BioImage Segmentation Comparison Utility and Interactive Tool
BioImage Segmentation Comparison Utility and Interactive Tool
About
What is BISCUIT?
BISCUIT is an open-source platform for visually comparing state-of-the-art cell segmentation models on microscopy images.
Designed for bioimage analysts, life scientists, and data engineers, BISCUIT helps you:
- 🔬 Run multiple SOTA segmentation models (e.g. Cellpose, Omnipose, StarDist) on the same image dataset,
- 👁️ Visually assess segmentation quality through side-by-side comparisons,
- ⚙️ Deploy flexibly:
- use Google Colab for quick browser-based access,
- or install locally to leverage HPC resources (tested at ETH Zurich).
BISCUIT is built for scenarios where visual quality and biological plausibility matter more than abstract metrics — helping you choose the right segmentation model by seeing what really works on your data.
Origins
BISCUIT (BioImage Segmentation Comparison Utility and Interactive Tool) traces its roots back to the Scientific IT Services (SIS) at ETH Zurich. Spearheaded by Thomas Wüst and Franziska Oschmann, the goal was to create an intuitive yet powerful platform for visually comparing cell segmentation models on microscopy data.
The initial version of BISCUIT was built on top of the MIDAP framework—a modular system originally developed for analyzing data from Mother Machine experiments. MIDAP’s flexible architecture made it a solid foundation for building interactive segmentation workflows.
Development continued in close collaboration with the ETH Zurich imaging center ScopeM, with practical contributions and technical leadership by Szymon Stoma and Andrzej Rzepiela. Together, the team laid the groundwork for a tool that aspires to become a versatile visual benchmarking suite for segmentation models in the life sciences.
Useful Links & Credits
- Scientific IT Services (SIS): ethz.ch/de/die-eth-zuerich/organisation/abteilungen/informatikdienste.html
- MIDAP framework (Maddle Machine image analysis): github.com/Microbial-Systems-Ecology/midap
- ScopeM at ETH Zurich: scopem.ethz.ch
- Image & Data Analysis of ScopeM : let-your-data-speak.com
Media
Benefits
1Zero Setup
Use state-of-the-art segmentation models directly from your browser — no installation needed. Run on your own microscopy data in minutes.
2Scalable by Design
Once you find a suitable model, you can seamlessly switch to high-performance computing (e.g. ETH Zurich clusters) for large-scale analysis.
3Visual First
Compare models through visual outputs — focus on what actually matters in bioimage segmentation: the biological plausibility of the result.
How to start?
Zero installation
You don’t need to install anything. Just open the notebook in Google Collab — everything happens step by step inside the Jupyter cells.
✅ Follow the instructions provided inside the notebook cells
✅ Example images are included so you can start right away
✅ You can also upload your own images to test how different models work on your data
That’s it. Simple. Visual. Interactive.
Service
Contact us to discuss customization.
BISCUIT can be extended and adapted to better serve the needs of individual research groups and institutions. Together with the Scientific IT Services (SIS) and the ScopeM image analysis platform, we offer two directions of support as a service.
If you’re interested in service or would like to discuss your specific needs, please don’t hesitate to get in touch. The teams at ScopeM and SIS are happy to explore how BISCUIT can support your research.
Custom model training and deployment
We can assist with retraining segmentation models on your own data, adjusting architectures or parameters as needed. These customized models can then be directly deployed and used within the BISCUIT interface.
Model deployment for groups & institutions
Models can also be installed on-demand for specific scientific groups or facilities — enabling private or large-scale use of the BISCUIT platform on institutional infrastructure.
Contact
Let's talk.
Email
© 2025