Scholarlink

<p>Alzheimer's disease (AD) is a degenerative brain condition that is often difficult to diagnose in its early stages. This research explores the utilization of deep learning methods to enhance early AD diagnosis by combining hippocampal subfield volumes and cognitive assessments. Recent neuroimaging advances allow accurate hippocampal subfields measurement, while cognitive assessments show the functional consequences of neurodegeneration. This research proposes a novel Feature Enhanced Wasserstein Generative Adversarial Transformer-based Adaptive Grey Cosine Evolution (FEWGAT-AGCE) approach to improve diagnostic accuracy. The approach uses the feature enhanced cascaded DenseNet121 (FE-CDNet) for feature extraction which combines cascaded convolutional neural network (cascaded CNN) with DenseNet121 to extract hierarchical features from pre-processed data. This model is augmented by a feature enhancement module (FEM) that incorporates global attention mechanisms to improve feature representation. A Wasserstein generative adversarial network with gradient penalty (WGANGP) is employed to generate high-quality synthetic data, which enhances model training stability and performance. The cross domain transformer (CDTformer) refines AD classification and detection, while hyperparameter optimization is performed using Adaptive Levy Grey Wolf Optimization with Cosine Evolution Strategy (ALGW-CES). The FEWGAT-AGCE method achieves a diagnostic accuracy of 98.72% and a precision of 97.59%, which demonstratessignificant improvements in AD early detection and classification leading to more timely and effective interventions.</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

FEWGAT-AGCE: a novel deep learning approach for early Alzheimer's disease detection using hippocampal subfields and cognitive assessments

  • 期刊论文

  • 发表日期：

    2025年9月10日

查看全文
  • M. Thamizharasi,
  • M. Lakshmi

摘要

Alzheimer's disease (AD) is a degenerative brain condition that is often difficult to diagnose in its early stages. This research explores the utilization of deep learning methods to enhance early AD diagnosis by combining hippocampal subfield volumes and cognitive assessments. Recent neuroimaging advances allow accurate hippocampal subfields measurement, while cognitive assessments show the functional consequences of neurodegeneration. This research proposes a novel Feature Enhanced Wasserstein Generative Adversarial Transformer-based Adaptive Grey Cosine Evolution (FEWGAT-AGCE) approach to improve diagnostic accuracy. The approach uses the feature enhanced cascaded DenseNet121 (FE-CDNet) for feature extraction which combines cascaded convolutional neural network (cascaded CNN) with DenseNet121 to extract hierarchical features from pre-processed data. This model is augmented by a feature enhancement module (FEM) that incorporates global attention mechanisms to improve feature representation. A Wasserstein generative adversarial network with gradient penalty (WGANGP) is employed to generate high-quality synthetic data, which enhances model training stability and performance. The cross domain transformer (CDTformer) refines AD classification and detection, while hyperparameter optimization is performed using Adaptive Levy Grey Wolf Optimization with Cosine Evolution Strategy (ALGW-CES). The FEWGAT-AGCE method achieves a diagnostic accuracy of 98.72% and a precision of 97.59%, which demonstratessignificant improvements in AD early detection and classification leading to more timely and effective interventions.