Scholarlink

<p>A chronic and irreparable neurodegenerative condition known as Alzheimer’s disease predominantly affects the neurons in the brain and causes a slow but significant deterioration in the normal cognitive function of an individual. A detection framework for Alzheimer’s disease using deep architectures from the Magnetic Resonance Imaging (MRI) images is implemented. The MRI images are initially acquired from standard databases. The acquired images are inputted into the Vision Transformer-based Weighted Fusion of Convolution Neural Networks (ViT-FusCNN). Instead of using traditional convolutional and pooling layers in the Residual Network (ResNet), the Visual Geometry Group (VGG), CNN architecture, and Inception models are utilized as the feature extractor to obtain the valuable features from the input MRI images. The multi-scale weighted fusion of features technique is applied to multiply the features with weights, for which the weights are tuned utilizing the newly Enhanced Mother Optimization Algorithm (EMOA). Then, the disease classification is done by utilizing the remaining layers of the CNN. Effective detection and classification of Alzheimer’s disease is obtained by this modified architecture. The accuracy of the proposed EMOA-ViT-FusCNN is 94.31. Accordingly, the accuracy values of the existing works such as VGG16, Inception-Net, ResNet, and, ViT-FusCNN are 86.87, 88.65, 90.56, and 92.31. Several performance measures are used to examine and assess the suggested model’s functioning.</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

ViT-FusCNN: vision transformer-based weighted fusion of convolution networks for Alzheimer’s disease detection from MRI images with enhanced mother optimization

  • 期刊论文

  • 发表日期：

    2025年5月12日

查看全文
  • Revathi Mohan,
  • Rajesh Arunachalam,
  • Neha Verma,
  • Thomas Bernatin

摘要

A chronic and irreparable neurodegenerative condition known as Alzheimer’s disease predominantly affects the neurons in the brain and causes a slow but significant deterioration in the normal cognitive function of an individual. A detection framework for Alzheimer’s disease using deep architectures from the Magnetic Resonance Imaging (MRI) images is implemented. The MRI images are initially acquired from standard databases. The acquired images are inputted into the Vision Transformer-based Weighted Fusion of Convolution Neural Networks (ViT-FusCNN). Instead of using traditional convolutional and pooling layers in the Residual Network (ResNet), the Visual Geometry Group (VGG), CNN architecture, and Inception models are utilized as the feature extractor to obtain the valuable features from the input MRI images. The multi-scale weighted fusion of features technique is applied to multiply the features with weights, for which the weights are tuned utilizing the newly Enhanced Mother Optimization Algorithm (EMOA). Then, the disease classification is done by utilizing the remaining layers of the CNN. Effective detection and classification of Alzheimer’s disease is obtained by this modified architecture. The accuracy of the proposed EMOA-ViT-FusCNN is 94.31. Accordingly, the accuracy values of the existing works such as VGG16, Inception-Net, ResNet, and, ViT-FusCNN are 86.87, 88.65, 90.56, and 92.31. Several performance measures are used to examine and assess the suggested model’s functioning.