Scholarlink

Accurate segmentation mask generation is critical for single-cell analysis workflows. While established semi-automated tools require expert intervention, emerging approaches aim to eliminate human guidance through fully automatic segmentation models. However, the suitability of automatically generated cell segmentation masks as reliable alternatives to expert annotations remains uncertain. This study evaluates different imaging mass cytometry (IMC) datasets by feeding them to a variety of generalist cell segmentation models and comparing the outputs with corresponding segmentation masks. Performance is assessed using instance segmentation metrics which are also viewed in the light of an upper bound determined by inter-annotator agreement.

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

Quantifying Inter-annotator Agreement and Generalist Model Limitations in Imaging Mass Cytometry Single Cell Segmentation

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Johannes Schuiki,
  • Markus Steiner,
  • Heinz Hofbauer,
  • Stephan Drothler,
  • Giulia Pessina,
  • Richard Greil,
  • Nadja Zaborsky,
  • Andreas Uhl

摘要

Accurate segmentation mask generation is critical for single-cell analysis workflows. While established semi-automated tools require expert intervention, emerging approaches aim to eliminate human guidance through fully automatic segmentation models. However, the suitability of automatically generated cell segmentation masks as reliable alternatives to expert annotations remains uncertain. This study evaluates different imaging mass cytometry (IMC) datasets by feeding them to a variety of generalist cell segmentation models and comparing the outputs with corresponding segmentation masks. Performance is assessed using instance segmentation metrics which are also viewed in the light of an upper bound determined by inter-annotator agreement.