Savitzky-Golay filtering to improve individual alpha frequency estimation

Slides presented at the Neural Oscillations Research Group, Centre for Human Brain Health, University of Birmingham, 10th July 2018.

Abstract:

In the wake of the 'reproducibility crisis' on the one hand, and the increasing trend towards 'big data' on the other, the need for efficient, automated, and fully-replicable data analysis techniques has never been more pressing. This talk presents a recently developed tool that aims to realise these ideals in the context of individual alpha frequency (IAF) estimation^[1]. `restingIAF` is an open-source MATLAB package^[2] that implements Savitzky-Golay filtering^[3] in order to smooth and differentiate power spectral density estimates derived from resting-state M/EEG. This routine attempts to parameterise the bounds of the individual alpha band and extract the alpha peak and centre of gravity frequencies. Individual summary statistics are then computed for subsequent group-level analysis. In this talk, I will summarise some of the difficulties and pitfalls that attend conventional approaches to IAF estimation, before characterising the key methodological features of the `restingIAF` approach.

^[1] Corcoran, A.W., Alday, P.M., Schlesewsky, M., & Bornkessel-Schlesewsky, I. (2018). Toward a reliable, automated method of individual alpha frequency (IAF) quantification. Psychophysiology, e13064. doi: 10.1111/psyp.13064

^[2] https://github.com/corcorana/restingIAF (for a Python implementation, see https://gitlab.com/palday/philistine)

^[3] Savitzky, A. & Golay, M.J.E. (1964). Smoothing and differentiation of data by simplified least squares procedures. Analytical Chemistry, 36(8), 1627-1639. doi: 10.1021/ac60214a047