Scholarlink

<p>Magnetic resonance imaging (MRI) plays a pivotal role in the diagnosis and treatment planning of brain tumors. However, accurately segmenting brain tumors from MRI scans remains challenging due to their complex backgrounds and heterogeneous boundaries. Traditional methods often struggle with these complexities, while manual analysis demands substantial time and medical resources, placing a burden on healthcare systems. This paper introduces LP-Swin, a novel end-to-end automated medical image segmentation method that integrates edge-aware information extracted via the Laplacian Pyramid to guide Swin Transformer-encoded features. The proposed model enhances multi-scale feature fusion during decoding, resulting in improved segmentation accuracy, particularly at tumor boundaries. Experimental evaluations on the medical segmentation decathlon (MSD), BraTS 2021 datasets, and automated cardiac diagnosis challenge (ACDC) datasets demonstrate that LP-Swin achieves state-of-the-art performance, with significant improvements in Dice scores and HD95 metrics compared to previous models. Our findings underscore the potential of combining frequency-domain and spatial-domain information in the segmentation of lesion margins in precision medicine imaging. Our code is available at <a href="https://github.com/springbreeze7111/LP-Swin">https://github.com/springbreeze7111/LP-Swin</a>.</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

LP-Swin: enhancing MRI brain tumor segmentation through Laplacian pyramid-guided swin transformer

  • 期刊论文

  • 发表日期：

    2026年7月9日

查看全文
  • Jie Jiang,
  • Dangguo Shao,
  • Lei Ma,
  • Hongjuan Yang

摘要

Magnetic resonance imaging (MRI) plays a pivotal role in the diagnosis and treatment planning of brain tumors. However, accurately segmenting brain tumors from MRI scans remains challenging due to their complex backgrounds and heterogeneous boundaries. Traditional methods often struggle with these complexities, while manual analysis demands substantial time and medical resources, placing a burden on healthcare systems. This paper introduces LP-Swin, a novel end-to-end automated medical image segmentation method that integrates edge-aware information extracted via the Laplacian Pyramid to guide Swin Transformer-encoded features. The proposed model enhances multi-scale feature fusion during decoding, resulting in improved segmentation accuracy, particularly at tumor boundaries. Experimental evaluations on the medical segmentation decathlon (MSD), BraTS 2021 datasets, and automated cardiac diagnosis challenge (ACDC) datasets demonstrate that LP-Swin achieves state-of-the-art performance, with significant improvements in Dice scores and HD95 metrics compared to previous models. Our findings underscore the potential of combining frequency-domain and spatial-domain information in the segmentation of lesion margins in precision medicine imaging. Our code is available at https://github.com/springbreeze7111/LP-Swin.