Scholarlink

<p>Federated learning (FL) is a privacy-preserving paradigm that enables collaborative training of machine learning models using data stored across multiple silos. Alzheimer’s disease (AD) is a debilitating neurological condition and a major public health concern that demands reliable and efficient diagnostic methods. This study presents a novel framework that integrates EEG-based FL with Generative Adversarial Networks (GANs), Finite Impulse Response (FIR) filtering, and a Federated Learning-based Validation Searching Technique (FLVST) to improve the accuracy and robustness of AD classification. To overcome the challenges of limited EEG data availability, privacy concerns, and suboptimal model selection, our framework integrates FIR filtering for domain-specific feature extraction, GANs for augmenting training data, federated learning for privacy-aware distributed training, and FLVST for robust model validation and optimization. This combination provides a comprehensive solution, advancing both technical and ethical aspects of Alzheimer’s disease classification. In the feature extraction phase, FIR filtering isolates frequency-domain characteristics critical for distinguishing between AD patients and healthy controls. GANs augment the limited dataset by generating synthetic EEG samples, enhancing training diversity and reducing overfitting. FL ensures data privacy while enabling multi-institutional collaboration and FLVST optimizes model selection through validation searching. Simulation results show that the proposed framework achieves an accuracy of 99.67% and a sensitivity of 97.99% for early AD detection, with reduced training time compared to conventional approaches. Furthermore, the FL-based design makes the model scalable and adaptable to heterogeneous data distributions, facilitating deployment in real-world clinical environments.</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

Federated GAN based framework for Alzheimer disease classification using finite impulse response filter techniques

  • 期刊论文

  • 发表日期：

    2026年4月21日

查看全文
  • R. Kingsly Stephen,
  • J. Faritha Banu

摘要

Federated learning (FL) is a privacy-preserving paradigm that enables collaborative training of machine learning models using data stored across multiple silos. Alzheimer’s disease (AD) is a debilitating neurological condition and a major public health concern that demands reliable and efficient diagnostic methods. This study presents a novel framework that integrates EEG-based FL with Generative Adversarial Networks (GANs), Finite Impulse Response (FIR) filtering, and a Federated Learning-based Validation Searching Technique (FLVST) to improve the accuracy and robustness of AD classification. To overcome the challenges of limited EEG data availability, privacy concerns, and suboptimal model selection, our framework integrates FIR filtering for domain-specific feature extraction, GANs for augmenting training data, federated learning for privacy-aware distributed training, and FLVST for robust model validation and optimization. This combination provides a comprehensive solution, advancing both technical and ethical aspects of Alzheimer’s disease classification. In the feature extraction phase, FIR filtering isolates frequency-domain characteristics critical for distinguishing between AD patients and healthy controls. GANs augment the limited dataset by generating synthetic EEG samples, enhancing training diversity and reducing overfitting. FL ensures data privacy while enabling multi-institutional collaboration and FLVST optimizes model selection through validation searching. Simulation results show that the proposed framework achieves an accuracy of 99.67% and a sensitivity of 97.99% for early AD detection, with reduced training time compared to conventional approaches. Furthermore, the FL-based design makes the model scalable and adaptable to heterogeneous data distributions, facilitating deployment in real-world clinical environments.