Aperiodic activity as a biomarker of seizures and neuromodulation.

Abstract

Introduction: Mounting evidence suggests the efficacy of neuromodulation for epilepsy is mediated by network remodeling and neural state. Epilepsy network related pathophysiology has been associated with variation in the aperiodic exponent, which describes the inverse relationship between frequency and power and has been linked to synaptic-level processes. This study sought to assess relationships between periodic and aperiodic activity, disease state, and responsive stimulation.

Methods: Chronic intracranial EEG was recorded from 13 patients undergoing responsive neurostimulation for epilepsy. Recordings containing clinician-annotated seizures, stimulation triggered by device-detected interictal epileptiform activity (IEA), and stimulation-free interictal periods were analyzed. Multidien IEA cycles were identified, and recordings were classified by cycle phase. Power spectra were parameterized into periodic and aperiodic components using an established algorithm.

Results: The aperiodic exponent was larger during seizures and pre-stimulation intervals than stimulation-free interictal recordings, and decreased following stimulation. A rise in aperiodic exponent was observed in the 12 hours preceding seizures. Larger aperiodic exponent was observed during the rising phase of multidien IEA cycles. Periodic alpha and beta power were larger during seizures, pre-stimulation intervals, and high-risk (rising and/or peak) IEA cycle phases, whereas periodic theta and gamma power exhibited variable relationships. Periodic power did not change after stimulation or in the hours before seizures for any studied frequency band.

Conclusions: The aperiodic exponent was positively related to instantaneous and multidien disease state severity and negatively related to therapeutic stimulation. Aperiodic activity may emerge as a practical biomarker of disease state and treatment response to guide neuromodulation for epilepsy.