Entropy: A Promising EEG Biomarker Dichotomizing Subjects With Opioid Use Disorder and Healthy Controls

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Electroencephalography (EEG) signals are complex, non-stationary, and contain vital information about brain activity.
• Traditional linear methods struggle with EEG's nonlinear and random nature, limiting diagnostic accuracy.
• Accurate analysis of subtle EEG variations is crucial for diagnosing psychiatric disorders.

Purpose of the Study:

• To develop advanced signal processing algorithms for extracting subtle information from EEG signals.
• To enhance the classification accuracy of brain abnormalities, particularly in substance use disorders.
• To evaluate the efficacy of finite impulse response (FIR) filtering and entropy-based features.

Main Methods:

• Employed a finite impulse response (FIR)-based filtering process, moving beyond traditional time and frequency domain methods.
• Analyzed FIR subbands to extract feature vectors using various entropy markers.
• Utilized a multilayer perceptron classifier and compared performance using classification accuracies and ROC curve scores.

Main Results:

• The introduced methodology shows promising potential for clinical applications, particularly in differentiating substance use disorder subjects.
• Entropy estimators effectively distinguished between normal subjects and those with opioid use disorder.
• The theta frequency band in EEG data demonstrated significant capability across most entropy types, with nonextensive Tsallis entropy showing superior performance.

Conclusions:

• The developed FIR filtering and entropy analysis approach offers a promising tool for EEG-based medical data analysis.
• This method can serve as a clinical interface for diagnosing and monitoring psychiatric disorders, including substance use disorders.
• Entropy measures, especially Tsallis entropy in the theta band, are highly effective for classifying brain abnormalities from EEG signals.