Noninvasive Neuroprosthetic Control of Grasping by Amputees

Date

2014-12

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Smooth coordination and fine temporal control of muscles by the brain allows us to effortlessly pre-shape our hand to grasp different objects. Correlates of motor control for grasping have been found across wide-spread cortical areas, with diverse signal features. These signals have been harnessed by implanting intracortical electrodes and used to control the motion of robotic hands by tetraplegics, using algorithms called brain-machine interfaces (BMIs). Signatures of motor control signal encoding mechanisms of the brain in macro-scale signals such as electroencephalography (EEG) are unknown, and could potentially be used to develop noninvasive brain-machine interfaces. Here we show that a) low frequency (0.1 – 1 Hz) time domain EEG contains information about grasp pre-shaping in able-bodies individuals, and b) This information can be used to control pre-shaping motion of a robotic hand by amputees. In the first study, we recorded simultaneous EEG and hand kinematics as 5 able-bodies individuals grasped various objects. Linear decoders using low delta band EEG amplitudes accurately predicted hand pre-shaping kinematics during grasping. Correlation coefficient between predicted and actual kinematics was r = 0.59 ± 0.04, 0.47 ± 0.06 and 0.32 ± 0.05 for the first 3 synergies. In the second study, two transradial amputees (A1 and A2) controlled a prosthetic hand to grasp two objects using a closed-loop BMI with low delta band EEG. This study was conducted longitudinally in 12 sessions spread over 38 days. A1 achieved a 63% success rate, with 11 sessions significantly above chance. A2 achieved a 32% success rate, with 2 sessions significantly above chance. Previous methods of EEG-based BMIs used frequency domain features, and were thought to have a low signal-to-noise ratio making them unsuitable for control of dexterous tasks like grasping. Our results demonstrate that time-domain EEG contains information about grasp pre-shaping, which can be harnessed for neuroprosthetic control.

Description

Keywords

Brain-machine interface, Electroencephalography (EEG), EEG, Upper limb, Neuroprosthetic

Citation