Does brain connectivity hold the key to safer roads? EEG-based fatigue detection in young drivers using interpretable...

Area of Science:

• Neuroscience
• Road Safety
• Artificial Intelligence

Background:

• Mental fatigue is a major risk factor for road accidents among young drivers.
• The neural mechanisms of driver fatigue are not well understood.
• Accurate detection of driver fatigue is crucial for preventing accidents.

Purpose of the Study:

• To investigate the neurophysiological basis of driver fatigue using EEG.
• To develop an accurate and interpretable deep learning model for detecting fatigue in young drivers.
• To enhance road safety by identifying and mitigating fatigue-related risks.

Main Methods:

• Collected EEG data from 32 young drivers on real roads and in a simulated environment.
• Analyzed EEG signals to construct brain functional networks with topological features.
• Developed a deep learning model (MHSA-xLSTM) integrating multi-head self-attention and long short-term memory.
• Utilized Shapley Additive exPlanation for model interpretability.

Main Results:

• The MHSA-xLSTM model achieved 94.39% accuracy in detecting mental fatigue.
• Small-world attribute and node strength were significant features for fatigue recognition.
• Brain's self-regulatory capacity can mitigate fatigue-related impairments.
• Driving experience can enhance performance and reduce accident risk.

Conclusions:

• EEG-based brain network analysis and deep learning offer potential for real-time driver fatigue detection.
• Interpretable models can inform targeted interventions to reduce accident risks.
• Understanding neural mechanisms of fatigue is key to improving road safety for young drivers.