Scholarlink

<p>Mitosis in H&amp;E-stained histopathological images is a critical prognostic marker for tumor aggressiveness, aiding in precise diagnosis and treatment planning. Accurate mitosis segmentation is essential for cancer diagnostics and therapeutic decisions. This study proposes an ensemble KANs method with XAI for mitosis segmentation. A hybrid dataset combining MIDOG21 and CCMCT is utilized, followed by enhanced stain normalization using DDPM and CD-SNMF. The ensemble KANs model incorporates XAI to improve interpretability and reliability, fostering trust in AI-assisted diagnostic tools. The proposed method achieved high performance on the hybrid dataset, with Accuracy of 0.905, Precision of 0.89, Recall of 0.88, F1-Score of 0.885, and AJI of 0.75. Experimental results demonstrate the method’s effectiveness in accurately identifying mitotic cells, outperforming existing methods and enhancing digital pathology.</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

An ensemble KANs method with XAI for mitosis segmentation in histopathological images

  • 期刊论文

  • 发表日期：

    2025年7月30日

查看全文
  • Refik Samet,
  • Nooshin Nemati,
  • Emrah Hancer,
  • Serpil Sak,
  • Bilge Ayca Kirmizi

摘要

Mitosis in H&E-stained histopathological images is a critical prognostic marker for tumor aggressiveness, aiding in precise diagnosis and treatment planning. Accurate mitosis segmentation is essential for cancer diagnostics and therapeutic decisions. This study proposes an ensemble KANs method with XAI for mitosis segmentation. A hybrid dataset combining MIDOG21 and CCMCT is utilized, followed by enhanced stain normalization using DDPM and CD-SNMF. The ensemble KANs model incorporates XAI to improve interpretability and reliability, fostering trust in AI-assisted diagnostic tools. The proposed method achieved high performance on the hybrid dataset, with Accuracy of 0.905, Precision of 0.89, Recall of 0.88, F1-Score of 0.885, and AJI of 0.75. Experimental results demonstrate the method’s effectiveness in accurately identifying mitotic cells, outperforming existing methods and enhancing digital pathology.