Scholarlink

Epilepsy, a complex neurological condition, is identifiable through intricate brain signals generated by interconnected neurons. Analyzing these signals, often obtained through electroencephalograms (EEG), is challenging due to their complexity, noise, and high data volume. To address these challenges, we used Discrete Wavelet Transform (DWT), a mathematical technique for signal analysis. Our focus was on precise classification, particularly in identifying seizures and extracting valuable insights. High-performing machine learning (ML) classifiers were employed to overcome these hurdles. Our in-depth study involved a comprehensive comparative analysis of Deep Learning (DL) models, specifically Residual Networks (ResNet-18) and Deep Neural Networks (DNN), for binary classification in predicting epileptic episodes using EEG data. The ResNet-18 model showed efficient real-time prediction after just 10 epochs and outperformed the DNN model. Evaluation metrics included accuracy, precision, recall, and F1 scores, providing valuable insights for enhancing epilepsy prediction and patient care.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comprehensive Comparison of Machine Learning Models with DWT for EEG-Based Epilepsy Prediction: Including Residual and Deep Neural Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Rehab Naily,
  • Siwar Yahia,
  • Mourad Zaied

摘要

Epilepsy, a complex neurological condition, is identifiable through intricate brain signals generated by interconnected neurons. Analyzing these signals, often obtained through electroencephalograms (EEG), is challenging due to their complexity, noise, and high data volume. To address these challenges, we used Discrete Wavelet Transform (DWT), a mathematical technique for signal analysis. Our focus was on precise classification, particularly in identifying seizures and extracting valuable insights. High-performing machine learning (ML) classifiers were employed to overcome these hurdles. Our in-depth study involved a comprehensive comparative analysis of Deep Learning (DL) models, specifically Residual Networks (ResNet-18) and Deep Neural Networks (DNN), for binary classification in predicting epileptic episodes using EEG data. The ResNet-18 model showed efficient real-time prediction after just 10 epochs and outperformed the DNN model. Evaluation metrics included accuracy, precision, recall, and F1 scores, providing valuable insights for enhancing epilepsy prediction and patient care.