An EEG analysis framework through AI and sonification on low power IoT edge devices

Area of Science:

• Medical Technology
• Artificial Intelligence in Healthcare
• Signal Processing

Background:

• Electroencephalography (EEG) analysis is crucial for diagnosing brain disorders, particularly in neonates where seizure detection is challenging.
• High costs of traditional EEG analysis limit access in under-resourced regions.
• Machine learning (ML) offers potential for automating neonatal seizure detection, aiding clinical interpretation.

Purpose of the Study:

• To assess the feasibility of deploying an integrated neonatal EEG analysis framework on low-power IoT edge devices.
• To evaluate the performance of ML and sonification techniques for EEG analysis in resource-constrained settings.
• To develop an accessible and expandable platform for neonatal and adult EEG analysis.

Main Methods:

• Implementation of a comprehensive EEG analysis framework incorporating ML, sonification, and intuitive visualization.
• Deployment and testing on a low-power IoT edge device.
• Evaluation of accuracy and efficiency for ML and sonification components.

Main Results:

• Both ML and sonification techniques were successfully implemented on low-power edge devices without compromising accuracy.
• The developed framework demonstrated efficient processing of neonatal EEG data.
• The platform's architecture allows for scalability to adult populations and other EEG applications.

Conclusions:

• Low-power IoT edge devices are feasible for implementing advanced neonatal EEG analysis, including ML and sonification.
• This approach can significantly reduce the cost and improve the accessibility of critical neurological diagnostics.
• The developed platform offers a versatile solution for expanding EEG analysis capabilities in diverse healthcare settings.