Related Experiment Video
Updated: Jul 27, 2025
10:25
Deep Learning-Based Segmentation of Cryo-Electron Tomograms
Published on: November 11, 2022
9.0K
3D surface reconstruction of cellular cryo-soft X-ray microscopy tomograms using semisupervised deep learning.
Michael C A Dyhr1, Mohsen Sadeghi2, Ralitsa Moynova1
1Institute of Chemistry and Biochemistry, Department of Biology, Chemistry and Pharmacy, Free University of Berlin, 14195 Berlin, Germany.
Summary
Automated 3D segmentation using deep learning accelerates cell ultrastructure analysis from cryo-soft X-ray tomography (cryo-SXT) data. This method enables high-throughput quantification of nanoscopic cellular features, overcoming manual segmentation bottlenecks.
Area of Science:
- Cellular and Molecular Imaging
- Biophysics
- Computational Biology
Background:
- Cryo-soft X-ray tomography (cryo-SXT) provides high-resolution cellular ultrastructure without labeling.
- Fast data acquisition in cryo-SXT generates large datasets.
- Manual segmentation of cryo-SXT data is time-consuming.
Purpose of the Study:
- To develop an automated 3D segmentation pipeline for cryo-SXT data.
- To enable high-throughput analysis of cellular ultrastructure.
- To quantify nanoscopic morphological parameters of cellular structures.
Main Methods:
- Developed an end-to-end automated 3D segmentation pipeline.
- Utilized semisupervised deep learning for segmentation.
- Validated the pipeline on mammalian cell filopodia.
Main Results:
- The pipeline automates the segmentation of cryo-SXT tomograms.
- Achieved robust segmentation with limited manual annotations and varying conditions.
- Enabled extraction of 3D cellular ultrastructure and quantification of filopodia morphology.
Conclusions:
- Semisupervised deep learning offers an efficient solution for cryo-SXT data segmentation.
- The automated pipeline facilitates high-throughput analysis of cellular ultrastructure.
- This approach advances the study of nanoscopic cellular morphology.
