Time-frequency Relevancy Analysis between Local Field Potentials and Lever Pressing Motion of Rats

Cortical Local field potentials (LFPs) in primary motor cortex (M1) include significant movement information. As a movement decoding source, LFPs have the advantage of being long-lasting and stable. However, LFPs contain activity of neural ensemble from neighborhood or even distant cortical areas. Information fully relevant to the movement task is necessary to be extracted. Existing studies indicate the movement relevant information is embedded in different frequency ranges. The analysis generally assumes linear relationship between band-limited LFPs and movement or specifically bases on the decoding method. The contribution of each frequency band is exhaustedly search for a variety of frequency band combinations, which increases the computation complexity. Considering the distinct frequency ranges may also exhibit relevancy over time, we propose to analyze the time-frequency relevancy between LFP signals and the rat lever-press tasks with a local learning based algorithm method. This approach extracts the important time-frequency features from the nonlinear LFP data structure, and labels the sequence of importance for time-frequency features. Comparing to the greedy search, our method demonstrate more clear time-frequency contribution, especially 50-100Hz signal and 200-400Hz signal for holding and pressing respectively, and are validated by better decoding results.