Parkinson’s disease (PD) presents distinct clinical motor phenotypes - tremor-dominant (TD) and postural instability/gait disorder (PIGD) – and in some patients, the episodic and debilitating symptom of freezing of gait (FOG). These clinical characteristics differ in prognosis and responsiveness to therapies such as deep brain stimulation (DBS). Identifying electrophysiological markers that reflect phenotypes, symptoms, and motor states is critical for advancing therapeutic strategies and technologies, including adaptive deep brain stimulation (aDBS). However, traditional analyses of subthalamic nucleus local field potentials (STN-LFPs) may obscure relevant neural dynamics by conflating oscillatory and broadband background activity. This study investigates whether spectral parameterization of STN-LFPs into periodic and aperiodic components enhances the detection of Parkinson’s disease-related neural signatures associated with FOG, phenotypes, and movement conditions.

STN-LFPs were recorded intraoperatively from 35 hemispheres in 22 PD patients (thirteen bilateral, nine unilateral) during both rest and voluntary upper-limb movement. Power spectral density (PSD) was parameterized to separate periodic components (alpha, low beta, high beta) from the aperiodic background (offset, decay exponent, and knee frequency). Periodic power was analyzed with and without adjustment for aperiodic components, enabling more precise identification of true oscillatory changes. Two-way mixed ANOVAs were conducted to assess the effects of phenotype and motor state on spectral features, as well as the interaction between FOG symptoms and motor state. Spearman’s correlations were used to examine relationships between features and canonical motor symptoms of PD. Logistic regression models were built to predict phenotypes based on both adjusted periodic and aperiodic features.

Regarding phenotypes, TD patients showed significant movement-related suppression in adjusted low beta power (p = 0.003), whereas PIGD patients exhibited increased high beta power at rest. Aperiodic components, particularly the decay exponent, distinguished motor conditions and phenotypes, with a steeper decay observed in PIGD during movement, suggesting more inhibited STN activity consistent with the impaired motor adaptability of this group. Rigidity correlated with both periodic and aperiodic features (e.g., adjusted high beta, ρ = 0.601). A logistic regression model combining adjusted low and high beta with the decay exponent achieved improved phenotype classification (AUC = 0.83). Furthermore, adjusted alpha power emerged as a promising feature for differentiating between freezing and non-freezing PD patients.

Overall, our results indicate that decomposing STN-LFPs into periodic and aperiodic components uncovers neural dynamics in PD that extend those captured by classical analysis. These findings support the use of spectral parameterization as a valuable tool for developing adaptive and phenotype-sensitive DBS strategies and for deepening our understanding of the neurophysiological mechanisms underlying this important neurodegenerative disease.