EEG Emotion Copilot: Optimizing lightweight LLMs for emotional EEG interpretation with assisted medical record...

Area of Science:

• Affective Computing (AC)
• Brain-Computer Interface (BCI)
• Machine Learning for Healthcare

Background:

• Deep learning shows promise in EEG emotion recognition but faces challenges in real-time processing, personalization, and user interaction.
• Existing methods struggle with end-to-end emotion computation and seamless integration into clinical workflows.

Purpose of the Study:

• To develop and evaluate the EEG Emotion Copilot, a novel system leveraging a lightweight LLM for direct EEG emotion recognition.
• To generate personalized diagnostic and treatment suggestions and automate assisted electronic medical records.
• To enhance computational efficiency and user interaction in AC and BCI applications.

Main Methods:

• Optimization of a 0.5B parameter lightweight LLM for local deployment.
• Novel techniques in prompt data structure, model pruning, and fine-tuning for improved performance.
• Strategies for efficient deployment and enhanced participant interaction.

Main Results:

• The EEG Emotion Copilot demonstrated superior accuracy in emotion recognition and assisted electronic medical records generation compared to larger models.
• Achieved an enhanced intuitive interface for participant interaction.
• Showcased improved computational efficiency and performance.

Conclusions:

• The optimized lightweight LLM-based copilot advances AC applications in the medical domain for mental health monitoring.
• The system offers an innovative solution for real-time emotion analysis and personalized healthcare.
• The proposed system provides a scalable and efficient approach for integrating BCI and AC in clinical settings.