BigCell generates gigapixel digital cancer tissue images
BigCell: Generating Gigapixel Whole-Slide Images
Computer Vision and Pattern RecognitionMachine Learning
Summary
Cancer diagnosis relies on extremely large digital images of tissue samples, but existing computer models can only create small pieces of these images. The researchers developed BigCell, a system that creates full-size tissue images at the scale doctors use, using a new method that pieces together smaller parts smoothly. They trained their model on over 120,000 real images and showed it can even produce images guided by clinical text reports. This helps improve computer-based cancer diagnosis by providing more realistic synthetic images for training.
What this means in practice
- •For pathology ai developers: Generate realistic full-slide tissue images to improve training and evaluation of diagnostic AI models working at the whole-slide scale.
- •For medical imaging software teams: Create synthetic gigapixel pathology images conditioned on clinical text reports to enhance diagnostic tool capabilities.
Authors
Srikar Yellapragada, Alexandros Graikos, Zilinghan Li, Kostas Triaridis, Tarak Nath Nandi, Ji Dong K Bai, Beatrice Knudsen, Tahsin Kurc, Prateek Prasanna, Rajarsi R. Gupta, Ravi K Madduri, Joel Saltz, Dimitris Samaras
Abstract
Whole-slide images (WSIs) are gigapixel-scale digital scans of histopathology tissue and serve as the primary basis for cancer diagnosis. Existing histopathology generative models operate at the patch scale, producing fixed-size tiles of at most 1024$\times$1024 pixels which encompass only a small fraction of the WSI. In contrast, text reports for clinical diagnoses and slide-level diagnostic labels -- the supervision that exists at scale -- operate at the whole-slide level. We present BigCell, the first approach to generate coherent whole-slide histopathology images at native gigapixel resolution. BigCell trains a flow-matching transformer at 2048$\times$2048 pixels on over 120,000 publicly available WSIs spanning multiple organ types. Conditioning on embeddings from pathology foundation models and using an inference-time sliding-window pipeline, BigCell generates coherent images of up to 32,768$\times$32,768 pixels, outperforming prior tile generators on both tile-level and slide-level fidelity metrics. 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. 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. We demonstrate that augmenting with synthetic samples improves slide-level few-shot classification performance by up to $18 \%$ over baselines trained on real data.