The Entropy of Adaptively Segmented Beta Oscillations Predict Motor Improvement in Patients With Parkinson's Disease

Beta bursts of local fields potentials (LFPs) recorded from subthalamic nucleus (STN) have been recently proposed as a new temporal feature for patients with Parkinson's disease (PD). We introduce a new technique for the adaptive time-domain segmentation of STN-LFP recordings such that the constructed time segments are proportional to the duration of stationary beta activity. We investigated whether the spectral entropy of the adaptively captured beta oscillations can describe the improvement in motor signs following dopaminergic medication.

STN-LFP recordings from externalized chronic deep brain stimulation (DBS) leads were obtained in 9 PD patients. During this monitoring, each patient underwent 3 medication intake cycles where short acting agents (L-DOPA equivalent dose) were administered. We analyzed 2-minute resting state LFP data in each OFF and L-DOPA-induced ON medication states and constructed time domain segmentation of LFP signal in which the length segmentations are adapted to time-varying nature of the oscillatory activity.

Adaptively constructed segments were noted to be significantly longer in OFF- and shorter in ON-state (p<0.001). Interestingly, in the OFF state, the peak frequency of long beta bursts (>375 ms) was in the low range (12-23 Hz) of the beta spectrum, whereas shorter beta bursts (<375 ms) were widespread in the 13-30 Hz band. Measured clinical improvement was highly correlated with the difference in the spectral entropy of beta bursts between OFF and ON states (r = -0.83, p<0.01).

Our findings suggest that beta oscillations can be adaptively segmented without the use of a predetermined amplitude threshold, thereby allowing for objective quantification of burst itself. Compared to the shorter ones, longer oscillations with duration ≥ 375 ms were highly correlated with the clinical improvement, supporting a pathological role for them. Overall, these findings coupled with our adaptive approach could enable the quantitative use of temporal dynamics of beta activity in assessing severity of PD and improvements in motor features.