A novel virtual reality based finger movement training system to investigate mechanisms of training induced plasticity

An efficient way to elicit practice-dependent plasticity is repetition of elementary finger movements. The precise role of training kinematics has not been established because appropriate tools allowing the manipulation of training movement kinematics are lacking.

For this purpose, a virtual reality based finger movement training system was developed.

The system consists of a sensor glove, a database system, and training software. During a movement trial, the subject receives online performance feedback (three different visual feedback modes) on the spatiotemporal imitation quality of a previously recorded target movement.

In a proof-of-principle study, finger movements evoked by TMS over the left primary motor cortex (intensity 1.3xRMT, 120 stimuli) were recorded with a sensor glove (Wii-Glove, 10 sensors, right hand) before and after a 30 minute training session (divided in 6 blocks) in 21 healthy volunteers (9 f, age 27.3±6.0 y). Training consisted of complex movements involving synergistic flexions of the right thumb and index finger. A training block contained 120 trials. Every 10 trials off-line performance feedback was given and the visualization mode was cyclically altered to maximize attention to the task.

Training performance was quantified as a correlation coefficient between the conducted movements and the target movement. Using principal component analysis, TMS-evoked movements were described as a small set of kinematic synergies (TMS-(4)PCs). The quality of reconstruction of the target movement by linear combinations of the synergies served as a measure of training induced plasticity.

Performance of training movements improved during the training (1st block, r=0.52±0.14; 6th block r=0.70±0.10; p<0.001), indicating motor learning. The correlation coefficient between movements reconstructed from TMS-(4)PCs and the target movement increased after training (pretraining, r=0.65±0.20; posttraining, r=0.74±0.18; p=0.015). The gain of reconstruction quality correlated (r=0.62, p=0.002) with the mean training performance.

These findings demonstrate the feasibility of a new virtual training system allowing tight kinematic control of training movements. In addition, they suggest that the motor cortex builds specific memory traces of recently practiced movements of even considerable complexity. Finally, they indicate that the spatiotemporal consistency of the trained movements is a major determinant of plasticity.

Supported by DFG Cl 95/8–1.