A Hybrid Machine Learning and Explainable-AI Based Approach for Real-Time Epileptic Seizure Detection Using EEG Signals

View at publisher Summarise with VIRAT
Authors: Vaishali Mehta Wadhwa, Balajee Maram, Jyoti Garg, Poonam, Manvi Khatri, Poonam Rani Verma
Cited by: 0
References: 7

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Epilepsy-induced seizures are abrupt, unexpected electrical impulses in the brain that significantly impair one’s health and lifestyle. They can be life-threatening in severe cases. For enhanced diagnosis, monitoring of patients, and prompt treatment, early detection of seizures is crucial. In an effort to detect epileptic seizures, this study proposes a machine learning approach that employs six machine learning classifiers including Decision Tree, Random Forest, K-nearest neighbor, Support Vector Machine, Logistic Regression and XGBoost. First, the dataset was preprocessed and analyzed before and after resampling to address class balancing, a common problem in seizure classification tasks. The accuracy of XGBoost before resampling was 95%. After resampling, improved accuracies were recorded by SVM with 98% and XGBoost with 97%, which proves the importance of balancing techniques to increase resilience. This research gave top priority to two major factors: accurate classification and interpretability, important for medical use. This is realized through the amalgamation of LIME and SHAP into the system, for it provides both local and global interpretability by identifying the critical EEG patterns and features that are necessary for seizure diagnosis. These results indicate how the integration of explainable AI techniques with traditional machine learning classifiers can lead to surprisingly accurate and understandable, trustworthy clinical decisions. The proposed method has an accuracy-interpretability trade-off and can therefore be used as a reliable and therapeutically useful seizure-detection technique in its own right.

Fields of Study & Subjects

EEG and Brain-Computer Interfaces Emotion and Mood Recognition Epilepsy research and treatment

Publication details

Journal / source
Biennial International Conference on Nascent Technologies in Engineering (ICNTE)
Pages
1–6
Publication date
16 January 2026
Language
EN