Automatic segmentation of single cell anatomy
Aims & Objectives
Analysis of three-dimensional biological cell samples is critical for understanding the mechanisms of disease and for the development of specific treatment of disorders. Soft x-ray tomography (SXT) is a unique technology that can image whole intact cells under normal and pathological conditions without labeling or fixation, at high throughput and spatial resolution. Ongoing improvements in throughput and applications of SXT increase the demand for accelerated image analysis. At present, image segmentation is a primarily manual task, which is tedious and time-consuming.
In our project, we aim to tackle this major bottleneck in the SXT data analysis pipeline by implementing the semi-automatic segmentation application Biomedisa and developing a fully automatic segmentation pipeline based on a deep convolution neuronal network.
By creating an automatic segmentation platform for SXT data, we expect to provide high segmentation accuracies and significantly accelerate the segmentation of large and complex data of single cell data. Through quantitative automatic analysis of the morphological changes within individual cells in pathologic conditions, like viral infection, we expect to reveal new principles of virus-induced alteration and provide important information for the development of antiviral therapies. While the application of the segmentation pipeline will be focused on viral infection pathways, the vision for the project is to establish an automatic segmentation platform for single cell anatomy applicable to other pathologies, for example, cancer, with use cases reaching far beyond what can be currently foreseen.
Research topics
The EMCL provides the knowledge of high-performance computing and infrastructure set up to challenge the tasks of automatic segmentation of single cell anatomy in 3D. A key issue lies in the choice of an appropriate network, an efficient minimization function, and transfer learning to be applicable to a large variety of cells. The research is focused on the following topics:
- Deep Learning of computed tomography datasets
- Image segmentation of cell anatomy
- Computer Vision
- Creation of an online platform for anatomical analysis of single cells based on cloud technologies
Funding
- The project is funded by The framework of the Excellence Strategy of the Federal (BMBF) and State Governments of Germany, “Flagship Initiative Engineering Molecular Systems” and
- European Union Research and Innovation Act, project “CoCID”
People from the EMCL
Partners
- Centre for Organismal Studies (COS), Heidelberg University
- SiriusXT Limited, Dublin, Ireland
- Department of Infectious Diseases, Molecular Virology, University Clinics Heidelberg
Contact
Links
In preparation
Publications
- Ayse Aydogdu-Erozan, Philipp D. Lösel, Vincent Heuveline and Venera Weinhardt. 2023. ACSeg: Automated 3D Cytoplasm Segmentation in Soft X-Ray Tomography, DOI: https://doi.org/10.11588/emclpp.2023.2.94947
