Scholarlink

<p>The non-stationarity of EEG recordings, between-subject variability and the temporal dependencies of the seizure patterns make detection of epileptic seizures a difficult task from electroencephalogram (EEG) signals. To overcome these difficulties, this paper presents an Artificial Rabbits Optimization-tuned Attention-Based Bidirectional Recurrent Neural Network (AROABiRNN-ESD) which is optimized by Artificial Rabbits Optimization (ARO) algorithm for accurate, robust epileptic seizure detection. To handle these challenges, this paper introduces an Artificial Rabbits Optimization-tuned Attention-Based Bidirectional Recurrent Neural Network (AROABiRNN-ESD) optimized by the Artificial Rabbits Optimization (ARO) algorithm for accurate and robust epileptic seizure detection. The proposed framework proposes bidirectional temporal feature learning along with attention mechanism to effectively capture the forward and backward EEG dependency and focus the most discriminative EEG signal segments related to seizures. Furthermore, systematic hyperparameter tuning (ART) is used to increase the stability of the convergence, minimize manual parameter selection and improve the model generalization. Experimental results on the UCI Epileptic Seizure Recognition database indicate that the proposed AROABiRNN-ESD model outperforms the traditional machine learning methods and existing deep learning approaches in terms of accuracy, precision, sensitivity, specificity, F1 score and AUC with a score of 97.0%, 96.6%, 95.8%, 97.9%, 96.2% and 98.4% respectively. Comparative evaluation and ablation analysis additionally substantiate the role of attention-guided temporal learning and also optimization of parameters. In conclusion, the proposed framework helps to design an automated seizure detection system using EEG signals that is reliable, high performing and computationally efficient.</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

A Hyperparameter-Tuned Attention-based BiRNN framework using Artificial Rabbits Optimization for Epileptic Seizure Detection

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月4日

查看全文
  • J. Viswanath,
  • S. Annamalai

摘要

The non-stationarity of EEG recordings, between-subject variability and the temporal dependencies of the seizure patterns make detection of epileptic seizures a difficult task from electroencephalogram (EEG) signals. To overcome these difficulties, this paper presents an Artificial Rabbits Optimization-tuned Attention-Based Bidirectional Recurrent Neural Network (AROABiRNN-ESD) which is optimized by Artificial Rabbits Optimization (ARO) algorithm for accurate, robust epileptic seizure detection. To handle these challenges, this paper introduces an Artificial Rabbits Optimization-tuned Attention-Based Bidirectional Recurrent Neural Network (AROABiRNN-ESD) optimized by the Artificial Rabbits Optimization (ARO) algorithm for accurate and robust epileptic seizure detection. The proposed framework proposes bidirectional temporal feature learning along with attention mechanism to effectively capture the forward and backward EEG dependency and focus the most discriminative EEG signal segments related to seizures. Furthermore, systematic hyperparameter tuning (ART) is used to increase the stability of the convergence, minimize manual parameter selection and improve the model generalization. Experimental results on the UCI Epileptic Seizure Recognition database indicate that the proposed AROABiRNN-ESD model outperforms the traditional machine learning methods and existing deep learning approaches in terms of accuracy, precision, sensitivity, specificity, F1 score and AUC with a score of 97.0%, 96.6%, 95.8%, 97.9%, 96.2% and 98.4% respectively. Comparative evaluation and ablation analysis additionally substantiate the role of attention-guided temporal learning and also optimization of parameters. In conclusion, the proposed framework helps to design an automated seizure detection system using EEG signals that is reliable, high performing and computationally efficient.