Estimation of Head Accelerations in Crashes Using Neural Networks and Sensors Embedded in the Protective Helmet

Area of Science:

• Biomechanics
• Neuroscience
• Artificial Intelligence

Background:

• Traumatic Brain Injuries (TBIs) are a severe consequence of Powered Two-Wheeler (PTW) crashes.
• Existing head kinematics tracking devices have limitations for real-world road applications.
• Accurate head acceleration data is crucial for early TBI diagnosis and treatment.

Purpose of the Study:

• To develop a new technical solution for estimating head kinematics using accelerometers in motorcycle helmets.
• To create a methodology employing deep Artificial Neural Networks (dANNs) for calculating linear and rotational head acceleration.
• To assess the accuracy of the dANN model in estimating head accelerations during simulated PTW crashes.

Main Methods:

• Integration of accelerometers into a motorcycle helmet design.
• Development of a deep Artificial Neural Network (dANN) model.
• Utilizing a finite element model of a helmet coupled with a Hybrid III head model for data generation.
• Training the dANN using virtual accelerometer data from the helmet's inner surface to predict head accelerations.

Main Results:

• The dANN model accurately estimated linear and rotational head acceleration time patterns.
• High correlation coefficients (R2 values from 0.91 to 0.97) were observed between reference and estimated acceleration values.
• The system demonstrated good accuracy across various impact conditions requiring medical attention.

Conclusions:

• The proposed helmet-based accelerometer system with dANNs offers a reliable method for estimating head kinematics in PTW crashes.
• This technology has the potential to improve early TBI diagnosis and management.
• The findings support the use of advanced AI in developing safer protective gear for motorcyclists.