Related Experiment Video
Updated: Aug 6, 2026
12:08
From Voxels to Knowledge: A Practical Guide to the Segmentation of Complex Electron Microscopy 3D-Data
Published on: August 13, 2014
BOUNDARY-AWARE INSTANCE SEGMENTATION IN MICROSCOPY IMAGING
Thomas Mendelson1, Joshua Francois2, Galit Lahav2
1The School of Electrical and Computer Engineering, Ben-Gurion University of the Negev.
Summary
This study introduces a new method for cell segmentation in microscopy, improving the separation of touching cells without needing user prompts. The approach enhances accuracy in dense cell populations for better cellular dynamics research.
Area of Science:
- Bioimage analysis
- Computational biology
- Cellular imaging
Background:
- Accurate cell segmentation is crucial for cellular dynamics studies.
- Existing foundation models struggle with dense microscopy scenes, requiring extensive prompting.
- Separating touching or overlapping cell instances remains a significant challenge.
Purpose of the Study:
- To develop a prompt-free, boundary-aware instance segmentation framework for microscopy images.
- To improve the separation of adjacent cell instances in dense environments.
- To enhance the geometric consistency and accuracy of cell contour modeling.
Main Methods:
- Predicting signed distance functions (SDFs) instead of binary masks for cell contours.
- Utilizing a learned sigmoid mapping to convert SDFs into probability maps for sharp boundary localization.
- Employing a unified Modified Hausdorff Distance (MHD) loss integrating region and boundary terms for training.
Main Results:
- The proposed framework achieves robust separation of adjacent cell instances.
- Demonstrated improved boundary accuracy and instance-level performance on high-throughput microscopy datasets.
- Outperformed recent SAM-based and foundation-model approaches in challenging dense cell segmentation tasks.
Conclusions:
- The prompt-free, SDF-based approach offers a significant advancement in cell instance segmentation.
- Enables more accurate and automated analysis of cellular dynamics in complex microscopy data.
- Provides a geometry-consistent method for modeling cell boundaries, overcoming limitations of current models.
