NeuroTech ML

EEG Analysis of Stress Anticipation

A machine learning project for analyzing EEG data to detect and classify stress response anticipation during delayed auditory stimulus.

This project examines how the brain responds to stress when anticipating delayed auditory stimuli. The analysis focuses on the relationship between delta (1-4 Hz) and beta (13-30 Hz) frequency bands, which has been linked to stress regulation.

The experiment presented participants with two types of auditory beeps:

• DelayedBeep: Stress-inducing delayed tone
• NonDelayedBeep: Control, normal tone

Using EEG recordings from multiple participants (Andy, Joycelynn, Okyanus), we analyzed the brain's electrical activity to classify these different states.

Key Findings

All recordings achieved above-chance accuracy, ranging from 68.33% to 81.67%, with the highest accuracy of 81.67% from the JoycelynModerate2 dataset.

This demonstrates that the EEG patterns associated with delayed and non-delayed conditions are distinct and can be reliably differentiated using Common Spatial Patterns (CSP) and Linear Discriminant Analysis (LDA) techniques.

Technical Details

The analysis pipeline includes:

• Loading and preprocessing EEG data
• Extracting and cleaning event markers
• Creating MNE Raw objects and applying filters (1-30 Hz)
• Performing ICA for artifact removal
• Creating epochs around events of interest
• Computing band power (delta and beta)
• ML classification using CSP + LDA
• Visualization of evoked responses

EEG Channels: P4, Pz, P3, C3, Cz, C4, F3, F4

Sampling Rate: 250 Hz

Filtering: 1.0-30.0 Hz

Epoch Window: -0.5 to 1.5 seconds around events