Scholarlink

<p>As a major worldwide health issue, liver tumor necessitates prompt diagnosis and precise treatment. The use of medical imaging, especially Computed Tomography, is essential for classifying abnormalities in the liver. Automating liver tumor segmentation is difficult, owing to liver's intricate structure and variability in illness prognosis. These problems are overcome with the aid of introducing a novel WRAU-Swin-Trans with optimized model for liver tumor segmentation. To improve the quality of input image and standardize input data, preprocessing processes are essential, which include image enhancement using Modified Contrast Limited Adaptive Histogram Equalization and reshaping. A novel WRAU-Swin-Trans model, which integrates wavelet transformation, attention scale mechanism with residual U-Net and swin transformer to segment data after preprocessing. For obtaining better outcomes of liver tumor segmentation, this study utilizes Dwarf Mongoose Optimization (DMO) model, fine-tuning the parameters of WRAU-Swin-Trans system, thereby improving predictive performance of tumour. For evaluating the performance of DMO optimized Wave-RAU-Swin-Trans model, python software is implemented. The obtained outcome is compared with the state-of-the-art models that visible that higher accuracy of (96.98%) with minimal errors and better performance analysis. The findings signify that with the use of proposed work, liver tumor detection and treatment planning is done early, whereby preventing lots of death of both women and men in worldwide.</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

Integrating swin transformer and attention scaling into residual U-net for accurate liver tumor segmentation in medical imaging

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月2日

查看全文
  • Saravanan G,
  • Geetha Palaniappan,
  • Mathana J. M

摘要

As a major worldwide health issue, liver tumor necessitates prompt diagnosis and precise treatment. The use of medical imaging, especially Computed Tomography, is essential for classifying abnormalities in the liver. Automating liver tumor segmentation is difficult, owing to liver's intricate structure and variability in illness prognosis. These problems are overcome with the aid of introducing a novel WRAU-Swin-Trans with optimized model for liver tumor segmentation. To improve the quality of input image and standardize input data, preprocessing processes are essential, which include image enhancement using Modified Contrast Limited Adaptive Histogram Equalization and reshaping. A novel WRAU-Swin-Trans model, which integrates wavelet transformation, attention scale mechanism with residual U-Net and swin transformer to segment data after preprocessing. For obtaining better outcomes of liver tumor segmentation, this study utilizes Dwarf Mongoose Optimization (DMO) model, fine-tuning the parameters of WRAU-Swin-Trans system, thereby improving predictive performance of tumour. For evaluating the performance of DMO optimized Wave-RAU-Swin-Trans model, python software is implemented. The obtained outcome is compared with the state-of-the-art models that visible that higher accuracy of (96.98%) with minimal errors and better performance analysis. The findings signify that with the use of proposed work, liver tumor detection and treatment planning is done early, whereby preventing lots of death of both women and men in worldwide.