Scholarlink

Polarization is an intrinsic property of light, and the Mueller matrix offers a comprehensive description of its interaction with matter. Mueller matrix imaging is a label-free, noninvasive, multiscale technique that provides quantitative, subwavelength information, ideal for probing complex pathological tissues. Yet, extracting disease-specific polarization features that remain robust to experimental variations and sample orientation has proved challenging. This chapter introduces both supervised and unsupervised machine-learning frameworks for pixel-level polarization feature extraction and representation, specifically designed to integrate polarization imaging and digital-pathology workflows. Together, they reveal “invisible” microstructures as explicit polarization features, laying the foundation for polarization-based digital pathology.

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

Machine Learning for Feature Extraction in Mueller Polarimetry Based Digital Pathology

  • 章节

  • 发表日期：

    2026年

查看全文
  • Valery V. Tuchin,
  • Tatiana Novikova,
  • Lihong V. Wang,
  • Dmitry A. Zimnyakov,
  • Hui Ma,
  • Marina V. Alonova,
  • Jiachen Wan

摘要

Polarization is an intrinsic property of light, and the Mueller matrix offers a comprehensive description of its interaction with matter. Mueller matrix imaging is a label-free, noninvasive, multiscale technique that provides quantitative, subwavelength information, ideal for probing complex pathological tissues. Yet, extracting disease-specific polarization features that remain robust to experimental variations and sample orientation has proved challenging. This chapter introduces both supervised and unsupervised machine-learning frameworks for pixel-level polarization feature extraction and representation, specifically designed to integrate polarization imaging and digital-pathology workflows. Together, they reveal “invisible” microstructures as explicit polarization features, laying the foundation for polarization-based digital pathology.