Scholarlink

<p>The automatic epileptic seizure detection method has recently been developed using electroencephalography (EEG) signals and the machine learning approach. But, the correlation and contextual information among different channels of EEG signals are not obtained precisely from the previous methods. This research introduces a novel approach, the DBN–GC–CRF framework, which integrates glial chains and conditional random fields to specifically address these identified issues. First, feature vectors for statistical measures, power measures, and Zhao-Atlas-Marks domain using a sliding window, are retrieved EEG signals. The DBN–GC–CRF algorithm is then used for extracting the high-level feature and generating its label sequence. Finally, the classification output for EEG data to classify epileptic seizures and non-epileptic seizures is obtained through the K-nearest neighbour algorithm in the decision merge layer. The proposed methodology is implemented on two publicly available epilepsy datasets namely Bonn and CHB-MIT EEG data. The output demonstrates that this approach can extract inter-correlation and contextual information from the EEG channel using the glial chains. The accuracy percentage obtained in the detection of epileptic seizures is 99.10% and F-score is 0.9712 for the Bonn dataset. Similarly, for CHB-MIT, an accuracy of approximately 99.56% and an F-score of 0.9833 were obtained. According to the results, the proposed method outperforms other deep classifiers.</p>

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

An Integrated DBN–GC–CRF Framework for Accurate Epileptic Seizure Identification Using EEG Signals

  • 期刊论文

  • 发表日期：

    2026年3月17日

查看全文
  • Dhamodharan Srinivasan,
  • M. Kalpana Chowdary,
  • C. Ganesh,
  • Sultan Ahmad,
  • A. E. M. Eljialy,
  • Mohammed Yousuf Uddin

摘要

The automatic epileptic seizure detection method has recently been developed using electroencephalography (EEG) signals and the machine learning approach. But, the correlation and contextual information among different channels of EEG signals are not obtained precisely from the previous methods. This research introduces a novel approach, the DBN–GC–CRF framework, which integrates glial chains and conditional random fields to specifically address these identified issues. First, feature vectors for statistical measures, power measures, and Zhao-Atlas-Marks domain using a sliding window, are retrieved EEG signals. The DBN–GC–CRF algorithm is then used for extracting the high-level feature and generating its label sequence. Finally, the classification output for EEG data to classify epileptic seizures and non-epileptic seizures is obtained through the K-nearest neighbour algorithm in the decision merge layer. The proposed methodology is implemented on two publicly available epilepsy datasets namely Bonn and CHB-MIT EEG data. The output demonstrates that this approach can extract inter-correlation and contextual information from the EEG channel using the glial chains. The accuracy percentage obtained in the detection of epileptic seizures is 99.10% and F-score is 0.9712 for the Bonn dataset. Similarly, for CHB-MIT, an accuracy of approximately 99.56% and an F-score of 0.9833 were obtained. According to the results, the proposed method outperforms other deep classifiers.