Harmful Brain Activity Classification using EEG Signals

The accurate classification of harmful brain activity is crucial in medical contexts, such as epilepsy monitoring units and brain-computer interface (BCI) systems. In this study, we focus on developing robust algorithms to automatically detect and classify harmful brain events using electroencephalography (EEG) data. Our goal is to enhance diagnostic accuracy, reduce false positives, and enable faster intervention. We examined machine learning models to detect seizures and other pathological states using EEG recordings, such as XGboost, Support Vector Machines (SVM), KNN, Random Forests, and Decision Trees. We evaluate our approach on two datasets—the DEAP dataset (originally designed for emotion analysis) and the HMS-Harmful Brain Activity Classification dataset from Kaggle. Our proposed model achieves promising results, with 99.2% accuracy on the DEAP dataset and an impressive 96% accuracy on the HMS dataset using XGboost. Our work contributes to advancing medical diagnostics and improving patient care by automating the detection of harmful brain activity from EEG signals.