Epileptic seizure detection using machine learning on electroencephalogram data

Abstract

It is estimated that more than 1% of people in the US have epilepsy, a life-threatening neurological disease characterized by recurrent, unprovoked seizures, due to abnormal electrical activity in the brain. The diagnostic process for epilepsy is very extensive and results in many misdiagnoses. We hypothesized that the implementation of machine learning, specifically utilizing Support Vector Machine (SVM), on preprocessed electroencephalogram (EEG) data would lead to improved accuracy in detecting epileptic seizures. Our study explored the application of machine learning in epileptic seizure detection using EEG data and aimed to improve the accuracy while limiting false positives. The study utilized a preprocessed EEG dataset and evaluated five machine learning models—Logistic Regression, K Nearest Neighbors (kNN), Random Forest, Neural Network, and SVM. We optimized model performance using hyperparameter tuning, the process of optimizing the parameters of a machine learning model to improve its performance, with a particular emphasis on recall. Results reveal that the SVM model outperforms others, achieving an accuracy of 96.77%, precision of 94.27%, and recall of 88.87%. We concluded by underscoring the need for further research to enhance model metrics, encompassing diverse datasets, alternative preprocessing techniques, and addressing privacy issues. This work contributes to advancing epilepsy diagnosis through machine learning applications, with implications for future developments in healthcare.