Scholarlink

<p>The increasing prevalence of Alzheimer’s disease (AD) in the elderly population highlights the urgent need for improved early detection tools. Diagnostic procedures nowadays depend significantly on evaluations, cognitive testing, and cutting-edge brain imaging techniques like positron emission tomography (PET) and magnetic resonance imaging (MRI). Using positron emission tomography (PET) and magnetic resonance imaging (MRI) separately may reduce the accuracy of early-stage AD diagnoses, even though PET and MRI give information on brain activity. Because AD-related brain alterations are so subtle, successful detection and analysis need complex, integrated methods. To overcome these challenges, the study introduces a method that combines metabolic data acquired from positron emission tomography scans with information from magnetic resonance imaging (MRI). This work introduces an Adaptive pulse coupled neural network (APCNN) for Medical image fusion, which dynamically adjusts parameters to improve the clarity and preservation of details in fused images, mainly when dealing with medical images. By retaining elements, like edges and textures, pulse-coupled neural networks contribute to diagnosing AD. A multi-scale feature extraction network built using a residual convolutional network and a pyramid-pooling module is proposed to manage the challenges of multi-scale feature extraction in AD neuroimaging data. Multi-scale feature extraction network overcomes the imbalance in positive and negative sample distribution by integrating target loss into its classifier, thereby considerably boosting the accuracy of feature extraction from multi-modal fused images. Finally, a novel meta-heuristic Feature Selection approach that integrates Thermal exchange optimization with the artificial hummingbird algorithm is presented to handle the high-dimensional feature space of neuroimaging data. The results show that the proposed integrated method can improve feature extraction and classification accuracy over different datasets, which supports the effectiveness of these proposed methods and highlights their potential for improving early AD detection.</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

Adaptive neurofusion: unveiling early Alzheimer’s signs with multi-scale feature extraction

  • 期刊论文

  • 发表日期：

    2025年9月8日

查看全文
  • S. Rajeswari,
  • K. Swathi

摘要

The increasing prevalence of Alzheimer’s disease (AD) in the elderly population highlights the urgent need for improved early detection tools. Diagnostic procedures nowadays depend significantly on evaluations, cognitive testing, and cutting-edge brain imaging techniques like positron emission tomography (PET) and magnetic resonance imaging (MRI). Using positron emission tomography (PET) and magnetic resonance imaging (MRI) separately may reduce the accuracy of early-stage AD diagnoses, even though PET and MRI give information on brain activity. Because AD-related brain alterations are so subtle, successful detection and analysis need complex, integrated methods. To overcome these challenges, the study introduces a method that combines metabolic data acquired from positron emission tomography scans with information from magnetic resonance imaging (MRI). This work introduces an Adaptive pulse coupled neural network (APCNN) for Medical image fusion, which dynamically adjusts parameters to improve the clarity and preservation of details in fused images, mainly when dealing with medical images. By retaining elements, like edges and textures, pulse-coupled neural networks contribute to diagnosing AD. A multi-scale feature extraction network built using a residual convolutional network and a pyramid-pooling module is proposed to manage the challenges of multi-scale feature extraction in AD neuroimaging data. Multi-scale feature extraction network overcomes the imbalance in positive and negative sample distribution by integrating target loss into its classifier, thereby considerably boosting the accuracy of feature extraction from multi-modal fused images. Finally, a novel meta-heuristic Feature Selection approach that integrates Thermal exchange optimization with the artificial hummingbird algorithm is presented to handle the high-dimensional feature space of neuroimaging data. The results show that the proposed integrated method can improve feature extraction and classification accuracy over different datasets, which supports the effectiveness of these proposed methods and highlights their potential for improving early AD detection.