{
  "id": 19533594,
  "title": "Training AI to see cells in context improves detection in pathology images",
  "url": "https://urgent.news/2026/11/11/training-ai-to-see-cells-in-context-improves-detection-in-pathology",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-11T18:13:30.000Z",
  "source": {
    "name": "Medical Xpress",
    "slug": "medical-xpress",
    "url": "https://medicalxpress.com/news/2026-11-ai-cells-context-pathology-images.html"
  },
  "original_language": "en",
  "account": "Researchers at Koç University have developed a technique that enhances AI's ability to identify cells in digital pathology images. The approach, called topology-guided hard example mining (TG-HEM), teaches AI to consider how cells are arranged within tissue rather than focusing solely on individual pixel errors. Published in Medical Image Analysis, the study introduces TG-HEM, which helps models concentrate on regions where predictions fail to capture the spatial organization of cells. By incorporating information about the surrounding cell arrangement into the training process, TG-HEM improves cell counting and localization compared to baseline models and existing hard example mining approaches. The technique uses topological data analysis to quantify differences in cell organization, assigning greater weight to regions with larger structural discrepancies. TG-HEM also places additional emphasis on difficult pixels within those regions, allowing the model to learn from both local detection errors and broader differences in cell arrangement. The researchers evaluated the approach using eight AI models and two pathology datasets, demonstrating improvements in cell detection without additional computational overhead. The study highlights the potential of more accurate cell detection to strengthen automated measurements in digital pathology, providing more reliable image measurements that could support more consistent tissue assessment. The KUCell dataset and implementation code have been released to enable other researchers to evaluate and build upon the approach.",
  "summary": "Artificial intelligence can help identify cells in digital pathology images, but densely packed tissue remains a challenge: neighboring cells may be merged, counted twice or missed entirely. Researchers at Koç University have developed a training strategy that improves detection by teaching AI to consider how cells are arranged within the tissue.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}