Chapters Transcript Video Walk this Way: New Approaches to Driving Changes in Walking Patterns Ryan Roemmich presents at the Johns Hopkins Department of PM&R’s Grand Rounds on February 20, 2018. Recently started um where I was waiting a little bit because I know that, you know, the therapy team, but I just realized this, I realize that, I mean, the A PP A and I assume that those folks who did not go to the meeting, that PC for those who went, I, I think you just need to get more interesting speakers to come. How you wouldn't say that and now that you said it. Yeah. Um so uh now, actually, it's, it's good, it's fun. It's a pleasure to, to this. Uh uh So, Brian, this professional environment is uh uh working and studies in which is a BMG. Yeah. Um So Ryan joined the department now this year last summer and he is um he did his phd in Human Sciences in the University of Florida in Gas. And then he came to the and then decided to stay here. Um So lot of people who have not, I think it's very interesting about uh what the work that, that am and so on. It is very, very helpful to the patient. So, uh thanks and thanks Sam for the invitation to do this. I'm looking forward to it. So, um as Paulo mentioned, my labs right next door over in Kennedy Krieger, uh we study walking. And so the two primary questions that really drive my research are one. why do people walk in the ways that they do? And then two, how can we help people with gate dysfunction to walk better? And of course, these questions aren't independent of one another. Uh If we can understand what's important to a patient, we can then begin to target that through rehabilitation to help them walk better. So for instance, if we have a patient who prefers to walk slowly, because if they walk faster, then they start to have develop balance difficulty. Uh we wouldn't want to force them in our interventions to walk faster. But rather we try to target that balance difficulty in hopes that that would eventually lead them to walking faster. So we really try to focus on this kind of um kind of multifactorial approach where we try to answer these two questions both independently and then in combination to, to improve rehabilitation. And so to start off with, I have no uh relevant financial disclosures. Uh and then here are the three objectives that you'll see listed on, on the hand out there that were also in the email. And so when going through these objectives, I think it's actually easier for me to think of it as kind of three discrete different questions that I'm gonna be talking about today. And so these questions first, uh the first two questions that I'm going to ask today, we will basically cover with studies of healthy young adults. And so when we're studying healthy young people, mostly what we're studying is we're trying to understand the mechanisms that these people use to change their walking patterns or to learn how to walk in different ways. And so the first question that I'm gonna be talking about today, is this really kind of broad basic question about how healthy adults use feedback um to learn a new walking pattern. And then after this, uh perhaps an logical next question is once we've learned this new walking pattern, how do we actually store this for future use? So if we have a patient who comes into the clinic and we are able to make some kind of change in their walking pattern. On day one, we don't want them to lose those gains they've made on the first day when they come back for a second day, we want them to be able to store what they learned on that first day when they come back to the clinic for a second or subsequent visit. So, moving beyond just how they learn these patterns initially, uh we also do a lot of work trying to understand how these patterns are saved from day to day or from training session to training session. And then finally, kind of a different approach that we've been working with recently is to try to understand if people will actually begin to walk in new ways, if we can do some kind of intervention that makes that new way to walk easier than the patient's current kind of undesired pattern. So in other words, if we want them to change how they walk, how important is it that this new pattern that we want them to be able to achieve? How important is it for that to cost actually less energy than the pattern that they're currently walking with? And so I'll walk through these kind of step by step as we're going through the presentation today. So more broadly, uh when we're talking about gate rehabilitation, there are obviously many different approaches that we can take uh depending on the patient's deficits depending on what kind of equipment is available in the lab. And also depending on our understanding of how these two things can be used uh together. And so one kind of common way to rehabilitate patients is through this kind of conventional physical therapy approach where the patient's movements are instructed or encouraged by an external source of feedback. Whether that's a therapist like you see here in the picture actually describing specifically what she desires the patient to be doing. Um whether that's some kind of extrinsic feedback, like a game where the patients are getting feedback about how they're doing and they can kind of try to change their performance to improve at the game. Some kind of external um feedback source is providing this information to the patient to help them guide changes in their walking pattern. And then of course, you have kind of the opposite of that where there's also this kind of boom in mechanical devices that actually change the patient's walking pattern for them. So for example, this kind of fancy uh exoskeleton that you can see here, it's not telling the patient what to do. It's actually mechanically changing those movements for the patient. So in both these first two scenarios, you have some kind of external source that's driving the change in movement in the patients. But what I study a lot in our lab is something kind of in the middle. And so what I'm going to be talking about a lot today is the split belt treadmill, but it doesn't necessarily have to be a split belt treadmill. What the splitt belt treadmill kind of signifies in this uh line up of different kind of trial types here is that we can create situations where the device isn't actually changing the patient's movement for them. But what the device does is it causes the patient to change their own movement. And we think this is an important distinction to make because we can actually do things on the split belt treadmill to make it. So that the locking pattern that the person learns on the split belt treadmill doesn't actually depend on any source of extrinsic external feedback anymore. So in other words, people can learn a new way to walk on this treadmill, get off the treadmill. And without having to think about, you know, changing their movement in response to how the therapist was instructing them or having to have this external device actually moving for them, they can actually take what they've learned from this treadmill and carry it into their daily life. And so again, I'll be focusing on this uh for the uh the first two thirds of the presentation today. And so one thing I want to make clear before I go too far uh with our split belt studies is that we aren't necessarily just studying the split belt treadmill. Uh What we're actually more interested in studying is the type of learning that occurs when you're on a split belt treadmill. And we often consider this, we call it locomotive adaptation. So when you're adapting to a new environment that you haven't experienced before, how do you use this stimuli that's coming in from some kind of external source? Like in this case, the split ball treadmill to change your movement. And so what's important to understand about this is again, it's not specific to the split belt treadmill, but it actually probably pertains to a lot of different environments where the outcome of your movement differs from the outcome that you expected. And so here, when we have somebody walk on the split belt treadmill, oftentimes we have one leg go faster than the other. And all the experiments that I'll show you that involve the split belt treadmill, the one leg is moving twice as fast as the other. But you can also kind of envision scenarios like using virtual reality or if you're kind of waiting one leg more than the other in the left to try to retrain strength on one side more than the other. Any time you enforce something that makes the consequences of that movement differ from what the person expected. You're probably inducing this type of adaptation. So if somebody is walking in virtual reality, you can change the sensory outcome of their movement. They might be expecting, they'll walk at a certain speed. But then you give them a flow in the virtual reality that slows that speed down. So you're constantly having to adapt to these new surroundings and these new settings that you're placed in. But when you walk on the slip belt treadmill, some pretty interesting things happen. Um This is my most commonly tested subject and my personal favorite subject. It's my wife walking on all these videos. Uh When we have somebody walk on the treadmill, what we always measure basically our primary outcome is step length asymmetry. And so a step length is simply as you can see on the top right here. It's simply the difference in the anterior, posterior position of the feet at heel, strike So, in other words, how far do you place 1 ft ahead of the other when you're walking on the treadmill? So most of us without any sort of neurological damage or disease or any sort of musculoskeletal damage, we walk in daily life. We prefer to walk using symmetric walking patterns. However, you can imagine that when you start walking on the treadmill with one belt moving twice as fast as the other, that symmetry quickly disappears. And so here is she walking on the treadmill. Her right foot is actually going to be walking twice as fast as the as the left in this video. And this is describing this initial period here. Um It's actually about the 1st 30 seconds of her walking on the split belt treadmill. So again, if you look closely and I'll play this video again, once it finishes, you can see that her right leg is placed much more closely to her left leg than her left is to the right when she takes a step. So in other words, she's stepping very shortly with that right foot and then the left foot actually extends out. And of course, this is probably what you would expect because that right foot is getting continuously kicked out from under her by that faster belt. But the really interesting thing about this is so this is actually she walking alone in this room with me just controlling the treadmill. There's no feedback nobody is giving her, any instructions or advice on how to walk on this thing. And if I simply just give her 10 minutes of exposure to walking on that treadmill, you can see here. Oops, I think my battery just died. You can see here by the end of the adaptation condition that she begins to walk more symmetrically. And so again, even without any instruction, this is what she's gonna look like after 10 minutes of practice. But that's sorry, that's the first one again. And so you can see if you hadn't been able to see the belts moving. And I hadn't told you anything about the treadmill. She doesn't look quite normal, but she looks much more normal than she did at the onset of that split in the belts. So again, just by experiencing this asymmetry or this kind of limping that you experienced on the treadmill, people actually begin to update how they need to be moving in this environment to meet the new environmental demands. And so this is what we're studying. We study how people actually adapt to these new environments and learn new walking patterns. And the reason that we know she's learned a new walking pattern is now in the next video that I'll show you, we actually are going to move both belts back to tide speeds. So in other words, this is just a normal treadmill. It happens to have two belts on it, but they're moving together just like a treadmill would in the gym or at home. And so what's going to happen here is you can see this after she's learned this pattern for about 10 minutes, she, her, her nervous system wants to keep using it even though perhaps the environment has changed. And so when we bring the belts back to tide speeds, we'll actually notice that she begins to limp in the opposite direction. And so now both belts are moving slowly. So they're both moving at the exact same speed that the left belt had moved in the previous video. And what you're going to see now is the exact opposite type of walking pattern. So again, even though the belts are tied, and I she's actually told that these belts are going to be tied before she begins to walk, she begins to limp in the opposite direction because she's learned how to walk in this new environment. And she expects that environment to persist even if you explicitly tell her that it's not going to. And so this is where this becomes useful. You can probably already see with stroke patients. Um If we take someone who's had a stroke and put them on the split belt treadmill, we can actually, they begin to counteract the limp that they walk with in daily life. But one of the problems is of course that you can see in this time series is that that learn pattern if you give people enough, excuse me, experience and exposure that learned pattern will begin to die out. So people will actually leave our lab walking more normally, if we give them about 10 minutes of exposure to just normal treadmill walking after the split bell. But for those patients obviously want them to go take it with them. So, understanding both how they learn this and then how to keep it over a prolonged period of time are both important things to study when we are talking about this type of motor learning. So the first series of experiments that I'm gonna show you um are kind of geared at answering this question about feedback and how that might either enhance or not enhance uh this type of adaptation that I just showed you on the previous slide. And so as I mentioned, this type of learning doesn't require any sort of instruction or any sort of advice from the experimenter. And so when you're walking on it, people adapt to the treadmill primarily through proprioception. And so what we're going to ask first is, well, what happens if we also show people information about how they're limping? If we give them some explicit awareness of how severe their limp is on the treadmill, will they use that extra information about we, what we term an error or basically the limp in their walking pattern? Will they use that extra information about their limp to basically fix it faster? And so here's she again using uh this feedback system that we developed, you're gonna see two dots kind of bouncing up and down on the stream. The blue dot on the right shows the step length of the right leg and the red dot on the left shows the step length of the left leg. So when she begins walking, you're going to see those dots move kind of up and down the shuffle board looking type of series of targets here. And what that represents is basically how far she's stepping on the treadmill. So if she started all the way at the back of the treadmill, stepped all the way to the front, the dot would pop up at 15, halfway would be between seven or eight and so forth. So here she's actually gonna be doing the opposite uh configuration of split dot walking that I just uh compared to what I just showed you on the previous slide. So here she's actually gonna be walking with her left leg walking twice as fast as the right. And so what you'll see at first is that the left dot's gonna be landing a little bit shorter than the dot on the right, because she's stepping a little bit shorter with that left leg, you'll be able to see she's able to get these dots to match inside the same target, uh much faster than she probably would have been on the last slide without having any sort of this feedback. So you can see, she's a little asymmetric, but even within three or four steps, she's already able to start hitting the same target with both dots while she's walking, even though the left leg again is moving twice as fast as the right. And so we tested a whole group of people and we found that, yeah, this, this result held even when we tested several participants. And so here, the green group was walking with one belt twice as fast as the other while receiving that feedback about their step length. And the blue group was uh doing the same split belt walking test but not receiving any feedback. And so we found that these uh participants that received the feedback were able to actually uh acquire the symmetric pattern much faster than the patients who didn't or the participants who didn't. And again, this is all healthy young individuals. But what was interesting about this is that normally when we see that a learning rate increases, we find that the unlearning rate also increases. And so all of us probably know somebody who can just pick up a sport really quickly and then switch back and forth between different sports. And they're very good at kind of changing their movement patterns uh very quickly. That's kind of what we see here too. So if you're able to quickly acquire the symmetric pattern while you're walking on the slip treadmill, you're usually able to switch back to normal walking, uh very quickly afterwards as well. But when we look at these kind of washout curves, so when we bring both belts back to tide speeds, after the participants have learned the split belt pattern with either feedback or no feedback, we actually find that both groups uh begin to de adapt or wash out at a very similar rate. And so we didn't see this relationship. Yeah. Sure. Oh, sorry. No, neither group had feedback at all during the de adaptation. So they just both wash out good question without having any sort of the screen is just off in front of them and they're walking on the treadmill. So again, tide belts here, no feedback and both groups wash out at the same rate. So this was surprising to us because normally again, if you see someone learn faster here, they also unlearn faster here. So we began to dig into this and really try to understand OK, how are these people using this feedback to acquire that symmetric walking pattern um in this first exposure to the split belt treadmill? And so here, I'm shifting the axis a little bit, but uh the interpretation is still the same. So we this is just kind of a learning curve that we use to model our data computationally. And so as the learning curve approaches one that means that the participant is walking more symmetrically on the split belt treadmill and then just the opposite here after the dash red line as that line approaches zero, that means the participant is walking more symmetrically when the belts are tied. And so just like I showed you on the previous slide, participants gradually learn uh to walk symmetrically in that asymmetric environment and then they gradually unlearn that pattern when they're uh brought back to tide speeds. And so we put this on this scale to begin to, as I mentioned uh model it and one model that we, we commonly try to apply to different types of motor adaptation is this two state model. And so here's just one of the states and I'll walk you through this really quickly. Basically, the idea of this model is to observe what's happening now on this current stride that you're walking with and then try to predict what's going to be happening on the next stride. So try to predict the learning that's going to occur from one stride to the next. And so X here on this equation is simply the stride that you're walking with. So X of N plus one is your next stride X of N is the current stride that you're walking with. So you can see there's two terms here on the right side of the equation, there's an A times X and A B times ee is simply the error. So in split, but walking, we consider the amount of limp you have as the error. So how big is your limp? And then X again, is just what are you currently doing in the current stride? And so what these two terms give us is a, is simply a remembering term. So from one stride to the next, how much are you going to remember about what you're doing right now? And then on kind of the flip side B is a learning term. So it doesn't care about remembering, but what it wants to do is eliminate the error as much as possible. So it wants to get rid of that limb. And so you can see uh like a very simple uh representation of this might be, if you had an A of one, you would be remembering everything about what you're currently doing when you take your next step. And then if maybe B was 0.5 that means you're correcting half of the limp that you have. So from one stride to the next, your limp would decrease by about half if A was one and B was 0.5. But what we think actually contributes to this total learning is two different states, a fast state and a slow state. And so you can see the fast state has exactly the same uh structure in the equation here as the slow state. But what we do when we model it is we actually put constraints on these different A and B terms uh to generate a faster and a slower process. And so with the slow process, we really up this a term such that the slow process really is biased towards remembering what you've done previously and continuing to do that in the future. On the flip side, the fast process is really biased towards learning. So it's, it really wants to rapidly correct errors as quickly as possible. And so what we can see here is when error is very large, the fast state is very active because you have a lot of error that you're trying to get rid of. There's a big limp in, in your walking pattern that you don't like to experience. And so when this fast process gets geared up at the beginning of adaptation, it gradually begins to fall off as the air gets smaller because again, it doesn't become so important. Once you're able to eliminate some of that error, and then just the opposite as you begin to eliminate a lot of the error, the slow process starts to build up. So it's really trying once you get in kind of a good pattern that you're feeling, you know, you're walking symmetrically on the split belt treadmill, error is small, the slow process begins to continue to try to keep this pattern going. And so that's why we think there might actually be that after effect that occurs when we bring the belts back to tide speeds. You can see that slow process is very slow to de AAPT as well. So it remains biased towards that split belt walking pattern. But anyway, we just use these two processes and then we sum them together to get the total amount of learning. So if we add this dashed fast line here to this more solid slow line, that equals this total uh blue learning rate that we see. So in other words, we just think the learning rate is composed of these two processes fast and slow the slowest bias towards remembering the fastest bias towards learning. So one easy way to explain what happened when we gave participants feedback was to simply increase the parameters on these models, such that people are just learning faster. So if you do that, you can generate the same kind of uh output that we saw in the previous slide, green is with feedback, blue is with no feedback. And by simply increasing the parameters on that model, you can make the model run faster to learn faster and it begins to walk more symmetrically faster. However, there's another explanation that we considered where we thought that perhaps this extra visual information that you're receiving actually doesn't play into the learning at all, but it actually triggers a separate kind of conscious correction process. So instead of feeding into this learning model and being able to actually learn the new walking pattern faster, maybe you're just kind of using this fudge factor in your nervous system where you're trying to satisfy what you're seeing on the screen, you're trying to walk symmetrically on the screen, but it doesn't actually play into the learning of the new walking pattern that's happening underneath. And so instead of changing those parameters and making the faster process even faster and the slower process even faster, we consider that there might just be kind of this additional process working learning where you're consciously correcting uh the movement. So in other words, the output is not accelerated learning or rather the output is just the same learning. But now you've added this separate conscious correction process to it. So in other words, you might have something like this where those two, the dash blue line and the and the light solid blue line remain constant whether you have feedback or not, but you just bring the separate process online where you're using the visual feedback to consciously correct your movement. And so then we used the predictions from these models to actually test them to see what was actually going on uh in the participants that we tested. And so how we did this was we considered a group where we gave them baseline walking. So they just walked normally on the treadmill for two minutes prior to exposure to the split belt treadmill. And then once they were on the splitt belt treadmill, we gave them feedback for only one minute and the belts remained split. But we took the feedback away. The reason we did this was because if the feedback had actually caused accelerated learning, we would expect that taking the feedback away would have no effect because the feedback had already helped them learn faster and helped them achieve a symmetric walking pattern faster. However, if the feedback was driving this conscious correction process, we should see this kind of strange effect where they can learn really or they appear to learn really rapidly at the beginning. And then once we remove that feedback, they actually fall back and become more asymmetric because again, they were relying on that feedback to be symmetric. And they hadn't actually learned using the feedback. And so these are kind of competing predictions of the models that we thought were able to actually help us tease out? OK. Are they learning faster or are they using this separate conscious correction process? And then really interestingly, this is actually probably my favorite part of this experiment. Uh We thought that you should theoretically be able to do the same kind of test in the wash out period. So in other words, if we gave people feedback during only the 1st 30 seconds of the wash out, they should be able to use that feedback to restore a more symmetric walking pattern in a normal symmetric treadmill environment. But if you pulled that feedback away, and it was driven simply by conscious correction, they should actually begin to start limping kind of spontaneously again. So even though they had been able to eliminate that limp using visual feedback, if it was conscious correction and not actually learning, the limp should reappear once we remove the feedback, even though again, they were walking in a tide belt, normal treadmill environment. And so again, this prediction would look similar to this where they have this kind of fast unlearning driven by the feedback, then you turn the feedback off and they should actually bounce back up and become more asymmetric. And so we tested both of these conditions in different groups. Again, these are all healthy young individuals. And we found that our results actually played out very similarly to what we had predicted in the model. And so here, when we gave participants about a minute of feedback and then pulled it away, we found that their limp basically reappeared in both the split belt treadmill, walking environment. And then even if we let them walk in this little treadmill, walking setting for 10 minutes and then wash them out with feedback, they could get very close to zero using the feedback. But once we took it away, you can see they bounced back up on the red curve here. So this showed us that the the behavioral findings that we got from our healthy participants actually fit very well with our conscious correction model and not so well with this accelerated learning model. So kind of the important takeaway from this particular set of experiments is that even if somebody is able to do something faster, that doesn't necessarily indicate faster learning. And so for providing feedback over the top of some other kind of learning mechanism, we need to really be careful about understanding how these two mechanisms are interacting. And so when we read a lot of the literature, especially some of the motor learning li excuse me, motor learning literature uh from the eighties and nineties, there's kind of this contrasting uh these contrasting viewpoints about whether more feedback is better or whether more feedback is worse. And what I'm telling you with the walking data that we have is that it's not necessarily better or worse, it's just different. And so we need to really be careful again about understanding how these mechanisms are interacting or not interacting. And in this case, they're actually working in parallel. So people are able to use the feedback to make voluntary changes to their movement over the top of making this more these more subconsciously driven changes in movement um that are caused by the split belt treadmill itself. So we can both, we can leverage, if you're in the clinic, you can actually leverage voluntary correction by telling somebody, OK, take a bigger step, take a bigger step while they're walking on the split belt treadmill and potentially get the gains of they can walk more symmetrically faster with your instruction, but then they can get the lasting effects from the split belt treadmill underneath because again, these two processes are working in parallel. Yeah. And so then again, the second part of my talk, as I mentioned at the beginning is going to be geared more towards understanding. OK, once we've actually learned a new walking pattern, um how do we store that pattern for future use? So if we're delivering multiple sessions of therapy, how should we be delivering these sessions so that people can actually take what they've learned from one session and carry it over to the next. And so this is something we've studied in the lab for several years. And we, we term this kind of phenomenon where relearning occurs faster than original learning as savings. And so if everybody thinks about riding a bicycle, when you learn how to ride a bicycle, the first time you were on it, you're probably very wobbly, it's very difficult, but you can put the bicycle away, go to bed, get up the next day, pull it back out and you're gonna be better at it the second day than you were the first day, even though you haven't been, you know, riding it since you were first on it. And so we see the same thing with our split be walking uh paradigm if we give people four exposures of split, but walking, which are represented by these colored blocks here and we separate them by little bouts of tie belt walking that wash out the split belt learning effect. What we find is that from one session to the next, people gradually begin to learn faster. So you can see this first blue line is relatively slow. That's the first exposure. The green line is the second exposure. The yellow or orange line is the third and the red is the fourth. So from one session to the next people actually begin to hold on to what they had learned previously, even if we wash out that learning um between each one of the exposures. But what we don't really know is how different types of practice or different types of training structures uh influence this day to day savings. So one question that I was really interested in understanding is if I want people to be able to hang on to what they had learned on the first day as best as possible when they come in for the second day or the second training session, how should I deliver that training on the first day? So that that happens. And that's kind of the question that I'm gonna be showing you some information about on these next few slides. So the first experiment that we did uh was we, we manipulated the ways in which we delivered the initial exposure to the split belt uh paradigm. And so here what I'm showing you uh any time you see a block again, that means that the belts are split. And then if there's a flat line, that means the belts are tied. So this is again a study of healthy young people where we bring them in, we give them 10 minutes of exposure to walking with one belt moving twice as fast as the other. Then we give them 10 minutes of walking with both belts moving at the same speed. And then we re expose them to the same split belt configuration that they experienced on this first training session. And so when you look at what the time series here looks like, uh You can see that during that initial exposure to the split belt, they learn relatively slowly. But then even after washout, when they've begun to walk uh symmetrically again, even in a normal environment, when we re expose them to the same split that they experienced initially, they learn much faster. And so this is something that I again just showed you on the previous slide. This was kind of our control group uh for this set of experiments. But a lot of times in our patients uh handling this abrupt split and the treadmill belts can be relatively difficult. Uh It puts a pretty strong demand on balance. And so another option that we've been looking at is actually delivering this training more gradually. So when people experience the gradual split belt perturbation, the fast belt eventually speeds up to two, twice as fast as the uh slow belt, but we can deliver that more gradually over time. So here the belts actually don't reach that 2 to 1 split until nine minutes after the training has begun. So again, it's this kind of gradual ramp up of one belt speed that makes it a little bit easier for the patients to walk on it, but it might have actual consequences for the learning. And so that's what we, what we wanted to know here is if we deliver that perturbation more gradually, how does that affect the, the participants ability to actually store that pattern if they're re exposed to the same exact split? And so what we find here is that if we're comparing the savings or the kind of relearning period from the gradual group to just the naive first adaptation from the abrupt group, we don't see any difference at all. And so, in essence, what this is showing you is that if the person has learned the split belt pattern through a gradual induction of the perturbation, when they are re exposed to that same pattern, abruptly, they act like they've never really seen it before. And so there's none of the savings carrying over from that first learning bout to the second when the first learning bout is delivered gradually. And I thought, well, that's kind of strange. So there were some things here we needed to control for. Obviously, you can see that the amount of time spent at the full perturbation and the gradual is much shorter than the 10 minutes that the abrupt group received. So that was the second thing that we controlled for we delivered the perturbation again, gradually over nine minutes. And then once the perturbation was in full effect, uh we let it stay for 10 minutes and then again, wash the participants out and re expose them to the same 2 to 1 perturbation a second time. And here, what you can see is the result is actually kind of somewhere in the middle between the first two. They certainly don't relearn as fast here as the abrupt group did, but they certainly do relearn faster than the gradual group did. So this showed us that the kind of time that you have to practice that new walking pattern plays some role in your ability to save the pattern from one session to the next, but it doesn't necessarily tell the whole story. And so then we thought, OK, well, what's the effect of just simply having this kind of abrupt change in the environment when you experience this abrupt perturbation? It's a pretty kind of jarring sensation. So it does trigger your nervous system. Ok. This is definitely something different than what I've been doing before. I need to really learn how to walk in this new situation. So that, ok, well, if we limit the, the amount of practice or training that they have at the full 2 to 1 split, but still deliver them with an abrupt perturbation, how does that influence your ability to save what's learned from one session to the next? And strikingly, this looks a lot like what we saw up here with the kind of extended gradual group here. You can see that it's again, not as fast as what you might expect from a full 10 minute exposure to an abrupt preservation, but certainly faster than what we saw in the gradual group. So we thought here is we actually identified two different characteristics about how the training is delivered that facilitate the savings from one session to the next. It seems that you need an abrupt exposure to that uh perturbation and then you need some extended practice once you are at the kind of full uh full split and belt. And so if you have one or the other, you just simply have this kind of intermediate savings between the two and not kind of full savings like you see in the abrupt or no savings like you see in the gradual. But then this is, this is a little bit unsatisfying to us because this kind of abrupt, like what does this actually mean for this to be abrupt or what does the abrupt allow you to do that? The gradual doesn't allow you to do? And so one thing that I began to think about was maybe this abrupt change in your environment triggers something that allows you to actually really map out this new environment that you're in, really map out these new settings. So that in case you see them again, you know how to react to them in the future. And so we recreated the experiments that I just showed you. But instead of giving them a second exposure to the split belt condition, we gave them this recall task. And so what the recall task was was instead of walking uh with the belts out of fixed perturbation. The left belt moot, which was the slow belt in this first exposure here, the left belt moved at the same speed that it had during the first exposure. But now participants had a keypad that they could use to control the right belt. And so what we told them was after they had washed out here with the tide belts, excuse me, um use this pad to make the treadmill move as it had during the first exposure. So, in essence, we were testing whether or not they had kind of an explicit awareness of the environment that they had walked in thinking that maybe the gradual induction of the perturbation actually knocks down your aware. It makes you feel like the perturbation is actually smaller than it was. And then when you experience that abruptly, a second time, you thought it was bigger than the first one because you didn't have this accurate representation of how big the perturbation actually was during the first time because it had been induced gradually. And so when we did this task, what we found is that here, I'm showing you the right belt speed uh on the y axis. So the actual right belt speed that it had been moving during the first training session was 1.4 m per second. The two abrupt groups, the pink and the uh orange up here, they perform significantly more accurately than the two gradual groups. So even if you just had this slight kind of, I think this was a two minute exposure to the abrupt condition that was enough to trigger the your nervous system to really think. OK, this is different. I need to know what this new environment is. And by knowing that new environment, once you've walked in it for a while, then you have a walking pattern that you've paired with this environment that facilitates the savings. And so when we were doing this task, though, one thing that we wondered about was when people are clicking this button and changing the treadmill speed to tell us how they thought the treadmill was moving. Are they actually basing that on a knowledge of the environment or are they basing that on a knowledge of how they're moving? So, in other words, are they just increasing the speed of the treadmill until they feel like they're walking in the same way that they were earlier? Or do they actually know something explicit about the environment? And so to test this, we did this kind of what we call the opposite group, uh which was actually exposed with the um left belt going twice as fast as the right. And then what we told them was ok. You had, you were walking with the left belt, they knew that the, the left belt was faster than the right. Then we washed them out, give them the recall task. And here what we say is now we want you to make the right belt move like the left belt had previously. So here the key was, they can't just stop clicking the button once they feel like they're walking similarly to how they had during the initial training period, because they hadn't walked in that situation before. But if they actually had knowledge of the environment, they should be able to flip that easily. Because if you knew what the speed was, you just keep increasing until the speed matches. And that's actually exactly what we found. We found that even if they hadn't experienced that environment before, they knew enough about it that they could accurately recreate it with the other limb. So having this kind of abrupt change in your environment actually uh allows you to develop a nice knowledge of that environment. So that if you re encounter it in the future, you say, OK, I've done this before. I know how to walk in this situation and that facilitates your savings. But of course, the practice also plays a role because you need to have that pattern that's been reinforced for that specific environment. So the takeaway from this first experiment is simply that you need to have an accurate knowledge of the environment that you're walking in. And then you need to associate an appropriate walking pattern with that environment in order to facilitate this faster relearning from day to day. And so we don't suffer other kind of manipulations to see uh to test about how other factors might influence savings as well. And so this is kind of a combination with the first experiment that I showed you the first set of experiments that I showed you, we were curious about the role of feedback uh in, in savings. And so here we collect the two groups. Um these are similar to those first two groups that I showed you. Again, if the participants walk on the split belt treadmill with the visual feedback, they acquire that symmetric pattern faster than if they don't walk with the visual feedback, they have similar wash out. And then we wonder, OK, if we give them a second exposure and there's no visual feedback during the second exposure, will that kind of faster acquisition that they use during the first learning period carry over to the second? And what we found is that's actually not true. And so regardless of whether or not the participants uh learned with or without feedback during this initial exposure to the split belts, they don't relearn any faster the second time. So it doesn't seem like that again, you're this is becoming kind of a common theme if the participants use the feedback during that initial period and then you take it away, it doesn't really have any lasting effects once it's gone. Whereas the split belt pattern, obviously, they still show savings. So they still have this faster relearning. They've still learned something from the treadmill. But that feedback was just used again as kind of a conscious correction mechanism during initial learning. And then we ran a control group where we thought, OK, maybe this could, maybe they have similar savings here because they both got to the, they eventually got to the same amount of learning on day one. So we truncated this group to be about a minute long so that we chopped it off here where one group had significantly more uh symmetry in their gate pattern than the group without feedback. And what we found is even when we did that. So when it appeared, that one group had acquired a more symmetric pattern than the other group. Uh And again, this was driven by the same feedback that I showed you earlier. Once you wash them out, the savings again is very similar between the two groups. So the takeaway here is again, the feedback seems to facilitate this faster acquisition, but it doesn't really have any lasting effects on the learning once it's taken away. Uh And then we also did uh a separate experiment in our lab recently where we began to wonder, OK, why is this uh why does this knowledge of, of environment seem to be important or how is this savings being facilitated? Uh And we thought that looking at kids might actually be an interesting way to show kind of not, we can't get precise insight into where in the brain this is happening, but it might lead us in the right direction. And so we tested several age groups of Children in the same kind of two exposure learning paradigm. So they walk on the split belt on one day, they got washed out, they came back the next day. And what we find here is that kids of at least 12 years of age show relatively normal savings and these are all abrupt perturbations. So the blue is day two, the orange is day one. If you were at least 12 years old, you showed faster relearning when you came back the second day than you did the first day. However, this was not seen in the kids that were less than 12 years old. So in the kids, 6 to 11, we actually found that they didn't have savings day today. So when they came back on the second day, they basically walked on the treadmill as if day one had never existed. They didn't remember at some level of consciousness, um what they had learned from one day to the next. And so we're thinking, ok, maybe this might be influenced by some sort of higher order cognitive mechanism where you're actually trying to map out your environment. But something that kind of occurs on the scale of development and starts to mature around the age of 12. So again, the takeaway from this last uh group of studies was simply that this first one seems a little obvious, but I'll explain a little more in detail. Um We should, we should really consider everything that we're trying to do uh in our training paradigm based on what we would hope the patient can accomplish. I mean, obviously, if we have a stroke patient who can't walk on the split belt treadmill, unless we induce the split gradually, we need to induce it gradually. But then we need to be prepared to know that this may not carry over from one day to the next as well as if we're able to deliver it more abruptly. And so really tailoring the training paradigm to the goals of the training I think is a critical step for this type of research and actual clinical implementation. And again, not all learning is equal. And so just because it appears that you've learned to walk in the split belt environment, doesn't mean that that's going to necessarily carry over from one day to the next as well as if you've learned it in a different way. And then finally, as I mentioned on the previous slide, different facets of learning appear to develop at different time courses across adolescence. And then this is the uh the final kind of set of experiments that I'll talk about today is what we've been doing most recently. And so one question that I'm really interested in is is this one that's on the screen now? And so what I've been really thinking about is how do we deliver some kind of uh therapy or some kind of intervention such that the outcome or the desired walking pattern that we want the patient to achieve through the intervention actually has value to them. So of course, when we have a stroke patient that has uh asymmetric steps come into our lab, we want them to be able to walk more symmetrically. But I think it's really important that we begin to understand or begin to think about. Why would that patient want to walk more symmetrically? And so here, what I'm showing you is this is a group of seven stroke patients uh again, step lengthy symmetry on the y axis here and striding over on the x axis, they're walking for four minutes on our treadmill and the red in the middle is their preferred asymmetry. So if you don't tell them to do anything in particular, this is the amount of limp or the amount of asymmetry that they show while walking on the treadmill, then we give them the same kind of visual feedback that I showed on the previous slides and they actually have considerable flexibility in their gate symmetry And so if I tell them to walk less symmetrically than they do in their daily life, they're able to do that. And that's what I'm showing you here um in this green data. And then conversely if I tell them to walk more symmetrically, so they're walking with these little targets on the screen and I tell them to try to step into the same target as best they can. Uh they're also able to do that. And so then it a pretty obvious question that emerges out of this is why don't they do that? Why don't they walk more symmetrically if they have the capability of doing so? And so that's what this next set of experiences is kind of geared to address. And one thing that we think might be playing into that is the energetic cost of, of symmetric walking for for persons posts stroke. And so we know from many years of research, uh this is a paper about 60 years old by now. Um that most of the gate parameters that we select in daily life appear to be tuned to minimize the energetic cost of walking. So it's plotted here is velocity squared, walking, velocity squared along the X axis and then the energetic rate along the y axis. And so not surprisingly, the faster you walk, the heavier you breathe, the more calories you're using. But what's important to notice about this curve is that the Y axis is normalized to time. And so when we think about efficiency, we don't in, in, in a movement sense, usually we're considering efficiency per unit distance, not per unit time. When you think about your car, you don't think about how much the amount of gallons of gas you're burning, you think about the MPG that you're getting and our bodies work the same way. So if we now normalize this y axis to distance traveled, we always see this kind of characteristic U shape function where if you walk very slowly, it actually becomes a very costly way to walk. Uh, you're not breathing very heavily, you're not using very much, much oxygen, but you're also not traveling very far. And on the opposite end of the curve here, if you're walking very quickly, you're breathing heavily and you're moving quickly, but you're actually breathing heavier relative to what you could be doing at a different speed. And that's what we find is that people tend to walk at this speed that minimizes this cost function. In other words, people tend to walk at a speed that saves them energy per unit distance they're traveling. So they tend to walk at a speed that gives them the best MPG out of the oxygen that they're breathing in. And this is not specific to walking speed. Uh There's a variety of different studies that show the same effect across different gate parameters here. I'm showing you a step cadence So the the frequency of your stepping follows the same U shape function, your step width. So how far apart your feet are media laterally when your stepping follows the same U shape function, and then you also tend to begin to run at speeds at which it becomes more economical to run than walk. And interestingly, this isn't even a unique phenomenon to humans, but also elephants kind of lumber at a relatively energetically optimal rate of box turtles crawl at an energetically optimal rate. And even penguins which have a really strange um muscular skeletal structure, they also wadle that are relatively energetically. So this is seen to be something that's important that's conserved, not only across species, but also across a wide variety of different uh walking patterns that we can study in the lab, excuse me. And so what we wanted to know then, um is we get so kind of habituated to walking in these specific ways in daily life that will people actually change how they walk if we can design an intervention, such that a new way to walk, saves energy. And so that's gonna be the primary question that I'm going to be trying to answer uh on these next few slides. So to put it a little more clearly, uh this is a common model that we see in walking rehabilitation. If you test a patient group that tends to have a high energetic cost of walking, whether that's multiple sclerosis or stroke, even Parkinson's disease. The, the approach is commonly to try to understand how to change their movement to reduce that energetic cost of walking. But what we're doing is we're kind of flipping that around. We're saying if you can design your intervention such that the pattern that you want the patient to be walking with becomes more energetically efficient for them, maybe that can drive the patient to adopt that pattern on their own. And so that's what we'll be doing on the next few slides, both in healthy individuals first and then moving it into stroke patients. So again, you've seen this before, this is kind of our basic set up where they have the TV. Now we're also measuring the metabolic cost or the energy cost of their walking using this metabolic cart. And they're able instead of seeing simply step length or how far they placed 1 ft head of the other on the screen. Now, now we give the patients actually this online feedback so they can constantly see the position of their feet while they're walking on the tread. It gives them a little bit more control over the screen in front of them and will be doing in these experiments is we actually create this kind of custom treadmill controller that when it's engaged, the participants actually have control over the speed of the treadmill depending on how they walk on the screen. So they would learn a variety of different walking patterns using this visual feedback. And depending on which one they begin to walk with the treadmill can actually change speed. And I'll explain this more how these are set up on the next few slides. But this is just something to keep in mind as we're going forward is that we can actually begin to reshape this energetic landscape by having the person actually control the treadmill speed themselves by walking in different ways using the visual feedback. And so we had this uh the first study is in healthy individuals. So on day one, we brought them in and we simply just measured uh the energetic cost at a variety of walking speeds. And then we also had them walk using a variety of different walking asymmetries again using the visual feedback in front of them. So what you're seeing on the y axis here for speed is just simply the speed in meters per second ss indicates their self selected overground walking speed. And what you're seeing on the bottom here is the symmetry in the targets they step to with left minus, right. So if you see this 2 to 3 here, that indicates that their left foot step, two or three targets ahead of the right, each target is about 10 centimeters wide. So roughly uh 20 to 30 centimeters apart uh for the 2 to 3 condition and then the negative just indicates the opposite. So the right was stepping ahead of the left. And when we measure these things out, we can uh acquire this relatively characteristic relationship between walking speed and the energetic cost of the walking pattern on the Y axis. So again, very slow walking drives up the energetic cost is very high and then it tends to bottom out here somewhere around the self selected walking speed and then increases again with increases in speed. And then as you would probably also expect as the limp size increases, the energetic cost also increases. And again, these are all healthy individuals, they don't limp at baseline, they don't limp with their preferred walking. But when we induce that limp by having them walk with different stepping targets on the screen, we can measure that cost. And so again, when they limp with this kind of small limp, so they're walking with one leg stepping two or three targets ahead of the other. It increases by about 50 mL of oxygen per kilogram per kilometer walk. And then that doubles basically that difference doubles once they're able to actually, once they're generating these even more asymmetric steps, when we have them take the really big limps where one leg has to step at least four targets ahead of the other. But what's important about this slide is that we now have two different factors that we can use to manipulate the energetic cost of movement. We can manipulate speed and we can manipulate asymmetry to generate a different energetic landscape and that's what we do on day two. This looks a little bit confusing at first, but the gray lines indicate that the symmetry that the person had to walk with and the black lines indicate the speed. And so these are all prescribed by us. So basically what we do is we say if you want to walk with a speed of 0.5 m per second, then you can just walk with symmetry. And so you can probably see what's beginning to happen here. If they want to walk symmetrically, now they have to incur a high energetic penalty from walking very slow. And then if we move forward, if you want to walk at 1 m per second, then you walk with this right footstep and two or three targets ahead of the left and so on. And we devise these pairs such that the energetic landscape now looks like this. And so again, I'm showing you the energetic cost of walking on the Y axis and then the different combinations that we set up, we programmed into this controller on the X axis. And so the two important takeaways here are now walking symmetrically, which is the difference between 01. That means both targets are either stepping to the same target or one target apart that becomes a very costly way to walk. Because if you start walking like this, the treadmill slows down to the 0.5 m per second speed, which again, is a very costly speed to walk. At. Now, on the other side, if you want to walk at your self selected speed, now you have to walk with the right foots, stepping at least four targets ahead of the left. So that also becomes very costly because you incur a high penalty from limping on the treadmill. But then in the middle, if you select a speed that's slightly different than you're self selected, and you begin to limp a little bit more than you normally do within this landscape that we have designed these actually become the least energetically costly ways to walk. And so we thought, if we put somebody in this environment and they're actually trying to walk in the most energetically efficient way possible, they'll actually start walking either a little bit slower or a little bit faster than they're used to. And they'll limp a little bit more than they're used to in order to minimize that energetic cost. Because again, that's how we've set up this landscape. And so the question here is, will these healthy individuals actually begin to limp a little bit if limping saves them energy in this environment? And so here is, uh, Shane one more time doing this task. And so you'll see at the beginning, she is going to be hitting the same target. So she's just walking normally and the treadmill moves very slowly when she does that. But once she starts walking a little bit asymmetrically. So her right leg is stepping two targets ahead of the left. Once she does that a few times the treadmill will begin to speed up. And so she here she starts going a little bit faster and it doesn't happen immediately because we built a little bit of a buffer into it so that patients wouldn't be shifting the speed back and forth on a step by step basis, depending on which target they hit. And so after we gave these people the participants, these healthy young participants exposure to these different combinations, we let them have kind of their own exploration period for 10 minutes. After we had uh tested all these different conditions, they got to just play around with the controller and begin to experience how when I walk in these different ways, the the speed changes depending on how I'm walking on the screen. And then after they've done that for 10 minutes, we actually had them pick a pattern that they felt was most comfortable for them to walk in this environment. So we said, if you had to walk for five minutes using one of these combinations, which one would you pick. And so then again, this is the cost function. And when we plot out the proportions that the, the participants or the the conditions that the participants preferred, we saw that about 75% of the healthy young people that we tested began to actually limp in this environment where limping costs less energy than, than normal walking. Whereas only about 25% of them defaulted to this kind of habitual slow symmetric walking. And so this was exciting for us because it showed that these people are actually following this cost curve, they prefer to walk in the way that saves them energy, even if this way to walk is very strange and unusual to them. And we think maybe the reason that nobody picked this uh speed, even though it also is similarly low in cost to the speed that a lot of people pick is simply because this has a very uh high speed. And so if you have two different combinations here that give you basically the same cost per distance, you probably want to pick the one that costs you less in terms of cost per time. Because again, we had people walk for five minutes and not a set distance in this experiment. And so if it's costing you a lot of energy per time, uh then you probably will, will switch to the other one that doesn't cost cost quite so much per time, which is again, the slower speed doesn't cost quite as much per time. And then we ask, OK, well, in, in the real world, we tend to walk for a set distance, not necessarily for a set amount of time. And so if we do this experiment again, but instead of giving the participants, the test period for five minutes, we have them walk for a kilometer, how might this influence their choices? And so here they're actually able to monitor the distance they're walking while they're doing the same set up that we had shown on the previous slide. And so again, we've also sped this one up to see if, if speed might have some effect because on that last slide, as you saw, nobody picked the 2.0 m per second condition. So we switched these two speeds such that now the energetically optimal condition required a slight limp and a speed a little bit faster than the participants walked at their self selected pace. But again, you'll see the curve is is very similar. So in this environment, a walking combination that required a slight limp and a speed slightly different than preferred was still the energetically most efficient way to walk. And so when we had people again do the same task only with the distance task rather than with the time task, that's the result from the previous experiment. And then we see the result is very similar when they do the distance task. So whether we're testing part time or per distance, people tend to converge on this energetically optimal solution, even if that solution is a little bit asymmetric and at a different speed than the participants prefer to walk at. And so then the big question that we're ultimately trying to get at with this study is then if we can actually make healthy participants begin to limp by making limping cost less energy than symmetric walking. Can we do the opposite in the stroke patients? So if we create this artificial environment in our lab where walking symmetrically actually begins to save the stroke patients energy, will they follow this kind of energetic landscape and also begin to walk more symmetrically. And so we tested them in a very similar protocol to what I just showed you for the healthy individuals. We use less speeds and we use less asymmetries just to minimize fatigue in our in our participants. So we brought them in on day one, we tested them at a slow speed, self selected speed and a fast speed. The fast speed and the slow speed were both functions of the person's self selected speed because we didn't want to push people too hard. So they couldn't actually produce the faster pace. And then the symmetries vary depending on their preferred. And then we had them do one again, that was less symmetric than they preferred and one that was more symmetric than they preferred. And so I mapping these out again on day one. sorry, you can't see that that's just cost again. On the, on the y axis, the person's post stroke, just like the healthy individuals. We saw that the energetic cost of their walking increased dramatically once they began to walk at the slow speed. And then decreased when they walked at the self selected speed or fast speed. But the interesting thing about our stroke, participants that we studied was when we manipulate the step length as symmetry, we don't see that same symmetry effect on the energetic cost of the movement. So no matter whether they walked less symmetrically than they do in their daily life, whether they walked, walked with their preferred symmetry or whether they walked more symmetrically, the energetic cost of the walking pattern was very similar across all three conditions. And so, in other words, persons post stroke, or at least the the subgroup that we tested, they don't receive that same energetic benefit from walking symmetrically um that healthy individuals do. And so we wondered is if we enforce that, then will they actually begin? If we create a situation where symmetric walking does save them energy, will they then begin to walk uh more symmetrically to, to actually be able to save themselves energy? And so that's what we did here. We created another controller where if the participant um walks with their preferred symmetry here in the dark green, the treadmill slows down to the slow pace. If they walk um with their more symmetric walking pattern, then they can walk at their self selected speed. And if they walk with a less symmetric walking pattern, then they walk at the fast speed. And so when we enforce this controller, the energetic landscape looks something like this where now, their preferred asymmetry is paired at the slow speed. So in effect, their preferred asymmetry becomes very costly. Whereas walking either less symmetrically at the fast speed or more symmetrically at the self selected speed, actually begin to save them energy. So if they want to save energy, they have to manipulate their step blank asymmetry in some way or another. Or if they're comfortable just walking with their preferred asymmetry, the treadmill is going to move very slowly. And so when we actually exposed them to this environment, and again, mapped out the proportion of participants picking each condition, we found that actually none of them preferred to keep walking with their normal everyday asymmetry. And they all actually changed their asymmetry in order to change the speed of the treadmill. And so we're excited by this because we think it shows that yeah, there's nothing really tying these people necessarily to their daily kind of preferred limping pattern. Um But if we can do something that kind of encourages them or provides some kind of benefit uh for them to change their asymmetry, then they'll actually begin to do this just as we're seeing in, in uh in this study here. The the of course, the million dollar question is how do you move this out of the lab and begin to do it in a clinical setting? How do you make symmetric walking costless energy for somebody in their daily life? But that's uh an area that we're actually moving into. Now with a graduate student in the lab, she's trying to really understand how the muscle patterns that people walk with after a stroke contribute to this higher energetic cost that they're incurring when they try to walk symmetrically. And hopefully, by understanding really um the muscles involved here, we can target those specifically on a patient by patient basis to hopefully try to reduce that energetic cost of walking symmetrically to again, encourage these patients to, to walk a little bit more even with each side. So the takeaways from this study are uh again, I'll just put them out quickly because we've covered most of them. One that people will walk in unusual ways to save energy if that, if the situation uh requires that two, that energy costs can actually be used to drive people to change their walking patterns. And then three that persons post stroke may actually walk more symmetrically if we can develop some kind of intervention uh that makes symmetric walking cost less energy for them. And so moving forward, I, I touched on this a little bit uh a second ago, but we just really want to understand how we can do this in daily life so that we can actually uh drive changes in these patients. So uh before I stop, I'll just acknowledge Amy and Chris Haass. These are my two mentors. So most of the things that you see up here are uh done. In Amy's lab. Uh But Chris continues to be a big influence on me. And then also thank a lot of our lab members who were involved in a lot of this research and in helping collect patients and uh and analyze data and things like that. And then also my funding from NIH. So uh thanks everybody for listening and I'll put a summary slide up here and be happy to uh answer any questions, small location, know providing them more efficient, right? A box to metric other than treating them with the right. Yeah. Yeah. Talking right. And so that's a question that I get every time I present this is if they are optimal at their current asymmetry, why do you want to make them more symmetric? And so if you were, if you were to take this curve that you see here on the bottom, right? And you plotted this against a healthy person walking at the same levels of asymmetry, you would find that these two points are relatively similar, but that this point is significantly reduced. Uh So when we, when you are Iowa symmetrically that provides us some energetic benefits. So we think that there is room for improvement there that we could theoretically through some kind of. And again, we haven't shown we haven't been able to do this yet. Uh But we think that there's room here to be able to knock that down. So one we think that there's, if you could make them walk more symmetrically, you could do that in a way that also reduces the energetic cost associated with that. But then we also know there are a variety of biomechanical reasons to walk symmetrically, it's more stable, it leads to less uh stress on joints. Um And so we, and it's also just kind of a, it's stable in a, in a kind of biomechanical sense, but then it's also stable in kind of a control sense too. It makes you less susceptible to perturbations and things like that. And so there's these biomechanical reasons, but the main thing is we think that there's kind of ample room here to play with that. And if we can knock it down, they would not only experience the benefits of these biomechanical benefits they're talking about, but they would actually probably cost them less energy to. So when you're, when you're trying to get stroke patients to walk symmetric, do you have any information on that? Right. So we actually uh somebody in our lab is testing that right now. Uh Merritt is running some of those experiments. Uh Last time I saw that the the data was, it looks similar to the healthy where you can see the same kind of after effects. So stroke patients can benefit if you're just trying to get them to carry it over into their daily life, they can benefit from it. We haven't done to my, I don't think merit did the the savings. So will one training session kind of bleed over into the next as well if you do it gradually versus abruptly. But we can uh so part of this stuff that this energetic stuff is um motivated by the kind of question you're asking because one of the frustrating things with the stroke patients is once we put them on the split belt treadmill, yeah, we can make them limp or we can make that limp go away nicely for a few strides. But eventually it comes back. And so we really wanted to know like why does it keep coming back? Like what is it about this limping pattern that is really desirable for these patients? And so we're thinking this might actually kind of provide some information with that the way they make, they make him by moving here, right? Uh So that's a great question and we're actually writing that up now. Uh But I didn't present on it. It's kind of this paradoxical thing where I'm just showing you step length asymmetry. So it's a very, it's like a very endpoint measurement, very discrete measurement. What they actually do is they take as you probably predict they take advantage of what they they can do. So they in a way they walk more, they take more symmetric steps by walking more asymmetrically if that makes sense. So they can place their feet symmetrically. But to do so maybe they have to really like crank at the hip and they can't actually get this maybe knee extension that you're hoping for or something like that. So they actually, it depends on the deficit is the, the short answer. But we're also like if you walked into the lab while I was doing this test and they were doing the more symmetric step length condition, the person doesn't, you wouldn't say that person's walking symmetrically, they might be stepping symmetrically, but their kinematics are not symmetric. And so we're trying to understand what about those kinematics? We can change to reduce that energetic cost of the symmetric walking. Um But they accomplish it now by taking advantage of what they still are able to do. So if they can flex at the hip and that allows them to get the foot forward. Um And they do that. But when we tested these people, we had six, I think a majority of the people that we tested um took shorter leg steps. So they tended to have more um kind of forward progression problems with the leg. And then the other person took a or the minority of people took a longer step um with the non oh sorry with the pre leg because they can't push off hard enough with the, with the pre leg. And so we have to be careful about like, you know, I just don't want to answer your question by saying they all do one thing because it really depends on the deficit that they have when they come into the lab. Um person goes into a 2 to 1 split velocity and they start becoming very symmetric, habituated to come back to what looks like more symmetrical matter what's happening with their energy efficiency. This is they, they may look more because they're the motor walking more now attainable, but they're actually on that fast leg taking much bigger. So, are they more energy efficient or not when they? Yeah, they are. And uh a post doc that was in Amy's lab before me. He actually did that experiment. And it's, it's kind of crazy like the, if you plot out the asymmetry versus the energetic cost on a time series, they're very well correlated. And so that's led some people to think that you actually adapt in order to save energy. I'm not in that camp but some people are. Uh and, but yeah, the energetic cost decreases as you adapt to the treadmill. Uh What would happen if, I mean, if there's a way to, if they normalize it more so they're more energy efficient, but one could block there being more of the levels or whatever. Right. Yeah. So I uh one thing that I've done is actually, and another graduate student in the lab has done is actually the opposite where we, we don't block them from. So the, the thing, the reason the thing about the adaptation is that it doesn't really rely on your movement, uh movement is kind of an outcome of it. But it relies on you experiencing this, we call it sensory prediction error. So the difference between what you expect and what you get. And so if we have somebody on the treadmill, And Andy long did this experiment in our lab, he's a graduate student, he, he gave visual feedback and basically said, no matter what the treadmill does step in these locations. And so he is in, in effect, pre prevented them from expressing the motor output of the adaptation. And so he didn't do what you were suggesting, but he did the opposite where he kept the energy that cost high of walking similarly. And then the second you take that feedback away, they start walking in the splitt be pattern as if they had learned it all along. So that's, that's why I said that I don't really subscribe to this adaptation is driven by the energetic cost of the movement because there you're, you're blocking the behavior and you still adapt. And so yeah, you wanna wait, wait them. Um Why don't they? Sorry. Uh It'll be this probation General Symmetry if they're able to do it and they can use them. Hm. Yeah, I see like we not in this task just walking around because it doesn't, yeah, you could flip it. So I mean, I sell it, I sell it as there's no energetic benefit to be gained by walking symmetrically but there's also no energetic to be benefit to be gained by walking with your preferred asymmetry. And so that's a good question. And that's something that we've had to address and in papers and grant reviews is uh this is I'm not saying that just to clarify, like, I don't, I don't think that energetics are the only thing that matters to a patient when they're walking. Uh If they gain stability within their current uh set of neuromuscular constraints by walking with a little bit of asymmetry. Of course, I don't think somebody if somebody can. Yeah, I was. But uh if so if somebody can basically, I mean, nobody's gonna save, fall down to save energy. I don't think, I think the people will maintain stability to cost at the cost of more energy. And I think that's probably what's happening uh in the preferred setting. But uh sorry about this. That was, thank you. Um Is that you just so they have arrived to this place, but it's become they will have to learn or whatever they have a all of this. So thank you. You uh from after which obviously they were addressing other problems right early on. They may know you v or Yeah, so um they are right to this place and is quite the fact that think they only are able to change that. Yeah, we, I mean, full disclosure, we won't know the answer to this question unless we can actually reshape this landscape in someone's daily life. Uh because even when they do this, it's a very cognitively driven task. And so they have to think about what they're doing, think about where they're stepping, it's not a natural type environment. So I think one thing that I, I really, if I could design any experiment and have it be done in like a day and then have the result I would be interested in knowing, taking an intervention that we know has a positive outcome, especially something with related symmetry. I I think that's easiest in this framework to study. So having a long term split belt training study and you would measure the energetic cost at a variety of speeds and a variety of asymmetries before and after the intervention. And then if you see that this curve has, we know that people will walk more symmetrically after the intervention. So are they walking more symmetrically because something about the intervention trained them to walk symmetrically or are they walking more symmetrically because the intervention has made it more energetically efficient to walk more symmetrically? And so that data obviously takes a long time to get. But um I think that is really the only way to really kind of get into this without doing these kind of lab based manipulations. If you added the condition of partial weight to check out the the staying asymmetric because they're more uh stable, take that out. If possible, how would that change? Great. That's a good question. We don't know. We haven't, we haven't, uh we talk about it a lot but we haven't actually done it looking at the interplay between kind of stability and, and, uh we might even be a little bit simplistic. And I mean, to me, those are the two big, big ones. Uh, and it might be, I think I'm probably being a little bit simplistic and focusing so much on those two. Um But we haven't done that yet and we thought about it and another thing that we haven't done um that this is a little bit limited by is right now, we always um we, we create symmetry being good by analyzing everything else. It doesn't actually have added benefits. Uh And so now we have a new term that we can do things like. Uh and that's because it's very hard to reduce the energetic cost. It's much easier to drive up the cost of movement than it is to reduce it. But we just have a new treadmill now in the lab that you can actually walk downhill. So downhill walking, as you can imagine is, is less costly than, than flat walking. So you could actually, if people walk more symmetrically allow them to walk up a little bit of a downhill slope, but now you can, there's no benefit to it. And it's not just here, we're just penalizing their preferred symmetry by linking that with a slow pace. Great. Thank you. All right, thanks everybody. Created by Related Presenters Ryan Roemmich, PhD Assistant Professor of Physical Medicine and Rehabilitation View full profile
Facebook
Twitter
YouTube
LinkedIn
Instagram
E-News
BROADCASTMED