A closed-loop BCI system for rehabilitation of the hemiplegic upper-limb: A performance study of the systems ability to detect intention of movement

Abstract

Rehabilitation of the upper-limb by means of a brain computer interface (BCI) is based on detecting an “intention to act” in the brain and transforming this intention into actual movement by using a motorized orthesis. The visual and proprioceptive feedback that results from the systems’ generated movement following the subjects’ captured intention, is thought to close the sensori-motor loop. Closing the loop is a key feature of BCI-systems that is thought to induce cerebral plasticity and thereby facilitating rehabilitation. Ten chronic patients post-stroke with a severe upper-limb motor deficit and 10 healthy controls were included in the study. To mobilize wrist and finger movement, we used a pneumatic orthesis that was controlled by the BCI during a wrist flexion/extension task. The main objective was to evaluate the systems’ CPS capacity to “classify” the observed brain activation as either movement or rest. Activation was registered with a 27-channel g tec EEG during two series of 40 randomized trials (20 rest, 20 movements). The sample frequency was 256 Hz. The BCI classifier was capable to distinguish between movements and rest with comparable performance for patients and controls and equal levels of true positives and negatives (78% vs 77% for patients and 81% vs 80% for controls). The CSP classifier used a wide 8–30 Hz-frequency band filter to evaluate EEG activity. By cutting this large band into smaller sub-units, we showed that the systems’ performance could be optimized by using subject-specific smaller frequency bands. The pneumatic orthesis should be improved. The present closed-loop BCI-system for rehabilitation of the paretic upper-limb was well able to recognize intention of movement in the brain. Its practical interest for rehabilitation post-stroke needs to be confirmed by a longitudinal efficiency study.