Exploiting Interlimb Coupling to Investigate Upper-Extremity Bimanual Loading in
Robot-Assisted Neuro-Rehabilitation on Subjects with Stroke
Impaired arm function is one of the most common outcomes for the estimated 700,000 stroke survivors in the US each year. The arm contralateral to the damaged hemisphere is often weak, while the opposite arm is usually intact functionally. Arm weakness in muscles contralateral to the stroke is caused predominantly by the inability to activate the corticospinal pathways needed to activate agonist muscles. Training methods are needed to stimulate these pathways to promote neurorehabilitation. One approach that has received considerable attention is the use of bilateral training. It is known that representations of specific muscles are present in the contralateral motor cortex, and that homologous regions of motor cortex in the two hemispheres are connected through the corpus callosum. Therefore, bilateral activity may result in enhanced activation of the cortical representations of weak muscles through activation of callosal pathways from the undamaged to the damaged hemisphere. In this dissertation, we performed experiments to investigate the optimal bilateral training parameters. Loading of the non-paretic limb during bilateral symmetric movements may enhance the interlimb coupling effect by increasing activity in the pathway from the undamaged hemisphere to the damaged cortex. However, prior studies in stroke survivors have found contradictory results. Different loading profiles and experimental conditions may explain these contradictory findings. To resolve this controversy, we have performed the first study that used robotic methods to systematically evaluate the effects of different loading profiles and amplitudes applied to the non paretic limb during bilateral symmetric elbow extension. In order for the robot to accurately load the non-paretic limb, we faced the technical hurdle of compensating for the intrinsic dynamics of the robot (MIT-MANUS). First, we developed and compared two inertia compensation algorithms for the robot. One of the methods used a novel algorithm for digital differentiation of the encoder signals from the robot. This new method reduced the robot intrinsic impedance up to 64%, and tangential force anisotropy was reduced by 74%. We then developed robot algorithms to provide three different loading profiles: inertial, constant and spring resistance. Second, we performed an experiment with post-stroke survivors that examined the effects on the kinematics of paretic limb elbow extension from loading of the non-paretic limb. We found that all of the bilateral movements were slower than unilateral movements, probably because of the constraint placed on the subjects to move the two arms in synchrony (verbal instruction). Increasing load level in the non-paretic limb improved speed, acceleration, EMG in the paretic limb within resistive loading under spring and constant force patterns. The constant loading appears to be most effective of the bimanual conditions, and inertial loading was the least effective. Increasing inertial loading actually decreased the speed of the paretic limb. In the previous study, the two limbs were coupled by the instruction to move the two limbs in synchrony. Data from this study suggested that paretic limb speed was highest in trials where the two limbs were highly synchronized. In a follow on study in healthy controls, we studied different methods to enforce greater synchrony between the two limbs. An EMG coherence method was used to assess the degree of interlimb coupling as a function of different coupling methods. To simplify the experiment, we studied isometric elbow extension and flexion torque generation instead of movement. During bilateral elbow torque generation, three types of coupling methods were tested: low difficulty visual coupling, high difficulty visual coupling, and haptic coupling through a mechanical apparatus. For elbow extension, 8-51 Hz coherence was higher in the haptic coupling condition compared to the two visual coupling conditions. The coherence was largest with agonist muscle pairs during the ramp phase of torque generation. No difference in coherence was seen across test conditions for elbow flexors. To our knowledge, this was the first study to use EMG coherence to evaluate bilateral coupling. In conclusion, we discuss the implications of these studies, and make suggestions for future work. The novel aspects of this work relate to the development of an active inertia compensation algorithm for the MIT-MANUS robot, first systematic evaluation of bilateral interlimb coupling in stroke survivors using robotic loading of the non-paretic limb, and first study to use EMG coherence to evaluate the degree of interlimb coupling as function of different coupling methods.
StatsViewed 63 times
Downloaded 23 times