Schedule of Events | Symposia

Improving the robustness of oscillation detection

Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Kieran Pawluk¹ (kpawluk@ualberta.ca), Tamari Shalamberidze¹, Jeremy Caplan¹; ¹University of Alberta

Neural oscillations at various frequencies are associated with many brain functions. To quantify them, it is important to distinguish oscillatory from non-oscillatory 1/f background activity. The BOSC method (Better OSCillation detection; Caplan et al., 2001) does this by estimating the background power spectrum and deriving detection thresholds from that estimate in order to disregard most background signal. When successful, this produces detection criteria that are fairly calibrated across frequencies. But if the background estimate is inaccurate, this can backfire and potentially introduce a bias across frequency. We used both real and synthetic signals to estimate the severity of these problems as they arose, focusing on shorter time windows and when substantial power existed at one end of the measured spectrum. With the goal of improving robustness in these conditions, we then compared various improvements to the background estimate. The combination that produced the best results involved removing high-power values across frequencies, using median rather than mean power values, and replacing ordinary least-squares regression with robust regression. When comparing the optimized BOSC method to the standard method, the standard method fared reasonably well aside from some extreme edge cases. Outcomes suggested that at very short time windows or when artifacts or lopsided power spectra are a concern, the combination of modifications in the optimized BOSC method could result in a more selective and thus more accurate fit to the coloured-noise background signal.

Topic Area: METHODS: Electrophysiology