Scholarlink

<p>This paper addresses key challenges in diagnosing Alzheimer’s disease (AD) using raw EEG signals, particularly the difficulty in identifying important features due to noise, signal abnormalities, and individual variability in brain function. These factors hinder consistent feature extraction and reliable classification across diverse patients, necessitating robust pre-processing, feature extraction, and optimization methods. To overcome these challenges, the study introduces a novel framework called Savitzky–Golay Denoising and Discrete Cosine Krawtchouk–Tchebichef Transform Optimized by Pied Kingfisher Algorithm (SGD-DCKTT-PKA). The proposed method begins by applying Savitzky–Golay Denoising (SGD) to enhance EEG signal quality, followed by a unique Normalized Discrete Cosine Krawtchouk–Tchebichef Transform (NDC-KTT) that extracts discriminative features crucial for identifying stages of cognitive decline. Subsequently, a Dual-Level Contextual Attention mechanism integrated with Cosine Convolutional Neural Network (DLCA-CCNNet) is employed for classification, providing robust feature learning. To further enhance diagnostic accuracy, the Pied Kingfisher Optimizer (PKO) is applied to fine-tune the model’s parameters. The technique is evaluated using the publicly available Dementia EEG Dataset, with a Python-based testing environment confirming the framework’s effectiveness. The SGD-DCKTT-PKA model achieved an outstanding classification accuracy of 99.9% and sensitivity of 99.8%, showcasing its potential to accurately differentiate between stages of cognitive impairment. These results highlight the method’s clinical relevance and suggest that such automated EEG analysis tools could complement or eventually substitute traditional clinical assessments for Alzheimer’s diagnosis. The study emphasizes the importance of repeated testing across varied datasets to ensure the model’s reliability, scalability, and applicability in real-world diagnostic settings.</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

EEG-Based Alzheimer’s Disease Diagnosis Using Savitzky–Golay Denoising and Discrete Cosine Krawtchouk–Tchebichef Transform Optimized by Pied Kingfisher Algorithm

  • 期刊论文

  • 发表日期：

    2025年7月24日

查看全文
  • A. Harshavardhan,
  • V. Jeyakrishnan,
  • Krishna Prakash Arunachalam,
  • Sampathirao Suneetha

摘要

This paper addresses key challenges in diagnosing Alzheimer’s disease (AD) using raw EEG signals, particularly the difficulty in identifying important features due to noise, signal abnormalities, and individual variability in brain function. These factors hinder consistent feature extraction and reliable classification across diverse patients, necessitating robust pre-processing, feature extraction, and optimization methods. To overcome these challenges, the study introduces a novel framework called Savitzky–Golay Denoising and Discrete Cosine Krawtchouk–Tchebichef Transform Optimized by Pied Kingfisher Algorithm (SGD-DCKTT-PKA). The proposed method begins by applying Savitzky–Golay Denoising (SGD) to enhance EEG signal quality, followed by a unique Normalized Discrete Cosine Krawtchouk–Tchebichef Transform (NDC-KTT) that extracts discriminative features crucial for identifying stages of cognitive decline. Subsequently, a Dual-Level Contextual Attention mechanism integrated with Cosine Convolutional Neural Network (DLCA-CCNNet) is employed for classification, providing robust feature learning. To further enhance diagnostic accuracy, the Pied Kingfisher Optimizer (PKO) is applied to fine-tune the model’s parameters. The technique is evaluated using the publicly available Dementia EEG Dataset, with a Python-based testing environment confirming the framework’s effectiveness. The SGD-DCKTT-PKA model achieved an outstanding classification accuracy of 99.9% and sensitivity of 99.8%, showcasing its potential to accurately differentiate between stages of cognitive impairment. These results highlight the method’s clinical relevance and suggest that such automated EEG analysis tools could complement or eventually substitute traditional clinical assessments for Alzheimer’s diagnosis. The study emphasizes the importance of repeated testing across varied datasets to ensure the model’s reliability, scalability, and applicability in real-world diagnostic settings.