Adaptation to constant-magnitude assistive forces: kinematic and neural correlates

Abstract

In many robot-assisted rehabilitation and motor skill learning applications, robots generate forces that facilitate movement performance. While there is some evidence that assistance is beneficial, the underlying mechanisms of action are largely unknown, and it is unclear what force patterns are more effective. Here, we investigate how reaching movements (and their neural correlates) are altered by 'assistive' forces. Subjects performed center-out reaching movements, under the influence of a robot-generated force, constant in magnitude and always directed toward the target. The experimental protocol included three phases: (1) baseline (no forces), (2) force field (with two different force levels, 3 N and 6 N, applied in random order), and (3) after-effect (no forces). EEG activity was recorded from motor and frontal cortical areas. In both movement kinematics and EEG activity, we looked at the effects of forces, of adaptation to such forces and at the aftereffects of such adaptation. Assistive forces initially induced a degraded performance and in general alterations in movement kinematics. However, subjects quickly adapted to the perturbation by improving their performance. With regard to EEG activity, we found (1) an increased beta band synchronization just before movements and an alpha band synchronization in the ipsilateral hemisphere, both proportional to force magnitude; (2) a gradual decrease in alpha band synchronization with practice in the contralateral hemisphere; (3) an increase in theta band synchronization in the later stage of the force epochs; and (4) an ipsilateral to contralateral shift (from baseline to aftereffect) of theta band synchronization. These results point to the need for a careful design of assistive forces to effectively facilitate motor performance and motor learning. Moreover, EEG signals exhibit distinct features related to force and adaptation. Therefore, at least in principle, the latter might be used to monitor the learning process and/or to regulate the amount of assistance.

Towards a Biomarker of Motor Adaptation: Integration of Kinematic and Neural Factors

“…Modifications in brain response were found at both planning and execution level during motion. During the planning of the motor scheme, the index ERDcueBeta and the related index ERSERDcueBeta revealed significant diversity between neural responses to A and B conditions, thus suggesting the presence of some specificity of planning in the beta band with respect to the environmental characteristics; this result seems to be consistent with previous studies [18]. On the other hand, the executive part of movement seems to be the one mostly involved in modifications: the two related indexes ERStargetBeta and ERDERStargetBeta revealed changes dependent on phase, as some specificity was put into evidence for the different stages of movement refinement.…”

“…Although the hodology (i.e. description of neural pathways) goes beyond the purpose of the present paper, we need to mention that the primary motor cortex (BA 4) also plays a primary role in the delivery of the motor command to the muscles, thus allowing a refined optimization of joint angles and the subtle tuning of torques [3], [18]. Electrically, a correlate is observed in the beta frequency band (18)(19)(20)(21)(22)(23)(24)(25), where some power decrease is found.…”

Robot assistance of motor learning: A neuro-cognitive perspective

“…For example, it creates a new dynamic environment in which movements are produced and to which humans adapt (cf. Emken et al, 2007a;Novakovic and Sanguineti, 2011;Richter et al, 2004). Robot assistance can become "normal" and motor learning specific to the particular dynamic environment.…”

Ipsilateral EEG mu rhythm reflects the excitability of uncrossed pathways projecting to shoulder muscles

“…In experiment 1, the regions of interest used for calculation of the LI were visually determined for each hemisphere. The reason why we applied a visual selection of LI was because i) visual inspection is an established method to evaluate EEG features such as artifacts [ 42 , 43 ], time-frequency map [ 44 ], and cortical topography [ 45 ]; ii) it helps us to avoid detecting regions which is not physiologically plausible, when an ERD accidently took large values at these regions due to electrooculogram and other noises. Based on visual inspection, we selected the areas where eye-blink noise and contact to the TMS coil or headrest were smallest, and were closest to the left primary sensorimotor cortex.…”