Detection of Epileptic Seizure Using Accelerometer Time Series Data and Hidden Markov Model

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Epilepsy affects 70 million globally, necessitating accessible seizure detection.
• Current methods like video-EEG monitoring are costly and hospital-dependent.
• There is a need for simpler, affordable remote seizure detection systems.

Purpose of the Study:

• To develop and evaluate a wireless remote monitoring system for epileptic seizure detection.
• To apply a Hidden Markov Model (HMM) for analyzing accelerometer data.
• To compare HMM performance against other machine learning models.

Main Methods:

• Utilized time-series accelerometer data from a wrist-worn device.
• Applied a Hidden Markov Model (HMM) to a reduced-dimension feature vector.
• Compared HMM with Support Vector Machine (SVM), logistic regression, and random forest models.

Main Results:

• The HMM approach achieved 95.7% seizure detection accuracy.
• A low false alarm rate of 14.8% was recorded.
• The system demonstrated a rapid run time of under 24 seconds.

Conclusions:

• The proposed HMM-based system offers an effective and efficient solution for remote epileptic seizure detection.
• This approach provides a more accessible and affordable alternative to traditional monitoring methods.
• The study highlights the potential of wearable sensors and advanced algorithms in managing epilepsy.