Non-invasive Brain-computer Interfaces for Control of Assistive Devices

Summary

Injuries affecting the central nervous system may disrupt the cortical pathways to muscles causing loss of motor control. Nevertheless, the brain still exhibits sensorimotor rhythms (SMRs) during movement intents or motor imagery (MI), which is the mental rehearsal of the kinesthetics of a movement without actually performing it. Brain-computer interfaces (BCIs) can decode SMRs to control assistive devices and promote functional recovery. Despite rapid advancements in non-invasive BCI systems based on EEG, two persistent challenges remain: First, the instability of SMR patterns due to the non-stationarity of neural signals, which may significantly degrade BCI performance over days and hamper the effectiveness of BCI-based rehabilitation. Second, differentiating MI patterns corresponding to fine hand movements of the same limb is still difficult due to the low spatial resolution of EEG. To address the first challenge, subjects usually learn to elicit reliable SMR and improve BCI control through longitudinal training, so a fundamental question is how to accelerate subject training building upon the SMR neurophysiology. In this study, the investigators hypothesize that conditioning the brain with transcutaneous electrical spinal stimulation, which reportedly induces cortical inhibition, would constrain the neural dynamics and promote focal and strong SMR modulations in subsequent MI-based BCI training sessions - leading to accelerated BCI training. To address the second challenge, the investigators hypothesize that neuromuscular electrical stimulation (NMES) applied contingent to the voluntary activation of the primary motor cortex through MI can help differentiate patterns of activity associated with different hand movements of the same limb by consistently recruiting the separate neural pathways associated with each of the movements within a closed-loop BCI setup. The investigators study the neuroplastic changes associated with training with the two stimulation modalities.

Eligibility

Age range

18–80 years

Sex

All

Healthy volunteers

Yes

Interventions

• Device
  NMES Feedback

  Electroencephalography (EEG) signals will be recorded from subjects as they perform cued tasks for flexing/extending their non-dominant hand. The signals will be processed and classified in real-time using machine learning algorithms to trigger electrical stimulation on the flexors/extensors of the targeted arm contingent to the detection of a subject-specific flexion/extension EEG patterns.

• Device
  Visual Feedback

  Electroencephalography (EEG) - recorded from subjects as they perform cued motor imagery (MI) tasks - are classified in real-time using a subject-specific BCI decoder,. The output classification probability of the decoder is accumulated using exponential smoothing and translated into continuous visual feedback by means of a bar - on a computer screen - that moves to the right or left in response to classification of one or the other MI task.

• Device
  TESS

  Transcutaneous Electrical Spinal Stimulation (TESS) is applied over the C5-C6 spinal segment for 20 minutes at 30Hz with 5kHz carrier frequency.

Location