Scholarlink

<p>Epileptic seizures emerge unexpectedly within the nervous system creating substantial threats to the safety of people and public. Various automated epileptic seizure detection approaches based on electroencephalogram (EEG) data now exist, yet most current methods focus primarily on seizure detection instead of predicting seizure types for personalized treatments. The authors developed an automatic seizure prediction system through Swish Beta activated LeNet-5 classifier (SWA-LeNet) for addressing the multi-type of epileptic seizure identification problem in multi-channel EEG data. The suggested method operates EEG data through two sequential phases. The first phase of signal processing utilizes Fourier-Legendre Polynomial (FLP) to generate coefficients that HgLEO optimizes for ictal as well as seizure episode prediction purposes. During the second stage of processing gamma frequency signals pass through combination FIR and IIR filters before Phase Amplitude based Synchro squeezing Time–frequency Decomposition (PA-STD) decomposes them using Intrinsic Mode Functions extraction. The combination of phase outputs contains features that undergo t-distributed Stochastic Neighbour Embedding (t-SNE) for dimensionality reduction before extraction. The SWA-LeNet classifier functions as the final decision system to divide eight different kinds of epileptic seizures. The experimental evaluation on standard EEG datasets shows the proposed method reaches an accuracy of 99% along with precision at 99.1% while recall stands at 99.02% and F-measure reaches 99% which exceeds current models. The SWA-LeNet model demonstrates excellent potential to help neurologists make precise early screenings of different epileptic seizure types.</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

Automatic Diverse Epileptic Seizure Detection Model Using SWA-LeNet

  • 期刊论文

  • 发表日期：

    2025年9月4日

查看全文
  • Milind Ashok Natu,
  • Mrinal Rahul Bachute

摘要

Epileptic seizures emerge unexpectedly within the nervous system creating substantial threats to the safety of people and public. Various automated epileptic seizure detection approaches based on electroencephalogram (EEG) data now exist, yet most current methods focus primarily on seizure detection instead of predicting seizure types for personalized treatments. The authors developed an automatic seizure prediction system through Swish Beta activated LeNet-5 classifier (SWA-LeNet) for addressing the multi-type of epileptic seizure identification problem in multi-channel EEG data. The suggested method operates EEG data through two sequential phases. The first phase of signal processing utilizes Fourier-Legendre Polynomial (FLP) to generate coefficients that HgLEO optimizes for ictal as well as seizure episode prediction purposes. During the second stage of processing gamma frequency signals pass through combination FIR and IIR filters before Phase Amplitude based Synchro squeezing Time–frequency Decomposition (PA-STD) decomposes them using Intrinsic Mode Functions extraction. The combination of phase outputs contains features that undergo t-distributed Stochastic Neighbour Embedding (t-SNE) for dimensionality reduction before extraction. The SWA-LeNet classifier functions as the final decision system to divide eight different kinds of epileptic seizures. The experimental evaluation on standard EEG datasets shows the proposed method reaches an accuracy of 99% along with precision at 99.1% while recall stands at 99.02% and F-measure reaches 99% which exceeds current models. The SWA-LeNet model demonstrates excellent potential to help neurologists make precise early screenings of different epileptic seizure types.