Quantifying cortex-wide traveling brain waves of complex patterns with a graph-based algorithm.

Traveling brain waves (TWs) are neural oscillations that propagate across the nervous tissue. Recently, wave detection algorithms have successfully linked TWs to various brain functions, including perception and movement. However, most existing approaches are not well-suited for large-scale recording systems that span multiple areas on highly curved surfaces. Here, we developed a framework that combines generalized phase analysis and a graph-based algorithm for detection of cortical traveling waves, enabling the decomposition of complex large-scale phase patterns into multiple simultaneously propagating TWs and the quantification of their properties in single-trial data. We applied the algorithm to hemispheric electrocorticogram recordings (82 and 122 channels) from two marmosets performing a visually-guided saccade task. We found that the 20-50 Hz activity in the occipital cortex during the first 100 ms following saccade offset forms a macroscopic TW. The wave originated in the primary visual cortex (V1) and propagated rostrally toward temporal and parietal areas. Moreover, the post-saccadic TWs originated from specific locations within V1 that were consistent with the retinotopic representations of the saccade targets. Wave parameters, including latency, duration, spatial coverage, and amplitude also depended on saccade direction. Combined with large-scale neural recordings, the graph-based framework introduced here provides a general approach for elucidating highly dynamic inter-areal communication and coordination mediated by TWs.