An LFP-based characterization of population-level heterogeneity during motor preparation & execution

The goal of the present work is to characterize mesoscale heterogeneity of population-level neural activity in the motor areas during motor preparation and execution using high-frequency local field potentials (HF-LFPs). Understanding the distribution of neural activity at this scale is necessary since ensembles of neural populations underlie most behaviors. HF-LFPs have emerged as a promising signal for such analysis due to their stability and relationship with cortical population dynamics. Here we analyzed HF-LFPs in non-human primates performing a reach-to-grasp task. We quantified three complementary aspects of heterogeneity: variance, latent dimensionality, and directional coherence. Heterogeneity varied by region and neuromotor state, and was specific to the motor behavior being implemented, even in regions exhibiting high correlation. In addition, more heterogeneous activity tended to better encode neuromotor states. Namely, regions and neuromotor states exhibiting higher variance and dimensionality yielded superior motor decoding performance. Further analyses involving low-dimensional linear projections of the signals revealed rich intra-region interactions between ensembles of populations. The present work helps establish intra-region heterogeneity as a multi-faceted and informative measure of neural activity, even on small spatial scales. It also establishes HF-LFPs as an effective signal for capturing complex intra-region dynamics, making it a suitable candidate for multi-region, systems-level analysis.