical Neuromodulation technologies so around the area of sort of by electronics as well as as well as in sort of Neuromodulation through electrical prescriptions and those sort of things but the the concept here is the fact that if you can continuously measure non-invasively essentially digital biomarkers of help from a person and then feed those back either through the physician in the loop sort of feed back to titrate care in the simplest case it could be take different medications change medication dosages while the person is at home based on their changing needs or if you have even automatics sort of autonomous systems that can modulate let's say physiology through Neuromodulation sort of approaches then you could essentially have better health and performance outside of clinical settings so a lot of what my lab does is twofold one is kind of elucidating these digital biomarkers or objective measures that can come out of wearable data let's say outside of clinical settings through animal human or cadaver studies supporting the discovery of those underlying markers and then the 2nd part of it is really working on the overall system of both thinking about the sensing and also the modulation So when I talk about 3 things today could move through them a little bit quickly.

[00:01:59]
Because I guess we have bought a half hour maybe leaving a few minutes of them for questions and feel free to interrupt at any point by the way with questions I think is more interesting so I'm really going to talk about mostly our work in the area of wearable biomechanics for assessing the health status outside of clinical settings and focusing primarily for this talk on acute knee injuries and the rehab process following that but we also have done quite a bit of studies and juvenile arthritis and some other conditions that are more muscular helped with the soldiers rather than me injuries in the last I'll introduce some of our latest work on noninvasive Neuromodulation for p.t.s.d..

[00:02:41]
So when I 1st started doing this work I didn't realize how common knee injuries and knee based patient visits are especially outside of just athletic populations so you of course hear about it all the time of some basketball player you really love who's on the or who is on your fantasy team if you do that sort of thing.

[00:03:02]
Let's say rolls an ankle or falls funny on the knee tears an a.c.l. tears in this because you've heard about that but it's actually very common in the lay population as well that say somebody might be sedentary for a while then maybe go play tennis for the 1st time in years they can absolutely get injured it happens very commonly So it's about 18000000 patient visits in the u.s. every year to knee injuries and the related disorders This is a slide that actually one of my Ph d. students prepared that I really like and now use it all the time so that's the great thing about you know the students do our work and then we get the benefit from it so this is just a nice way of capturing that not only our knees one of the most often or frequently injured parts of the body in all of these different activities but at the same time in green you have the ranking of how the knee ranks in terms of the majority of severe injuries in these sports and activities so you have things like obviously soccer and hockey you have military activity there in the middle you know basketball in any of these it is both one of the top 12 or at most 3 actually one or 2 in terms of how often is injured and also when it is injured the number of days that you're out of your sport or out of your activity is either greatest or 2nd greatest So this is the process of after you have a knee injury what happens in terms of the diagnosis and the rehab so the 1st thing that happens is that typically you go to a doctor who's a specialist on speed expressionless they may kind of wiggle your leg around.

[00:04:40]
It obviously is making light of it but they do these sorts of maneuvers to test the structural integrity in different planes so do you have maybe too much wiggle in the lateral direction verses in and out of in and out of plane direction and then of course if they determine there's a need for it then they may send you for imaging it's an m.r.i. m.r.i. or a c.t. scan following that you usually have some sort of treatment or therapy that maybe surgery to go in and let's say replace.

[00:05:13]
A ligament that was torn or to scrape some of the or remove some of the fluid or various different procedures. After which you have this long period sometimes of rehab so in the case of an a.c.l. tear for example and answer crucial crucial ligament tear you might be looking at 6 months to 9 months rehab where you're slowly and painfully in meticulously going from one activity to the next determining when you should ramp up the intensity of different activities once you move to plyometrics When should you move to cutting or more sports specific activities and then finally you have to make some sort of decision about if you're ready to return to your sport or activity that in the 1st place cause injury and when this decisions poorly made you have a very high risk re-injuring that joint or even during the contra lateral side.

[00:06:06]
So the concept really that my lab has initiated and that is has been really fun aspect of our work over the past 6 years that I've been here has been the thought that you could use acoustic emissions from the knee as a quantitative index of need help and essentially by looking at when the knee moves the way that surface is internally articulating and maybe rubbing on each other maybe popping or clicking out of place that you could measure that quantitatively with small contact microphones on the surface of the skin and then maybe that could provide us exactly this dish a good digital biomarker that we need to be able to say titrate care whether that's sounds like you're nice to billet he has increased by enough or you can do plyometrics activities or whether it's someone with arthritis or you might say that a certain therapy is working well or not for them as quickly as possible.

[00:07:01]
The next few slides I'm going to get into are some work that we've done in a cadaver model so if you're sensitive to seeing photos of cadaver legs then now would be a good time to check your e-mail or do some other activity so. We collaborated with School of Biological Sciences here at g.t. to do this has to be called School of blood physiology but same people there's a new crop Sealab there where they have you know where we've been able to order these fresh frozen cadaver limbs.

[00:07:31]
Essentially just the leg right cut to sort of the mid quad region and these legs are fresh frozen so all of the mechanical characteristics of the tissues of the bone the ligaments are preserved and the great thing here is that we can do something that obviously would not be ethical in humans we can induce an injury look before and after and compare the signatures of the sounds that we measure to determine if there are some specific characteristics or features that are representative of that underlying injury and how it affects the sounds from the joint.

[00:08:06]
So I guess the procedurally it's pretty simple the leg is mounted to the table and with a hand that's actuated flex and extended at a rate that we've determined is repeatable and also represents how we'd be doing this interview so just quick need physiology knee anatomy information just to help you understand exactly what we did hear all kind of cut to the chase I guess you have course you have the femur coming in from the top you have the figure learn to be coming in from the lot of those remain bones involved in between of course the knee has held together by soft tissue and in between the harder surfaces you have this nice miniscule disc that sits in between there and allows for.

[00:08:53]
Kind of. Friction less or low friction rubbing between the surfaces and if you look now so this is from a side view if you look top down at the tibial plateau then you'll see what's on the right over here and what you can notice is that this lateral and medial meniscus are kind of 2 elements of the ministers that make up this overall disc and one of the most common terrors that the most common knee injuries that people face is a small tear in the medial ministers and so that's the particular tear that we want to focus on here for the reason that it's common but also for the fact that the way that that would affect articulation of this joint we hypothesize would involve clicks and pops because it's a mechanical disruption to the interaction between their services.

[00:09:41]
So I have a Mt ph d. student who had actually done some orthopedic sort of stuff a little bit enough so that he was comfortable doing these surgeries and he's been awesome and he probably will end up being a phenomenal orthopedic surgeon of course it took some time for us to figure out.

[00:09:59]
Exactly how to do this what sort of thing matters in terms of disrupting the mechanical properties of the tissue there's all kinds of gross things you have to deal with when you deal with cadaver limbs so we had a lot of learnings but by the time the protocol got really.

[00:10:16]
Refined and was clean enough we essentially went through these 4 steps so there's the baseline recording these are now the coup stick signals measured from these contact microphones on the surface of the skin we have the baseline recording before there's any sort of surgery there's a sham surgery where we just open up then the capsule to the extent that we need to to be able to access them in this case and create this tear and the goal there is we want to make sure that just the fact that we're doing any sort of surgery to the joint does not disrupt the sounds themselves because they mind and we didn't see any big difference or the next step is actual minutes because they're so the small radio meniscus tear again on the medial side that's sort of representative of a common injury that you'd see just example of what the signal looks like after that tear we do start seeing some of these high frequency short in time events clicks or pops associated with that tear and then following There's the minutes at to me which is kind of one way that people deal with the minutes because there we just take out a large chunk of the meniscus we didn't know what to expect there we were just curious about how that might change the sounds.

[00:11:23]
Now we use this. Metric which is already used so acoustic emissions I should have mentioned is commonly used in other domains so civil engineering mechanical engineering for looking at structural health so the health of structures rather than the health of the body and there's this one metric that people use often called the b. value which is essentially a way that you sort of look at the distribution amplitude values of all of these main events that occur at these pops or clicks and the differences in the slope of the b. value versus the number of events that are recorded can tell you about and tell you about how many defects essentially there are in a surface so we were excited and also not that surprised then to see that that if we look at baseline and sham We have a higher b. value indicating a lower amount of defects when you go to the actual meniscus tear the value drops indicating more defects and them in a 2nd we didn't really have much of an impact beyond the tear once you already have that mechanical disruption there it already leads to these sorts of pops and clicks is what we noticed.

[00:12:33]
And we want to understand a little bit better why so we did some 3 d. video imaging of the joint while it was going through its normal articulation So this is the same cadaver joint that we were measuring the coup stick signals from and with that we were able to map how close the different surfaces were getting to each other so the heat map shown on top here is actually looking downward at the tibial plateau and anything in red is getting closer to the surface above it anything in blue is sort of further away so it's kind of a good map demonstrating as the person well as the leg goes from 120 degrees $280.00 degrees extension to the surface is get closer together or further apart again our hypothesis was that if the surfaces start to get closer together and then move further apart that that might be the instant where we get the loudest or most repeatable sounds that's what we wanted to figure out and we were thinking that because in the minister's tear example you get more of these loud pops and clicks that in the minutes because Terry example we might see that the surfaces are getting closer at one point than they do in the normal non-insured case and it's hard to tell of course from this but that's exactly what we're seeing so in the case of the ministers tear and them in a sec to me we're starting to get slightly closer surfaces the distance between smallest distance between surfaces on this whole tibial plateau is decreasing the other thing we notice is after the Surface get close together following that point then as they start getting further apart is where we start getting these pops and clicks and that happens in both directions both in the extension and flexion directions so it was just giving us a little more information about maybe why these things are happening and why in the minuscule staircase we might be seeing what we're seeing so let's see if this works.

[00:14:25]
It's always hit or miss whether audio works I've done in this room before but now we don't have h.d.m.i. So whose may not work we'll see. You know my hearing was let's play with it a little bit. He gave me 2 seconds to try it because it's fun if it works now I'll know a little bit well you'll hear this one.

[00:14:58]
Now it's just my computer not. For you if you're interested there's stuff online and I can always post the stuff later but I think you can just see on the waveform that in the a.c.l. tear case you get quite a bit of quite a bit of variability. That's kind of an interesting interesting thing and gross thing to listen to both move on so my lab isn't working on now taking this and actually building it into a wearable system so we could deploy it in a few different clinical sites for measurement we've been collaborating with one of my good friends who's a assistant professor in the med school at Northwestern on this and really it's been an interesting sort of embedded systems development effort where we have these 2 boards one of them is capturing 4 channels high resolution high bandwidth audio from these contact microphones that we use from a place called Knowles acoustics at the same time we have this other board that's collecting this bio impedance and skin temperature data that I haven't talked about much but really it gives us some insight into the amount of fluid that's in the joints space the composition of the tissue and also a little bit about maybe blood flow there locally as well sort of neutral nutrient blood flow so that's kind of the set up for the boards we've actually built these things there's there is no mas the etymology is my friend and collaborator on this and we have now a system so it's you know it's big at this point but it collects everything we need to and everything is housed so that stuff is recorded on the s.d. cards with mozzie we've developed.

[00:16:33]
These tools for being able to charge a battery just through you know connecting in the u.s.b. his group has great software to be able to then download all the data from nasty cards or equity to the computer from which we can access it here at Georgia Tech and so we're kind of set to be able to do collection in clinical settings and we've started one of the 1st studies is here is actually converse of Minnesota in collaboration with Hubert limbs group who's doing this Neuromodulation therapy in patients with rheumatoid arthritis and so as they're going through this therapy day to day we're getting all these clinical variables that give us information about the arthritis status and at the same time we're also measuring our sounds of mild Pete and stated so that we can start doing some correlation and see if this device works in that setting.

[00:17:25]
So. Maybe a good time to ask questions on that 1st stuff because we're going to get into the sort of totally different the other side of things now kind of more the modulation but yeah yeah. What are we trying to measure. What I mean what what we're trying to.

[00:17:52]
Yeah. Like what physiological aspect Yeah yeah so we're keeping a broader we're saying we want to measure joint health right in some context so if it's a case of arthritis for example most likely it's an index of inflammation so it's a quantum quantitative in the underlying surface roughness within the joint space that's reflective of inflammation in the case of acute knee injury it's probably something that has to do with misalignment or variability in the way that the joint moves internally while the person's doing a basic activity like extension flexion but what we're trying to measure there is no gold standard.

[00:18:33]
And that currently is out there for what we're trying to measure so it's kind of like a Vital Signs of the knee but there is no vital signs currently for the knee so I can't give you like a wonder why it's not like blood pressure where there's clinical blood pressure maybe you want to measure it with a wearable device there's no blood pressure for the knee we kind of go in they ask us about pain and symptoms they do some imaging to look at the structure so if there were a ligament torn and then that would show up they can sort of poke it and try to estimate if there's fluid swelling and and maybe make an assessment based on that but there's no you know your knee x. is a 4 there's nothing like that there's functional surveys but they're super Of course it's like do you have pain walking up stairs to you know do you encounter difficulty with normal daily living activities are you able to stand up out of a chair easily there's that kind of stuff and then it results in a score of like $86.00 or something but it's not.

[00:19:32]
But there's nothing really directly in the clinical world that we compare to. Yeah so we were definitely interested in that we did a study where people were doing sort of vertical leg press where they're laying back and pushing up on weights and where we increase the weights in sort of a gradated sort of way going from 0 percent body weight well like one percent body weight to 336600 percent body weight and we want to see how that impacted the sounds and that's why it impacts the sounds in a very monotonic and quantitative sort of way so I would say the loading forces internally are definitely a factor in what's going on I don't think it's the only factor but absolutely that's And for the Tuesday I gave a similar talk to the robotics group and there I was talking a little bit more about the fact that you could measure these loading forces in real time that could be valuable for so the wearable exoskeleton kind of community.

[00:20:44]
Yeah. Yeah so we're collecting the data set now with Emory orthopedics and sports medicine that is larger than any dataset we've had so far it's still only on the order of like $40.00 to $50.00 people but they're kind of all comers with knee pain so there are a lot of different injuries represented.

[00:21:18]
Of course the most common ones like a.c.l. minutes because they're c.l. are still going to dominate but with that data set I think we can start answering those questions so are there injuries specific features that we can key in on or combinations of features that I don't have an answer to yet but I think we might I would say that for the cute knee injury versus arthritis case for knee injuries we're finding that more of the time domain features and more variability in the signal over time as the person performing the same activity is an important feature whereas for the arthritis case actually spectral properties of the signal most likely due to the fact that the arthritis signal is interesting from the grinding standpoint where you have again this may be increased roughness of the circuit surface grinding against other surface whether it's cartilage degradation or whatever's happening internally to the inflammation versus in the joint injury case I think it's more variability in the internal alignment of the structures due to either let's say a torn ligament or maybe in the minutes because their case you just sometimes have that tear causing abrupt you know you can feel pops in your knee when you have this because there right yes.

[00:22:37]
Already. That's like. That. Little right over you I'm sorry I didn't show that data here but we actually are national data set was in people post surgery during rehab. And so looking at how the joint essentially goes from the repaired point out to you know several months post repair and there are changes in that direction as well but it's just that the cadaver model is of course easier for us in a controlled setting to study the fact of the injury in the 1st place and then we use that same biomarker Actually we already published that in the rehab dataset and it works well for the in vivo case as well so kind of we use the cadaver model's a way to understand 1st and then we tried the playa to the human data.

[00:23:38]
Yeah it's a great question. I don't get into the whole change of gears here although it's it's all related so this sort of therapy and. The idea with so the Vegas nervous large nerve is a cranial nerves it sends information both from the brain to the body and also from the body to the brain various organs including the heart and lungs and the Vegas kind of.

[00:24:05]
Connects into the brain stem at locations that are known to have impact on autonomic tone so in other words are sort of stress versus rest and I just kind of balance that a person encounters and so this is by the way in collaboration with Doug Bremner who's psychiatry and radiology person at Emory and also is really a world expert in p.t.s.d. and treats people with p.t.s.d. all time as well as doing a lot of really really fundamental imaging studies in p.t.s.d. that on covered important factors like the differences in the size of the hippocampus in people with p.t.s.d. compared to not in other sorts of really fundamental biological findings but you know I've been working on this project with our with our group.

[00:24:56]
And the concept that we had was that if you pair v.m.s. together with emotional or psychological stress that specific to that person's own trauma that led to p.t.s.d. that you might be able to down regulate activity and certain brain regions that normally are really overactive in response to that emotional stress So for example the middle of the insulin answer your simulate regions that if they get overactive in response the stress lead to the person not being able to sleep well having a lot of anxiety during the day can't go through the normal functions and so that was kind of the concept for the project and the idea was from the sole concept of closed loop therapy based on sensor feedback that.

[00:25:41]
Could measure these real time physiological markers that could tell us potentially in the wild whether the person is encountering that sort of trigger that leads to their p.t.s.d. symptoms and if so at that moment you could intervene and stimulate that of course is much bigger than a single project the idea with this 1st project was to fundamentally 1st assess number one if you pair noninvasive v.m.s. together with the psychological or emotional stress do you see a reduction in sympathetic In other words fight or flight response to this emotional stress than you would normally see compared to sort of a sham stimulation and the number 2 are there noninvasive biomarkers that can be measured with wearable devices that are seemingly very well suited to quantifying that difference in these people so that's not about them doing so it's a really involved protocol it's a 3 day protocol that the subjects go through we have 2 groups one that we've already published that's people with prior trauma but no previous no p.t.s.d. So those are kind of our controls and the 2nd group which we finished data collection on haven't published yet but it looks like the results are gonna be very similar which is people with p.t.s.d. both groups are divided into 2 subgroups one that sham subgroup meaning that they receive stimulation but that the stimulation is at a frequency where you can't capture the Vegas so you're just going to be kind of buzzing the surface of the skin and the 2nd group which is the active group they're receiving on invasive v.m.s. where it is actually in fact stimulating the Vegas And it with a methodology that's already been proven with f. m.r.i. e.g. and other sorts of methods to actually stimulate the brain nucleus so.

[00:27:31]
And then they go through a variety of these stressors so they'll have emotional stress which is specifically they'll write down their prior traumatic experience as a detailed script of it that script is read back to them it's commonly used in psychiatry sorts of experiments and also in certain therapies and then as soon as they have that script read to them afterwards the noninvasive Vega stimulation starts and so the ideas that that Biggest will stimulation will immediately start down regulating those regions of the brain that are normally going to sort of fly off the handle and then in addition to the traumatic stress or as we have some mental stress and mental arithmetic built into the study as well we also have high resolution pet imaging at the various points and blood biomarkers I'm not getting into all that right now but all of those are kind of at the point where we're publishing them currently So the idea specific to sort of this stock is that we've then been measuring with wearable sensors the cardiovascular both electrophysiology and cardiac mechanics and vascular measures in response to this to this emotional stress or so we focus on really 3 signals here one is the c.g. which probably you guys have heard of or at least seemed kind of an ambulance ride to sort of this electrical signal related to cardiovascular activity and then the 2nd signal is a signal that my lab studies heavily which is called the seismic cardiogram which is kind of like the mechanical thumps of the chest in response to the heartbeat and the blood movement and then the 3rd signals the photo op with him a gram which is a peripheral measurement of blood volume pulse and in this study really you can think of that as a measure of if the vessels are dilated or constricted in response to stress so we have sham group an active group here for a photo for them and people g. amplitude a measure of age that I lay ssion in this case and on the right projection period which is a measure cardiovascular really is a cardiac time interval measure that relates to sympathetic tone so in both of these cases what we observe is that for the active group compared to shim group there is significantly lower sympathetic activity greater parasympathetic activity with the v.m.s. as compared to share.

[00:29:40]
This is now not paired with emotional stimulus emotional stress is a just a vs a low so this result confirms what's already known in the literature from brain imaging and blood biomarker sorts of studies but is specifically with our wearable sensors now this result was totally new which is really addressing the hypothesis so in this case we are using a trance containing us vs the noninvasive e.-n. us together with this exposure to the personalized traumatic script paired stimulus and sham versus active group we're comparing the response in people g.m. platoon and prediction period and respiration rate and again in all 3 of those cases when you do pair the v.m.s. with the exposure to this emotional stress or you get reduced sympathetic reduce sympathetic activity in the active group or the sham group so really supporting our hypothesis for the study and this is something that one of my Ph d. students we just we just have this accepted in brain simulation for publication and we're currently working on writing up the p.t.s.d. results so I guess that that's it for now.

[00:30:53]
An open question thank you. It has already asked questions. Here's. Yeah so the biggest difference is the lack of blood flow but the actual mechanical properties of the tissue are really similar and that's why they use this for training of Orthopedic Surgeons that's how they that's how they practice.

[00:31:26]
Like a practice field. Are we able to what with the top transcripts. So you're saying for the healthy knees that we look at how do we know that they are in fact healthy or for the cadaver for the cadaver we have a really detailed medical work up and history on the person whether they've had arthritis or not or whether they have a yes so we have all of that kind of information of course there's some possibility that maybe the person maybe in their 20 had some kind of injury we don't know about it but I think that.

[00:32:21]
Number one I think the medical record is pretty thorough for the people that. That we end up getting the cadaver from and also when we open it I guess theoretically if there was something really grossly wrong with uneven we'd see it in there. No no I mean I guess we would know because.

[00:33:00]
I guess 1st of all. We haven't found anything surprising I mean they're surprising but there's not like you know knee injury type surprising when we opened up the knees for any of these specimens. And then I think that. We don't have enough data at the age of the cadaver limbs to really know for sure so a lot of the data we have is for younger population and even kids for a juvenile arthritis project that I can talk about all day but for the population so the cadaver limbs are usually 7080 years old.

[00:33:35]
Folks of died we don't have enough normative data sort of on that population to be able to tell and differences and that matter the age definitely matters with a joint biomarkers Yeah. Without it's about so our typical protocol is the person will sit and we'll do about 10 of these flexion extension exercises that's usually about 2 seconds each so the overall thing is technically it could be 30 seconds or so probably more like you know a couple minutes to take the data these really do squats as well for a dataset and if we're doing both knees.

[00:34:16]
Yeah stuff I lost him. Thank. You. All it has to be on there so we use double sided tape for them for their contact microphones in the in vivo testing so the microphone is directly contacting the skin because these are the really wide band accelerometers so any airborne sounds are not going to cup went to the microphone it has to be vibrations like a stethoscope more than it is just a like more than a something like this right so it has to be coupled all the skin and that is one of the biggest challenges so for the could our model we had the microphones on the skin to avoid any such you know rubbing of the mic and other sorts of noise from that it's really difficult back to go from what we have now to a fully braced type solution where there is no as he said it all will be difficult but that's true for even electrodes right it's it's hard I think to avoid it he said.

[00:35:18]
At least for this next system definitely it would be really cool if we didn't have to but but with the types of microphones that are already package that we're using is harder but maybe we can talk about better ways. That's definitely a limitation. Thank you Rob.