A study on combining local field potential and single unit activity for better neural decoding

Recent studies showed that the local field potential (LFP) in motor cortex carries information about parameters of limb movements and could be used as a candidate neural signal in brain‐machine interfaces to control external devices. However, it is yet to be clear how much information LFP can offer and how it can be effectively used in BMIs. In this article, we systematically evaluated the decoding performance of combining LFP and single‐unit activity (SUA) from the primary motor cortex of rats performing the lever‐pressing task. The results showed that the decoding power could be significantly improved by combining SUA and LFP in the time‐frequency mode, which is based on the separation of LFP into multiple frequency bands. Furthermore, we found that using all frequency bands might be the best choice because it yielded better or no significantly worse results than using low or high frequency bands only. This implies that different frequency components of LFP carry different information. Moreover, we demonstrated that the combination could stabilize the decoding performance even if SUA disappears over time. These results suggest that the different frequency components in the LFP play different roles in kinematics decoding and the combination of LFPs and SUA could be a promising strategy for improving neural decoding in brain‐machine interfaces.© 2011 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 21, 165–172, 2011