Electromyography Biomarkers for Quantifying the Intraoperative Efficacy of Deep Brain Stimulation in Parkinson's Patients With Resting Tremor
Kailiang Wang, Mathew R Burns, Dan Xu, Wei Hu, Shiying Fan, Chunlei Han, Qiao Wang, Shimabukuro Michitomo, Xiaotong Xia, Jianguo Zhang, Feng Wang, Fangang Meng
Abstract
Introduction: Deep brain stimulation (DBS) is an effective therapy for resting tremor in Parkinson's disease (PD). However, quick and objective biomarkers for quantifying the efficacy of DBS intraoperatively are lacking. Therefore, we aimed to study how DBS modulates the intraoperative neuromuscular pattern of resting tremor in PD patients and to find predictive surface electromyography (sEMG) biomarkers for quantifying the intraoperative efficacy of DBS. Methods: Intraoperative sEMG of 39 PD patients with resting tremor was measured with the DBS on and off respectively, during the intraoperative DBS testing stage. Twelve signal features (time and frequency domains) were extracted from the intraoperative sEMG data. These sEMG features were associated with the clinical outcome to evaluate the efficacy of intraoperative DBS. Also an sEMG based predictive model was established to predict the clinical improvement rate (IR) of resting tremor with DBS therapy. Results: A typical resting tremor with the peak frequency of 4.93 ± 0.98 Hz (mean ± SD) was measured. Compared to the baseline, DBS modulated significant neuromuscular pattern changes in most features, except for the peak frequency, by decreasing motor unit firing rate, amplitude, or power and by changing the regularity pattern. Three sEMG features were detected with significant associations with the clinical improvement rate (IR) of tremor scale: peak frequency power (R = 0.37, p = 0.03), weighted root mean square (R = 0.42, p = 0.01) and modified mean amplitude power (R = 0.48, p=0.003), and they were adopted to trained a gaussian process regression model with leave-one-out cross-validation procedure. The prediction values from the trained sEMG prediction model (1000 permutating times, p= 0.003) showed a good correlationship (r=0.47, p=0.0043) with the true IR of tremor scale. Conclusion: DBS acutely modulated the intraoperative resting tremor mainly by suppressing the amplitude and motor unit firing rate, and by changing the regularity pattern, but not by modifying the frequency pattern. Three features showed strong robustness and could be used as quick intraoperative biomarkers to quantify and predict the efficacy of DBS in PD patients with resting tremor.