Cognition requires the dynamic coordination of neural ensembles across multiple brain regions. This is one of the biggest neuroscientific questions: how do neural populations form transient communication networks in the service of cognition? One exciting candidate mechanism by which this occurs is through the coupling of neural oscillations between brain regions. These oscillations are a ubiquitous feature of electrophysiology, occurring across species. Despite their wide study, recent work has highlighted many pitfalls in analyzing oscillations, largely centered around three major issues: 1) Oscillations should be measured relative to the aperiodic (1/f) background because, strictly speaking, oscillations are defined as any regions of the power spectrum that rise above the 1/f background, which has itself been shown to be dynamic in relation to both cognition and disease; 2) Most tools for extracting and quantifying oscillations assume that they are sinusoidal despite the fact that they rarely ever are. Further, those non-sinusoidal features may carry critical physiological information; 3) Traditional methods can conflate bursting and non-bursting oscillations, despite the rapidly mounting evidence that the two oscillatory modes are distinct, and may even play different functional roles. In this project we will significantly expand upon analytic software and platforms, developed by my lab, to test the validity of our tools against real and simulated data. These tools are designed specifically to address the three major oscillation analysis issues outlined above. After testing, these analytic toolboxes will then be moved online, to permit cloud-based, large-scale analysis of oscillations, the 1/f background, non-sinusoidal waveform features, and oscillatory bursts. We will then leverage new, dynamic, interactive in-brower visualization tools for data processing and exploration. All of these will be done using open-source tools, built to industry standards of software development, in a transparent manner. 1
As neuroscience moves toward ever-larger, more complex datasets, new methods and analytic tools need to be developed to handle those increasingly larger datasets; at the same time, new standards need to be reached regarding how to analyze these data, and which features need to be reported on. Here we will develop novel analysis toolboxes for the field, allowing for cloud-based analysis and visualization of LFP, M/EEG, and ECoG data. These tools will significantly improve the ability for researchers to link field potential data to underlying physiology, as well as to cognition, behavior, and disease. 1
