Inferring hand movement kinematics from MEG, EEG and intracranial EEG: From brain-machine interfaces to motor rehabilitation

Abstract

The ability to use electrophysiological brain signals to decode various parameters of voluntary movement is a central question in Brain Machine Interface (BMI) research. Invasive BMI systems can successfully decode movement trajectories from the spiking activity of neurons in primary motor cortex and posterior parietal cortex. It has long been assumed that non-invasive techniques do not provide sufficient signal resolution to decode the kinematics of complex time-varying movements. This view stems from the hypothesis that movement parameters such as direction, position, velocity, or acceleration are primarily encoded by neuronal firing in motor cortex. Consequently, the fact that such signals cannot be detected using non-invasive techniques such as Electroencephalography (EEG) or Magnetoencephalography (MEG) has led to the claim that fine movement properties cannot be decoded with non-invasive methods. However, this view has been proven wrong by numerous studies in recent years. First, a growing body of research over the last decade has shown that the local field potential (LFP) signal, which represents the summed activity of a neuronal population, can encode movement parameters at a level comparable to unit recordings. These findings were confirmed in humans by the successful use of electrocorticography (ECoG) to achieve continuous movement decoding via invasive human BMI approaches. Very recently, a number of non-invasive studies were able to provide striking evidence that even surface-level MEG or EEG data can contain sufficient information on hand movement in order to infer movement direction and hand kinematics from brain signals recorded using non-invasive methods. Here we provide a brief review of this recent literature and discuss its importance on the future of BMI research and its implications on the development of novel motor rehabilitation strategies.