A Spark-based Analytic Pipeline for Seizure Detection in EEG Big Data Streams

Area of Science:

• Neurology
• Data Science
• Machine Learning

Background:

• Epilepsy affects 1% of the global population, characterized by recurrent seizures.
• Real-time seizure detection from Electroencephalogram (EEG) signals is crucial for epilepsy diagnosis and management.
• Continuous brain activity monitoring generates massive datasets, posing a big data challenge for real-time analysis.

Purpose of the Study:

• To develop and evaluate a Spark-based machine learning approach for real-time seizure detection in EEG signals.
• To address the big data challenges associated with continuous neurophysiological monitoring.

Main Methods:

• Utilized a Spark-based machine learning framework.
• Employed linear dimensionality reduction techniques for data processing.
• Implemented classification algorithms for seizure identification.

Main Results:

• Achieved an average accuracy of 99.32% across 24 patients.
• Demonstrated high average sensitivity (99.41%) and specificity (95.25%).
• Reported a low average latency of 0.38 ms for the seizure detection framework.

Conclusions:

• The proposed Spark-based machine learning approach is effective for accurate and efficient real-time seizure detection.
• This method offers a scalable solution for analyzing high-throughput neurophysiological data in epilepsy monitoring.
• The framework shows potential for improving the clinical diagnosis and treatment of epilepsy.