Multijoint arm stiffness during movements following stroke: implications for robot therapy

Abstract

Impaired arm movements in stroke appear as a set of stereotypical kinematic patterns, characterized by abnormal joint coupling, which have a direct consequence on arm mechanics and can be quantified by the net arm stiffness at the hand. The current available measures of arm stiffness during functional tasks have limited clinical use, since they require several repetitions of the same test movement in many directions. Such procedure is difficult to obtain in stroke survivors who have lower fatigue threshold and increased variability compared to unimpaired individuals. The present study proposes a novel, fast quantitative measure of arm stiffness during movements by means of a Time-Frequency technique and the use of a reassigned spectrogram, applied on a trial-by-trial basis with a single perturbation. We tested the technique feasibility during robot mediated therapy, where a robot helped stroke survivors to regain arm mobility by providing assistive forces during a hitting task to 13 targets covering the entire reachable workspace. The endpoint stiffness of the paretic arm was estimated at the end of each hitting movements by suddenly switching of the assistive forces and observing the ensuing recoil movements. In addition, we considered how assistive forces influence stiffness. This method will provide therapists with improved tools to target the treatment to the individual's specific impairment and to verify the effects of the proposed exercises.

Figure 1

Experimental Setup

Figure 2

Force Field, Structure of the basic trial.

Figure 3

a) Exemple of oscillation after negative force step. b) STFT, c) Reassigned Spectrograms, and d) time-varing stiffness, Subject S7, Force 5N, central target of circle B in Fig.1, last session.

Figure 4

Stiffness and damping estimations for S2 (Ash. 1+) for each target of the workspace, with 3 levels of aiding force. On top, velocity profiles to reach two targets (yellow and black squares, respectively)

Figure 5

Comparison between the stiffness metrics between the first and the last session using the same level of force. Force level was chosen as the minimal to initiate the movement. For significant statistical differences in the metrics populations see Table III.

Figure 6

Relarionship between Ashworth score and the determinants of the stiffness matrices averaged throughout the workspace. * indicates groups that are statistically different.