Model selection for spectral parameterization

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. Lay summaryBrain activity is composed of rhythmic patterns that repeat over time and arrhythmic elements that are less structured. Recent advances in brain signal analysis have improved our ability to distinguish between these two types of components, enhancing our understanding of brain signals. However, current methods require users to adjust several parameters manually to obtain their results. The outcomes of the analyses, therefore, depend on each users decisions and expertise. To improve the replicability of research findings, the authors propose a revised method to streamline the analysis of brain signal contents. They developed a new algorithm that defines the parameters of the analytical pipeline informed by the data. The effectiveness of this method is demonstrated with both synthesized and real-world data. The approach is made available to all researchers as a free, open-source app, observing best practices for neuroscience research.