Scholarlink

<p>Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that necessitates accurate and efficient imaging-based tools for early diagnosis and disease assessment. Although deep learning methods have shown promise in brain MRI analysis, reliable semantic segmentation of Alzheimer Segmentation remains challenging due to anatomical variability, subtle tissue degeneration, and ambiguous boundaries. This paper presents a Fuzzified Efficient U-Net model integrated with a Conditional Random Field (CRF) for precise semantic segmentation of Alzheimer-related regions from brain MRI. The proposed framework incorporates fuzzy logic within the network to model boundary uncertainty and enhance feature representation, while CRF-based refinement enforces spatial consistency in the final segmentation. An Efficient U-Net backbone is employed to maintain computational efficiency without sacrificing segmentation accuracy. The proposed method is evaluated on two publicly available Alzheimer-related MRI datasets. Quantitative results demonstrate consistent performance improvements over state-of-the-art methods, achieving a Dice Similarity Coefficient (DSC) of 92.7%, Jaccard Index (JI) of 86.4%, sensitivity of 91.2%, specificity of 94.8%, and an average Hausdorff Distance (HD) of 2.4 mm. In addition, the framework attains an inference time of 0.87&#xa0;s per scan, indicating its potential suitability for time-sensitive clinical applications.</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

Fuzzified eficient UNet model with conditional random field for semantic segmentation of Alzheimer’s from brain MRI

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月23日

查看全文
  • Nidhi Pandey,
  • Oshin Sharma

摘要

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that necessitates accurate and efficient imaging-based tools for early diagnosis and disease assessment. Although deep learning methods have shown promise in brain MRI analysis, reliable semantic segmentation of Alzheimer Segmentation remains challenging due to anatomical variability, subtle tissue degeneration, and ambiguous boundaries. This paper presents a Fuzzified Efficient U-Net model integrated with a Conditional Random Field (CRF) for precise semantic segmentation of Alzheimer-related regions from brain MRI. The proposed framework incorporates fuzzy logic within the network to model boundary uncertainty and enhance feature representation, while CRF-based refinement enforces spatial consistency in the final segmentation. An Efficient U-Net backbone is employed to maintain computational efficiency without sacrificing segmentation accuracy. The proposed method is evaluated on two publicly available Alzheimer-related MRI datasets. Quantitative results demonstrate consistent performance improvements over state-of-the-art methods, achieving a Dice Similarity Coefficient (DSC) of 92.7%, Jaccard Index (JI) of 86.4%, sensitivity of 91.2%, specificity of 94.8%, and an average Hausdorff Distance (HD) of 2.4 mm. In addition, the framework attains an inference time of 0.87 s per scan, indicating its potential suitability for time-sensitive clinical applications.