Detection of tremor bursts by a running second order moment function and analysis using interburst histograms

Abstract 

Introduction

Conventional linear signal processing techniques are not always suitable for the detection of tremor bursts in clinical practice due to inevitable noise from electromyographic (EMG) bursts. This study introduces (1) a non-linear analysis technique based on a running second order moment function (SOMF) and (2) auto- and cross-interburst interval histograms (IBIH) showing distributions of interburst interval EMG bursts of pathological tremors illustrating an application of the SOMF.

Materials and methods

EMG recordings from extensors and flexors of two patients with Parkinson's disease with a rest tremor and from a healthy subject during sustained muscular contraction were preliminary analyzed in a pilot study. The SOMF was obtained by repeated second order moment calculations within a window of fixed width W (time scale parameter) plotted as a function of time. Minimum SOMF values indicate local “moments of inertia” of each EMG burst. Bursts were detected and located when minimum SOMF values were below level L (decision parameter). Optimal settings of parameters W and L were calculated empirically for pathological tremor EMGs. Auto- and cross-IBIHs were obtained from minimum SOMF values of detected bursts.

Results

Tremor frequency and phase relation between EMG bursts from auto- and cross-IBIHs agreed with those derived from spectral analysis. Burst detection by SOMF has a high sensitivity and selectivity even with noisy background.

Conclusion

The SOMF is appropriate for detection of individual EMG bursts of pathological tremors. The technique is sensitive to non-stationary changes of tremor bursts regardless of their amplitude. IBIHs provide a measure of tremor frequency and phase difference between EMG bursts.

Keywords: Cogwheel phenomenon, EMG burst, Interburst interval histogram, IBIH, Parkinson's disease, Second order moment function, SOMF, Tremor, Time scaling