Scholarlink

Electroencephalography (EEG) is a widely used technique for monitoring the electrical activity of the brain. It plays a crucial role in detecting various neurological conditions, including epileptic seizures. The sudden occurrence of epileptic seizures can significantly impact an individual’s quality of life. Although portable EEG devices are increasingly accessible, real-time analysis remains challenging, requiring lightweight yet effective computational models. This study aims to differentiate between preictal and interictal states using a lightweight Transformer-based architecture known as Separable Vision Transformer. Our approach focuses on capturing the temporal-frequency patterns associated with seizure onset using spectrograms. The proposed model achieves an accuracy of 97.30%, sensitivity of 97.70%, and specificity of 97.00%, using only 9.9k trainable parameters.

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

Epileptic Seizure Prediction Using a Lightweight Separable Vision Transformer and EEG Spectrograms

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ana Fatima,
  • Sarfaraz Masood

摘要

Electroencephalography (EEG) is a widely used technique for monitoring the electrical activity of the brain. It plays a crucial role in detecting various neurological conditions, including epileptic seizures. The sudden occurrence of epileptic seizures can significantly impact an individual’s quality of life. Although portable EEG devices are increasingly accessible, real-time analysis remains challenging, requiring lightweight yet effective computational models. This study aims to differentiate between preictal and interictal states using a lightweight Transformer-based architecture known as Separable Vision Transformer. Our approach focuses on capturing the temporal-frequency patterns associated with seizure onset using spectrograms. The proposed model achieves an accuracy of 97.30%, sensitivity of 97.70%, and specificity of 97.00%, using only 9.9k trainable parameters.