ChatBCI-4-ALS: A High-Performance, LLM-Driven, Intent-Based BCI Communication System for Individuals with ALS

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Amyotrophic lateral sclerosis (ALS) causes severe motor and speech impairments, necessitating alternative communication methods.
• P300 speller brain-computer interfaces (BCIs) offer non-muscular communication but are often slow for ALS patients' needs.
• Existing BCIs may not fully address the need for efficient communication with minimal cognitive load.

Purpose of the Study:

• To introduce ChatBCI-4-ALS, the first intent-based BCI communication system specifically designed for individuals with ALS.
• To enhance communication speed and enable efficient expression of user intent beyond literal text.
• To develop novel semantic-based metrics for evaluating intent-based communication effectiveness.

Main Methods:

• Leveraging large language models (LLMs) within the ChatBCI-4-ALS system.
• Implementing a dynamic flash algorithm to optimize typing speed and intent recognition.
• Utilizing P300 event-related potentials (ERPs) detection for BCI operation.
• Introducing and applying new semantic-based quantitative performance metrics.

Main Results:

• ChatBCI-4-ALS achieved a record average spelling speed of 23.87 characters per minute.
• The system demonstrated a best-case spelling speed of 42.16 characters per minute.
• A peak information transfer rate (ITR) of 128.85 bits per minute was recorded.
• Online experiments validated the system's effectiveness in intent-based communication.

Conclusions:

• ChatBCI-4-ALS represents a significant advancement in P300 BCI-based communication systems for ALS.
• The intent-based approach and LLM integration enhance communication efficiency and user experience.
• The developed system offers a promising solution to improve the quality of life for individuals with ALS through advanced BCI technology.