Eurostars project E!11626 DeepSegment, ImaComp Consulting AB
| Reference number | |
| Coordinator | ImaComp Consulting AB |
| Funding from Vinnova | SEK 4 583 793 |
| Project duration | November 2017 - August 2020 |
| Status | Completed |
| Venture | Eurostars |
Important results from the project
The goal was to offer users a system for a complete and efficient segmentation of 2D and 3D structures in areas such as medicine, biology and materials science. The system should guide the user through all steps. The process must be user-friendly and be as automatic as possible, but also allow manual steps for eg review. Both traditional and deep-learning-based segmentation methods shall be supported and concrete solutions integrated. The system shall keep track of all parameters, images and results. It should also be possible to analyze the segmented structures.
Expected long term effects
A complete and user-friendly system for efficient segmentation of 2D and 3D structures in several research areas has now been finished completely according to the objective we set up. Beta versions are now out with a number of groups in both medicine and material science. The first major system will be set up at a German institute that deals with correlative microscopy where very complicated workflows with a focus on segmentation will be implemented with the system. Several other commercial system deals are planned thereafter.
Approach and implementation
The development work has gone very well. DeepSegment is a further development of the XamFlow system that was developed during a previous Eurostars project. This formed a very good basis for early prototypes to allow external research groups to evaluate and provide feedback on. XamFlow has also made it possible for each team to work independently and thus also efficiently. Until the Covid crisis, we had regular meetings in Zurich and Stockholm, which then turned into video calls. At the meetings, the work has been able to be synced and gradually improve the platform based on feedback.