A Machine Learning Enabled Wireless Intracranial Brain Deformation Sensing System

Area of Science:

• Biomedical Engineering
• Neuroscience
• Sensor Technology

Background:

• Traumatic brain injury (TBI) is often caused by intracranial brain deformation due to mechanical impact.
• Brain deformation is a complex viscoelastic process, not a simple rigid transformation.
• Accurate measurement of in situ brain deformation is crucial for understanding TBI mechanisms.

Purpose of the Study:

• To develop and validate a machine learning-enabled wireless sensing system for predicting intracranial brain deformation.
• To assess the system's accuracy in both in vitro and in vivo experimental models.

Main Methods:

• An implantable soft magnet and an external magnetic sensor array were used to create a wireless sensing system.
• Machine learning algorithms (random forests, k-nearest neighbors, neural networks) interpreted magnetic sensor outputs to predict deformation.
• Validation was performed using in vitro (PVC gel) and in vivo (rat brains) experiments.

Main Results:

• The system accurately predicted in vitro gel deformation, with an absolute error of 138 μm.
• In vivo experiments showed high accuracy in dead animal models (absolute error = 50 μm) and good accuracy in live animal models (absolute error = 125 μm).
• The machine learning models demonstrated strong performance in interpreting sensor data for deformation prediction.

Conclusions:

• The proposed machine learning-enabled sensor system is an effective tool for measuring in situ brain deformation.
• This technology holds promise for advancing TBI research and diagnostics.
• The system's ability to capture viscoelastic deformation offers a significant advantage over traditional methods.