Scholarlink

<p>Early diagnosis of Alzheimer’s disease (AD) is a significant issue because neural changes related to the disease are slow, discrete, and, in most cases, concealed by standard clinical tests. This study introduces a Multi-Perspective Topological Gramian Angular Field (MPT-GAF) system that has the potential to deliver reliable EEG-based AD classification. EEG signals are nonlinear and non-stationary in nature, which implies that the proposed method represents complex spatiotemporal behavior by encoding in a nonlinear phase space and transforming it into a topology. Pre-processing of EEG records entails the application of a fifth-order Butterworth bandpass filter (0.5–45 Hz) and Daubechies-4 wavelet denoising to eliminate artifacts and finally sub-resolution into five major frequency bands. All bands were sampled into image-like MPT-GAFs, and the phase correlation, nonlinear association, and interregional connection patterns were preserved. A weighted ensemble classifier based on a Convolutional Neural Network (CNN), Support Vector Machine (SVM) with Radial Basis Function (RBF) kernel, Random Forest and Bidirectional Long Short-Term Memory (Bi-LSTM) networks was then used to exploit both spatial and temporal representations. The experimental validation blocks in the OpenNeuro dataset achieved outstanding performance with 98.94% accuracy, 98.96% precision, and 98.94% F1-score with a significant statistical value (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(p &lt; 0.05\)</EquationSource> </InlineEquation>), which is an improvement of 17.3% over the current state-of-the-art methods. The proposed MPT-GAF framework offers a computationally efficient, noninvasive, and clinically interpretable solution to the problem of early stage AD detection through the effective modeling of the nonlinear neurophysiological dynamics of cognitive decline.</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 Multi-modal Ensemble Framework with Topological GAF Features for Accurate EEG-Based Alzheimer’s Diagnosis

  • 期刊论文

  • 发表日期：

    2026年5月6日

查看全文
  • Abhishek Mankar,
  • Suraj,
  • Sandeep Raj

摘要

Early diagnosis of Alzheimer’s disease (AD) is a significant issue because neural changes related to the disease are slow, discrete, and, in most cases, concealed by standard clinical tests. This study introduces a Multi-Perspective Topological Gramian Angular Field (MPT-GAF) system that has the potential to deliver reliable EEG-based AD classification. EEG signals are nonlinear and non-stationary in nature, which implies that the proposed method represents complex spatiotemporal behavior by encoding in a nonlinear phase space and transforming it into a topology. Pre-processing of EEG records entails the application of a fifth-order Butterworth bandpass filter (0.5–45 Hz) and Daubechies-4 wavelet denoising to eliminate artifacts and finally sub-resolution into five major frequency bands. All bands were sampled into image-like MPT-GAFs, and the phase correlation, nonlinear association, and interregional connection patterns were preserved. A weighted ensemble classifier based on a Convolutional Neural Network (CNN), Support Vector Machine (SVM) with Radial Basis Function (RBF) kernel, Random Forest and Bidirectional Long Short-Term Memory (Bi-LSTM) networks was then used to exploit both spatial and temporal representations. The experimental validation blocks in the OpenNeuro dataset achieved outstanding performance with 98.94% accuracy, 98.96% precision, and 98.94% F1-score with a significant statistical value ( \(p < 0.05\) ), which is an improvement of 17.3% over the current state-of-the-art methods. The proposed MPT-GAF framework offers a computationally efficient, noninvasive, and clinically interpretable solution to the problem of early stage AD detection through the effective modeling of the nonlinear neurophysiological dynamics of cognitive decline.