Identification of Methamphetamine Abusers Can Be Supported by EEG-Based Wavelet Transform and BiLSTM Networks

Area of Science:

• Neuroscience
• Computational Psychiatry
• Biomedical Engineering

Background:

• Methamphetamine (MA) abuse significantly impairs cognitive function.
• Automatic detection of MA abuse via neural activity remains underexplored.
• Distinguishing MA abusers from healthy individuals based on brain activity is crucial for intervention.

Purpose of the Study:

• To investigate neural activity differences between MA abusers and healthy individuals.
• To develop an automated method for MA abuser discrimination.
• To explore the utility of P300 event-related potentials in MA abuse detection.

Main Methods:

• Event-related potential (ERP) analysis to identify the P300 component.
• Extraction of wavelet coefficients and time-frequency domain features from the P300 component.
• Feature selection using F_score and classification with a Bidirectional Long Short-term Memory (BiLSTM) network.

Main Results:

• The P300 component in EEG signals differs between MA abusers and healthy individuals.
• The BiLSTM network achieved an 83.85% accuracy in detecting MA abusers.
• Optimized feature sets effectively discriminated between the two groups.

Conclusions:

• Neural activity, specifically the P300 component, provides a reliable biomarker for MA abuse.
• This study presents a novel, automated approach for MA abuse prevention and diagnosis.
• The findings support the use of EEG-based machine learning for neurological drug abuse detection.