Scholarlink

<p>Progressive neurodegeneration is the cause of Alzheimer’s Disease (AD), an intractable neurological condition. Detecting the condition at an early stage is crucial for both prevention and management. Deep learning methods used to multimodal imaging have been shown to substantially enhance AD recognition in computer-aided diagnosis (CAD). However, while high-quality imaging is essential for precise diagnosis and effective treatment planning, medical scans are frequently affected by noise generated during the acquisition process. In this paper, Two-Dimensional Quantum Morphological Haar Wavelet Transform (2DQMHWT) denoising is introduced for medical images. The 2DQHWT decomposition technique divides input images into developing a multiscale representation in many frequency bands. Features are captured at several levels in this depiction, enabling efficient extraction of both local and global image structures. Additionally, 2DQMHWT preserves spatial localization details during the decomposition process. Two-Dimensional Inverse Quantum Morphological Haar Wavelet Transform (2DIQMHWT) is also introduced to accurately reconstruct the noise-removed image. Feature Fusion Transformer (FFT) model is the process of integrating features extracted from various sources magnetic resonance imaging (MRI) and positron emission tomography (PET) into a one, more detailed feature vector. MRI features are extracted by Weight Optimized DenseNet-169 (WODenseNet-169) model. PET features are extracted byContext TransformationInception-ResNet-V3 (CTIREV3) model. Attention Weight Bidirectional Long Short-Term Memory (AWBiLSTM) two LSTMs make up the classifier, a sequence processing model, one of which takes input flowing forward and the other backward. AW reflects the features of the images; consequently, the model might be able to identify important image features in the set being evaluated. The AW approach enhances the feature maps of the classifier, with experiments carried out using the Global Alzheimer’s Association Interactive Network (GAAIN) dataset derived from neuroimaging sources. Including MRI and PET, show that the suggested approach works more effectively than the most advanced techniques for diagnosing AD.</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

Two-Dimensional Quantum Morphological Haar Wavelet Transform (2-DQMHWT) Denoising and Feature Fusion Transformer (FFT) Model for Alzheimer Disease Diagnosis

  • 期刊论文

  • 发表日期：

    2026年1月13日

查看全文
  • Sowgandh Krishnaa Nandamuri,
  • V. D. Ambethkumar

摘要

Progressive neurodegeneration is the cause of Alzheimer’s Disease (AD), an intractable neurological condition. Detecting the condition at an early stage is crucial for both prevention and management. Deep learning methods used to multimodal imaging have been shown to substantially enhance AD recognition in computer-aided diagnosis (CAD). However, while high-quality imaging is essential for precise diagnosis and effective treatment planning, medical scans are frequently affected by noise generated during the acquisition process. In this paper, Two-Dimensional Quantum Morphological Haar Wavelet Transform (2DQMHWT) denoising is introduced for medical images. The 2DQHWT decomposition technique divides input images into developing a multiscale representation in many frequency bands. Features are captured at several levels in this depiction, enabling efficient extraction of both local and global image structures. Additionally, 2DQMHWT preserves spatial localization details during the decomposition process. Two-Dimensional Inverse Quantum Morphological Haar Wavelet Transform (2DIQMHWT) is also introduced to accurately reconstruct the noise-removed image. Feature Fusion Transformer (FFT) model is the process of integrating features extracted from various sources magnetic resonance imaging (MRI) and positron emission tomography (PET) into a one, more detailed feature vector. MRI features are extracted by Weight Optimized DenseNet-169 (WODenseNet-169) model. PET features are extracted byContext TransformationInception-ResNet-V3 (CTIREV3) model. Attention Weight Bidirectional Long Short-Term Memory (AWBiLSTM) two LSTMs make up the classifier, a sequence processing model, one of which takes input flowing forward and the other backward. AW reflects the features of the images; consequently, the model might be able to identify important image features in the set being evaluated. The AW approach enhances the feature maps of the classifier, with experiments carried out using the Global Alzheimer’s Association Interactive Network (GAAIN) dataset derived from neuroimaging sources. Including MRI and PET, show that the suggested approach works more effectively than the most advanced techniques for diagnosing AD.