From spectral decomposition to clinical insight: aperiodic and oscillatory dynamics in the subthalamic nucleus of patients with Parkinson’s disease

Brain activity can be measured through electrophysiological signals that reflect the collective activity of neuronal populations. These neural field potentials can be recorded noninvasively (EEG, MEG) or invasively (iEEG, LFP) and exhibit power spectra comprising periodic (oscillatory) and aperiodic components. Traditionally, spectral analyses relied on band-limited power within canonical frequency bands, without explicitly separating these components. Increasing evidence, however, indicates that periodic and aperiodic activity reflect distinct neurophysiological processes and should therefore be analyzed separately. Separating periodic and aperiodic components—commonly termed spectral decomposition—is now frequently applied. Yet, the reliability of existing approaches has not been systematically established. To address this gap, this work evaluates two widely used spectral decomposition methods, specparam and IRASA, across multiple electrophysiological modalities. Empirical analyses reveal systematic deviations from idealized 1/f power laws and identify additional, interrelated challenges in separating spectral components. These findings delineate methodological limitations of current approaches and provide concrete recommendations for mitigating these challenges. Careful spectral decomposition that mitigates these challenges is critical for interpreting electrophysiological activity, including in neurological disease. Parkinson’s disease provides a compelling example of how spectral decomposition impacts clinical interpretation. While exaggerated subthalamic beta activity is a hallmark of the disorder, prior studies largely conflated periodic and aperiodic components. Applying spectral decomposition to subthalamic recordings revealed that isolated low beta oscillations correlate positively with motor symptom severity (ρ = 0.24, p < 0.05), confirming pathological low beta synchrony as a disease-relevant process. In contrast, low-frequency activity (ρ=−0.32, p < 0.001) and low gamma oscillations (ρ=−0.21, p < 0.05) showed prokinetic associations. Notably, aperiodic broadband power—interpreted as a proxy for population spiking activity—explained motor symptom severity best at the single-patient level (p < 0.05), underscoring a central contribution of aperiodic activity in the parkinsonian subthalamic nucleus. Long-term recordings demonstrate stabilization of periodic beta features after implantation and during chronic stimulation, supporting their potential use as chronic biomarkers, while network-level analyses, including cortical recordings, link high beta activity to pathological cortico-subthalamic synchronization. Together, these findings demonstrate that disentangling periodic and aperiodic spectral components enables a more precise and physiologically grounded understanding of parkinsonian neurophysiology, provided that methodological limitations are explicitly considered.