Center for Computational and Theoretical Biology

EM Segmentation


3D electron microscopy provides detailed ultrastuctural information, but manual segmentation of such images is still the de-facto standard. We develop methods to automatically extract quantitative stuctural information from such dense, large-scale image volumes from electron microscopy.


Targeted segmentation

Using classical watershed-based segmentation and object features, we develop tools to automatically extract well-defined structures from electron microscopy images, e.g. presynaptic vesicles from electron tomograms.

Trainable methods and deep learning

We apply trainable pixel classifiers and convolutional networks to explore large 3D EM volumes and to determine quantitative structural detail.


2017[ to top ]
  • FIJI Macro 3D ART VeSElecT: 3D Automated Reconstruction Tool for Vesicle Structures of Electron Tomograms. Kaltdorf, Kristin Verena; Schulze, Katja; Helmprobst, Frederik; Kollmannsberger, Philip; Dandekar, Thomas; Stigloher, Christian. In PLOS Computational Biology, 13(1), bll 1–21. Public Library of Science, 2017.