Separation of multi-channel spinal cord recordings using unsupervised adaptive filtering

In anesthetized animals, evoked motor signals descending through the corticospinal tract were recorded from the spinal cord with selectivity using multi-contact surface electrodes. However, the spatial selectivity needs to be improved for this approach to be used as a multi-channel neural interface. In this study, we applied the blind source separation (BBS) technique to improve the separation between the neural channels. The BSS algorithm improved the selectivity from an initial value of less than 1% to 91% although the signal-to-noise ratio of the signals was as low as 0.46 on average.