Discrimination and Characterization of Parkinsonian Rest Tremors by Analyzing Long-Term Correlations and Multifractal Signatures

Abstract

In this paper, we analyze 48 signals of rest tremor velocity related to 12 distinct subjects affected by Parkinson's disease. The subjects belong to two different groups, high-and low-amplitude rest tremors, with four and eight subjects, respectively. Each subject has been tested in four settings given by combining the use of deep brain stimulation and L-DOPA medication. We develop two main feature-based representations of such signals, which are obtained by considering (i) the long-term correlations and multifractal properties, and (ii) the power spectra. Such feature-based representations are initially utilized for the purpose of characterizing the subjects under different settings. Our results show that the effect of medication is clearly recognizable. In addition, when medication is used, we observe a change in the related signals from anti-correlated to long-term, positively correlated. In agreement with previous studies, we observe that deep brain stimulation does not significantly characterize neither of the two groups regardless of the adopted representation. On the other hand, the medication effect yields statistically significant differences in both high-and low-amplitude tremor groups. We successively test several different instances of the two featurebased representations of the signals in the setting of supervised classification and (nonlinear) feature transformation. We consider three different classification problems, involving the recognition of (i) the presence of medication, (ii) the use of deep brain stimulation, and (iii) the membership to the high-and low-amplitude tremor groups. Classification results show that the use of medication can be discriminated with higher accuracy, considering many of the feature-based representations. Notably, we show that the best results are obtained with a parsimonious, two-dimensional representation encoding the long-term correlations and multifractal character of the signals.

Bispectral Analysis of Parkinsonian Rest Tremor: New Characterization and Classification Insights Pre-/Post-DBS and Medication Treatment

“…• The proposed methodology achieves high performance, while maintaining a robust behavior across different conditions, in recognizing the effect of medication and DBS and classifying subjects to the HAT/LAT classes. • In agreement with previous studies [9], [15], [16], we also show that medication performs better than DBS in suppressing tremor and that a tremor rebound occurs 15 minutes after DBS is switched off [9]. We also observe a secondary rebound that was not reported in [9], 60 minutes after DBS is switched off.…”

“…One dataset that is used in the analysis of PD rest tremor includes data from velocitytransducing laser, aimed at the index finger of each subject [9]. Works that utilized this dataset so far used deep learning to predict different attack stimulations in DBS [14], employed discrete non-Markovian stochastic processes to study the Parkinsonian pathological tremor [15], or considered Long-Term Correlations (LTC) and Multifractal (MF) properties to tackle three different classification problems [16]. The Fast Fourier Transform (FFT) [17] and the Power Spectrum (PS) [9] have also been utilized to analyze tremor amplitude and frequency characteristics of resting tremor signals.…”

Machine Learning’s Application in Deep Brain Stimulation for Parkinson’s Disease: A Review

“…The fidelity of tremor detection in calibrating a closed loop adaptive system has been demonstrated in several studies [65,66]. The use of ML in tremor classification for DBS patients was also explored and deemed as a useful tool for adaptive, patient-specific optimization [49,62,63,[67][68][69][70][71]. Khobragade et al [62] examined the surface electromyography of two patients (one PD-DBS) over two sessions and analyzed the signals using an NN.…”

Data-Driven Prediction of Freezing of Gait Events From Stepping Data

“…While there are several recent contributions applying machine learning to Parkinsonian tremors (16) and gait analysis (17) as well as the freezing of gait detection (13,18), to our knowledge, no work has used measurements of stepping actions, using kinetic data alone to predict upcoming freezing events. Therefore, this study aims to forecast freezing events from the kinetic stepping data.…”