Reconstructing 3-Dimensional Hand Movements from Electroencephalographic Signals

Noninvasive observations such as EEG are relatively hard to interpret, due to low signal-to-noise ratio (SNR), low spatial resolution and reduced spectral content. However, much effort is put in interpreting the signals for the devise of non-invasive Brain Computer Interfaces (BCIs).

In this study, we aimed at correlating EEG activity with actual and imagined repetitive hand movements to targets positioned in 3D space (a block design), using Linear Regression (MLR) analysis and various time lags between motion kinematics and neural activity. We have found a strong correlation between EEG activity and hand movements (both actual and imagined), reaching r=.6, two times stronger than was reported in previous studies. Moreover, testing for the channels and time lags that provided best trajectory reconstruction revealed the brain areas and networks engaged in motor planning and execution.

These results may hopefully enable the devise of more proficient non-invasive BCIs for patients with a variety of neurological pathologies, including Multiple Sclerosis, ALS, stroke and Lock-In-Syndrome (LIS).