Human state anxiety classification framework using EEG signals in response to exposure therapy

Area of Science:

• Neuroscience
• Biomedical Engineering
• Psychiatry

Background:

• Human anxiety is a significant mental health issue requiring effective assessment tools.
• Current anxiety detection methods may lack objectivity or efficiency.
• Physiological signals offer potential for objective anxiety measurement.

Purpose of the Study:

• To develop an objective framework for human anxiety assessment using electroencephalography (EEG) signals.
• To quantify anxiety into two and four levels based on EEG data.
• To improve the accuracy and reduce the computational complexity of anxiety detection.

Main Methods:

• Utilized EEG signals from the DASPS database, recorded during exposure therapy.
• Pre-processed EEG signals, performed channel and frequency band selection (theta and beta bands).
• Extracted features, applied feature selection, and classified anxiety levels using machine learning algorithms, including random forest.

Main Results:

• Achieved 94.90% accuracy for two-level anxiety classification and 92.74% for four-level classification.
• The random forest classifier used a small subset of 9 and 10 features, respectively.
• The proposed framework demonstrated superior accuracy with reduced feature count compared to existing methods.

Conclusions:

• The developed EEG-based framework provides an objective and accurate method for state anxiety classification.
• The approach enhances efficiency by reducing computational complexity.
• This framework holds promise for improved mental health monitoring and intervention.