An Epileptic Seizure Detection System based on Mediapipe and Deep Learning

Epilepsy, a neurological disorder characterized by recurrent seizures, poses significant challenges in timely intervention and patient safety, affecting millions of individuals worldwide. Early and accurate detection of seizures is crucial for timely medical intervention and improved patient outcomes. This system holds particular significance for the elderly demographic, wherein prompt assistance is imperative to safeguard their lives. This study integrates MediaPipe Pose Detection alongside Machine Learning and Deep Learning techniques to enable the identification of individuals experiencing generalized epileptic seizures within the confines of their residences. Moreover, hyperparameter tuning has been performed on the machine learning algorithms to find optimal combination of hyperparameters for each algorithm. Experimental results show that highest average accuracy among machine learning models is achieved by SVM Classifier and Gradient Boosting algorithms at 88.08%, while Time series classification implemented using 1DCNN performs at 84.77% accuracy. It is noted that more varied data samples and standardized datasets would likely improve current results.