Scholarlink

<p>An abnormal electrical discharge in the brain that disrupts normal brain activity is an Epileptic seizure. These electrical disturbances cause changes in physical, sensory, or behavioral functions, including loss of consciousness, convulsions, and muscle twitching. Early diagnosis and accurate classification of epileptic seizures is essential for effective treatment. To address this, an S-Gaussian-Deep Neuro Fuzzy Network (SG-DNFN) is proposed for epileptic seizure classification. The Electroencephalography (EEG) signal is acquired, and features like Power Spectral Density (PSD), Corrected Conditional Entropy (CCE), spectral kurtosis, logarithmic band power, and pitch chroma are extracted. Feature selection is accomplished using Analysis of Variance (ANOVA), and classification is done using SG-DNFN, where the Membership Function (MF) for Deep Neuro Fuzzy Network (DNFN) is modified using S function and Gaussian MF. The proposed SG-DNFN achieved 94.037% accuracy, 92.903% True Positive Rate (TPR), 94.732% True Negative Rate (TNR), and 91.592% Negative predictive value (NPV).</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

SG-DNFN: S-Gaussian Membership Function-Based Deep Neuro Fuzzy Network for Epileptic Seizure Classification Using EEG Signal

  • 期刊论文

  • 发表日期：

    2025年8月25日

查看全文
  • Thejaswini B M,
  • George Glan Devadhas,
  • T Y Satheesha

摘要

An abnormal electrical discharge in the brain that disrupts normal brain activity is an Epileptic seizure. These electrical disturbances cause changes in physical, sensory, or behavioral functions, including loss of consciousness, convulsions, and muscle twitching. Early diagnosis and accurate classification of epileptic seizures is essential for effective treatment. To address this, an S-Gaussian-Deep Neuro Fuzzy Network (SG-DNFN) is proposed for epileptic seizure classification. The Electroencephalography (EEG) signal is acquired, and features like Power Spectral Density (PSD), Corrected Conditional Entropy (CCE), spectral kurtosis, logarithmic band power, and pitch chroma are extracted. Feature selection is accomplished using Analysis of Variance (ANOVA), and classification is done using SG-DNFN, where the Membership Function (MF) for Deep Neuro Fuzzy Network (DNFN) is modified using S function and Gaussian MF. The proposed SG-DNFN achieved 94.037% accuracy, 92.903% True Positive Rate (TPR), 94.732% True Negative Rate (TNR), and 91.592% Negative predictive value (NPV).