Complexity Analysis of Electroencephalogram Dynamics in Patients with Parkinson’s Disease

Guotao Liu, Yanping Zhang, Zhenghui Hu, Xiuquan Du, Wanqing Wu, Chenchu Xu, Xiangyang Wang, Shuo LiParkinson 2017 · DOI

TL;DR

This paper introduces a novel three-stage method combining Discrete Wavelet Transform (DWT), Sample Entropy (SampEn), and a Three-Way Decision Model (O_CCA) to accurately detect Parkinson's Disease (PD) from EEG signals. This approach achieves a high classification accuracy of 92.86%, outperforming methods without DWT preprocessing, and offers a more nuanced classification by including a "boundary" region for uncertain cases.

What Problem Does This Paper Address?

Parkinson's Disease (PD) is a significant neurodegenerative disorder affecting the motor system, with its diagnosis often relying on clinical observation of motor symptoms like tremor, rigidity, and bradykinesia. While EEG signals offer a non-invasive way to study brain dynamics, they are inherently noisy and non-stationary, making visual distinction between healthy individuals and PD patients challenging. Existing automated detection methods for PD using EEG have faced limitations in accuracy and efficiency. This paper aims to address this gap by developing a robust and accurate method for PD detection from EEG signals, leveraging advanced signal processing and classification techniques.

Methods

The proposed methodology is a three-stage pipeline:

  1. Discrete Wavelet Transform (DWT): This technique is used for signal preprocessing. The authors employ a three-level DWT to decompose the noisy EEG signals into approximation and detail coefficients. This process effectively separates frequency information and helps to remove noise, isolating more relevant signal components.
  2. Sample Entropy (SampEn): After DWT decomposition, SampEn is calculated for each EEG channel. SampEn is chosen for its ability to quantify the regularity and predictability of time series data, making it suitable for analyzing the complex dynamics of EEG signals. It's an improvement over Approximate Entropy (ApEn) due to its independence from record length and reduced computational complexity.
  3. Three-Way Decision Model (O_CCA): The SampEn values from each channel serve as features for a novel classifier, the Optimal Center Constructive Covering Algorithm (O_CCA). This model, based on rough set theory, extends traditional binary classification by introducing three decision regions: Positive (POS), Negative (NEG), and Boundary (BND). This allows for a more nuanced classification, where the boundary region indicates cases requiring further investigation. The O_CCA algorithm automatically determines these regions without requiring prior knowledge of parameters.

Datasets

The study utilized EEG data from 42 participants: 25 healthy subjects and 17 PD patients. The participants' ages ranged from 63 to 78 years. Exclusion criteria included other diseases that could affect EEG signals, such as insomnia or heart disease. The EEG data was recorded using 10-channel electrodes, with a sampling rate of 250 Hz, and each recording lasted 40 seconds. The experiment was performed using Matlab (R2014b, 64 bits).

Key Insights and Ideas

  • DWT is Crucial for Noise Reduction: The significant improvement in accuracy when DWT is used as a preprocessing step highlights its effectiveness in handling the inherent noise and non-stationarity of EEG signals, making subsequent analysis more reliable.
  • SampEn Captures Relevant Dynamics: SampEn proves to be a valuable feature extraction method for EEG signals in the context of PD detection, effectively quantifying signal complexity and regularity.
  • Three-Way Decisions Offer Nuance: The introduction of a boundary region (BND) in the classification process is a key advantage. This acknowledges the inherent uncertainty in some diagnoses and provides a more realistic representation of the classification outcome, guiding further clinical investigation.
  • O_CCA's Automatic Region Determination: The O_CCA algorithm's ability to define classification regions automatically, without relying on predefined parameters or loss functions, makes the model more objective and less prone to subjective bias.

Key Results

The proposed combined scheme achieved a classification accuracy of 92.86% for detecting PD patients from normal subjects. When DWT was omitted as a preprocessing step, the accuracy dropped to 88.10%, underscoring the importance of DWT. The study also compared the O_CCA model with traditional binary classifiers (SVM, KNN, NB, RF) using DWT-processed data. While SVM achieved a slightly higher accuracy (97.62%), it lacked the boundary region capability. The O_CCA model, with its 92.86% accuracy and inclusion of a boundary region, is deemed more suitable for PD detection due to its comprehensive approach.

Limitations and Future Work

  • Sample Size: The authors acknowledge that a larger number of subjects would be beneficial for generalizing the results and improving the model's robustness.
  • Channel Selection: They suggest that exploring attribute reduction, potentially by identifying only the most relevant EEG channels for PD detection, could further enhance sensitivity and efficiency.
  • Data Generalization: Future work could involve collecting more data related to PD and adhering to standard participant selection criteria to improve the generalizability of the findings.

Summary updated

Abstract

In this study, a new combination scheme has been proposed for detecting Parkinson's disease (PD) from electroencephalogram (EEG) signal recorded from normal subjects and PD patients. The scheme is based on discrete wavelet transform (DWT), sample entropy (SampEn), and the three-way decision model in analysis of EEG signal. The EEG signal is noisy and nonstationary, and, as a consequence, it becomes difficult to distinguish it visually. However, the scheme is a well-established methodology in analysis of EEG signal in three stages. In the first stage, the DWT was applied to acquire the split frequency information; here, we use three-level DWT to decompose EEG signal into approximation and detail coefficients; in this stage, we aim to remove the useless and noise information and acquire the effective information. In the second stage, as the SampEn has advantage in analyzing the EEG signal, we use the approximation coefficient to compute the SampEn values. Finally, we detect the PD patients using three-way decision based on optimal center constructive covering algorithm (O_CCA) with the accuracy about 92.86%. Without DWT as preprocessing step, the detection rate reduces to 88.10%. Overall, the combination scheme we proposed is suitable and efficient in analyzing the EEG signal with higher accuracy.
BibTeX
@inproceedings{liu2017complexity,
  author = {Guotao Liu and Yanping Zhang and Zhenghui Hu and Xiuquan Du and Wanqing Wu and Chenchu Xu and Xiangyang Wang and Shuo Li},
  title = {{Complexity Analysis of Electroencephalogram Dynamics in Patients with Parkinson’s Disease}},
  booktitle = {Parkinson},
  year = {2017},
  doi = {10.1155/2017/8701061},
}
Figure