Simulations reveal that beta burst detection may inappropriately characterize the beta band

Abstract

In neurophysiological research, the traditional view of beta band activity as sustained oscillations is being reinterpreted as transient bursts. Bursts are characterized by a distinct wavelet shape, high amplitude, and, most importantly, brief temporal occurrence. The primary method for their detection relies on a threshold-based analysis of spectral power, and this presents two fundamental issues. First, the threshold selection is effectively arbitrary, being influenced by both temporally proximal and distal factors in the signal. Second, the method necessarily detects temporal events, as such it is susceptible to misidentifying sustained signals as transient bursts. To address these issues, this study systematically explores burst detection through simulations, shedding light on the methods robustness across various scenarios. Although the method is effective in detecting transients in numerous cases, it can be overly sensitive, leading to spurious detections. Moreover, when applied to simulations featuring exclusively sustained events, the method frequently yields events exhibiting characteristics consistent with a transient burst interpretation. By simulating an average difference in power between experimental conditions, we illustrate how apparent burst rate differences between conditions can emerge even in the absence of actual burst rate disparities, and even in the absence of bursts. This capacity to produce misleading outcomes challenges the reinterpretation of sustained beta oscillations as transient bursts and prompts a critical reassessment of the existing literature. New and NoteworthyNeurophysiological research is experiencing a transformative shift in understanding beta band activity, moving away from the notion of sustained oscillations towards recognizing the significance of transient bursts. Here we show how the methods to detect such bursts are prone to spurious detections and can blur the distinction between sustained signals and transient bursts. Further, in realistic scenarios these methods can produce apparent behavioral associations where no such association exists.