A Bayesian Model-Selection Approach for Determining the Number of Spectral Peaks in Neural Power Spectra

Neurophysiological brain activity comprises rhythmic (periodic) and arrhythmic (aperiodic) signal elements, which are increasingly studied in relation to behavioral traits and clinical symptoms. Current methods for spectral parameterization of neural recordings rely on user-dependent parameter selection, which includes selecting the correct maximum number of spectral peaks to fit to the spectrum. This challenges the replicability and robustness of findings. Here, we introduce a data-driven model-selection procedure for determining the appropriate number of oscillatory peaks to fit to neural power spectra, based on the Bayesian Information Criterion (BIC). We present extensive tests of the approach with ground-truth and empirical magnetoencephalography recordings. Data-driven model selection enhances both the specificity and sensitivity of spectral decompositions. Overall, the proposed spectral decomposition with data-driven model selection reduces reliance on user-defined ‘maximum number of peaks’ settings, enabling more robust, reproducible, and interpretable spectral parameterizations.