Identify principles underlying the generalization of motor learning for neurorehabilitation
Perception studies towards improving gait neurorehabilitation
One of the major issues for rehabilitation robotics is that motor improvements are observed with the training device, but not maintained in the patient’s daily life. One of our projects towards addressing this issue is focused on quantifying sensation. This is important since sensory information plays a critical role for planning, executing, and corrective movements. The challenge with measuring sensation in people is that it is an internal variable; therefore, we use mathematical tools and perception experiments to estimate what individuals feel. There is evidence that stroke survivors have difficulty perceiving their asymmetric movement, and our studies will help us characterize this deficit and indicate if split-belt walking - in which the legs move at different speeds - can correct it. Importantly, as part of these studies, we are using a human-in-the-loop method to manipulate how people perceive their movement and create the illusion of error-free performance during split-belt walking. This strategy makes people believe that they’re walking normally even though their movements are subconsciously changing in response to speed differences between their legs. The overall purpose of this work is to design strategies to enhance the generalization of movements from devices like treadmills and exoskeletons to daily activities.
The role of naturalistic conditions on the generalization of motor learning
A central interest of our research has been the study of the motor system’s ability to generalize movements from trained to untrained.–for example, from walking on a treadmill to walking overground. Prof. Torres-Oviedo initiated this work during her postdoctoral experience (Torres-Oviedo and Bastian JNP, 2012; Torres-Oviedo and Bastian. JNeurosci, 2010). This has lead to our current research on principles mediating the generalization of learning. We are particularly interested to study locomotor adaptation with portable devices, which opens the exciting possibility to study motor learning under more realistic situations and potentially correct subjects gait more effectively.
Computations underlying the generalization of motor learning in locomotion
We investigate how and why motor errors (i.e., unexpected movements) have a strong effect on the generalization of recalibrated motor commands across walking conditions. For this, we are developing mathematical models showing that 1) people develop motor memories associated with specific perturbation directions and that 2) errors upon introduction or removal of a new environment enable the transition between motor memories appropriate to the context at hand. Our mathematical framework is enabling us to gain insights into savings and interference phenomena reported in the existing literature. This mathematical model has also been useful at making predictions about specific effects of error on the generalization that we are testing experimentally. This body of work helps us to address clinical problems through a combination of mathematical tools and experiments.
Cognitive aspects regulating motor generalization
There is growing evidence indicating that cognitive and motor processes interact in motor learning, but little is known about their relation in action selection and its changes with healthy aging. Our studies have demonstrated that action selection declines with healthy aging (Sombric et al., Front. Aging Neurosci. 2017) –that is, older adults have greater difficulty disengaging the treadmill-specific pattern when walking over ground than younger individuals. Interestingly, older adults also have difficulty switching decisions in cognitive tasks and we found that switching in cognitive and motor domains were associated. Moreover, we found that attention during adaptation also regulates the generalization of movements (Mariscal et al., bioRixv 2018). This line of research is exciting because there is little knowledge about the influence of cognitive processes in locomotor learning, which could be harnessed for training patients more effectively in the clinical setting.
Characterize motor learning in locomotion when humans experience new walking conditions
Studies on muscle activity bridging the knowledge gap relating between biomechanical outputs and neural control mechanisms
Most recently, my group developed analytical methods to quantify learning ability through the analysis of muscle activity and revealed age-related decline in motor learning not detectable from kinematics alone (Iturralde and Torres-Oviedo, eNeuro 2019), which is the predominant method in the field. This was important since muscle (feedback) responses upon environmental transitions in walking were thought to simply indicate a sudden change in the environment independently from sensorimotor recalibration. However, we show that the structure of this feedback activity is indicative of the motor system’s adapted state when walking in a novel environment. We also find that this recalibration process does not require intact cerebral structures even if feedback responses are impaired post-stroke (deKam et al.,NNR, 2018) and cerebral damage limits the changes in muscle activity in the altered environment. Taken together, our results suggest that the sensorimotor recalibration to update motor commands and the execution of those commands are partially dissociable processes. The execution of motor commands requires intact cerebral structures, whereas their recalibration does not (deKam et al., bioRixv 2019). From a clinical perspective, our results are interesting because they revealed awry muscle activity patterns that could serve as clinical targets. In addition, our findings suggested that sensorimotor recalibration post-stroke could be exploited to induce gait rehabilitation. This work is intended to bridge the gap between biomechanical outputs and neural control mechanisms.
Computations underlying sensorimotor adaptation in locomotion to understand the retention of motor memories for Neurorehabilitation
We are interested in developing mathematical models to understand sensorimotor recalibration. Our group has implemented a linear time-invariant state-space model to describe the step-by-step evolution of high-dimensional motor patterns in walking (Iturralde and Torres-Oviedo, presented at plenary lecture in the Advances in Motor Learning and Motor Control meeting). Most mathematical descriptions of adaptation dynamics have been done of unidimensional variables, such as scalar measures of force or movement. In contrast, our multivariate description of behavior has allowed us to systematically characterize the evolution of activity across multiple muscles. Our mathematical framework also enabled us to investigate the effect of “breaks” (i.e., time periods without movement) on motor memories developed by individuals with and without cerebral lesions. This line of research is relevant for understanding the retention of motor memories, which is critical for neurorehabilitation.
Interdependence in the adaptation of biomechanical variables to augment motor learning in stroke survivors
Walking is a high-dimensional behavior, challenging the identification of a single “teaching” signal driving locomotor adaptation. As a result, there has been little progress on identifying ‘rules’ governing motor learning in locomotion. We work towards this goal by identifying control variables that predict human behavior when experiencing new walking conditions, such as split-belt walking in which the legs move at different speeds. For example, my group has shown that the forces generated to propel one’s body forward constitute an important control variable regulating the adaptation of movements (Sombric et al., Front. Physiol. 2019), at the expense of step length asymmetry. This was a common principle underlying motor adaptation in healthy individuals and those with unilateral cerebral lesions (Sombric and Torres-Oviedo, bioRixv 2019), challenging the prominent view that step length symmetry is tightly controlled during bipedal locomotion. We have also found control variables in space and time that distinctively contribute to the adaptation of locomotor movements (Malone et al., JNP 2012; Gonzalez-Rubio et al., Front. Hum Neurosci. 2019) and we have demonstrated that cerebral lesions alter these control variables differently (deKam et al., bioRixv 2019). This research has an impact in rehabilitation as it provides insight into biomechanical variables that can be manipulated to augment motor learning in patients with unilateral cerebral lesions, such as stroke survivors.