Smart Seizure Detection System: Machine Learning Based Model in Healthcare IoT

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Epilepsy is characterized by recurrent seizures with diverse etiologies.
• Seizures are transient events, often leaving no lasting trace post-recovery.
• Electroencephalograms (EEG) detect abnormalities only during active recording.

Purpose of the Study:

• To develop a smart seizure detection system for Healthcare IoT applications.
• To address the challenges in analyzing electroencephalogram (EEG) data for epilepsy.
• To overcome limitations of manual seizure identification and its impact on patients.

Main Methods:

• An integrated methodology combining signal processing techniques (Discrete Wavelet Transform, Hjorth Parameters, statistical features) and classifier ensembles (Decision Trees, Logistic Regression, Support Vector Machine).
• Utilized Healthcare IoT for real-time data analysis.
• Compared results with kNN classifier, other datasets, and state-of-the-art techniques.

Main Results:

• Achieved remarkable accuracy, reaching up to 100% in specific experimental conditions.
• Demonstrated the effectiveness of combining diverse signal processing and machine learning approaches.
• Validated the system's performance against established methods and datasets.

Conclusions:

• The proposed smart seizure detection system offers a robust solution for EEG data analysis in Healthcare IoT.
• The methodology provides real-time data crucial for informed clinical decision-making.
• This advancement significantly contributes to improved patient outcomes in epilepsy management.