BigCell: Generating Gigapixel Whole-Slide Images

Key points

• Whole-slide images (WSIs) are gigapixel-scale digital scans of histopathology tissue and serve as the primary basis for cancer diagnosis.1
• Conditioning on embeddings from pathology foundation models and using an inference-time sliding-window pipeline, BigCell generates coherent images of up to 32,768$\times$32,768 pixels, outperforming prior tile generators on both tile-level and slide-level fidelity metrics.1
• To make gigapixel synthesis practical, we introduce an adaptive scheduling strategy that prioritizes important slide regions, speeding up image generation by over 3\ttimes with minimal quality degradation.1
• Furthermore, we propose the first text-to-image synthesis at 8k resolution by training a separate model that produces dense conditioning grids directly from pathology text reports.1

Sources (2)

• [1]BigCell: Generating Gigapixel Whole-Slide Images
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 07:33 PM

  We present BigCell, the first approach to generate coherent whole-slide histopathology images at native gigapixel resolution.

  Whole-slide images (WSIs) are gigapixel-scale digital scans of histopathology tissue and serve as the primary basis for cancer diagnosis.

• [2]BigCell: Generating Gigapixel Whole-Slide Images
  Hugging Face Daily Papers · Oct 9, 12:00 AM · same content

Extractive summary: sentences quoted from the sources.