An increasing number of labs around the world develop microscopes with unprecedented imaging capabilities. Today it is e.g. possible to follow the developmental process of various organisms from a fertilized egg to a hatching larva, and beyond, on the scale of individual cells. Such imaging pipelines produce, every day, terabytes of data containing information which is highly relevant to biologists.
BioImage Computing addresses the challenge of bridging the gap between data acquisition and scientific insight given these vast amounts of image data.
The purpose of this workshop is to draw the attention of the computer vision community to the still little known and explored “playground” offered by the bioimage community. During this workshop we will (i) give an overview on state of the art methods for bioimage computing (ii) give an overview on the freely available bioimage data that is already out there, ready to be downloaded and processed, including benchmark data and challenges, and (iii) foster the exchange between the computer vision community and biology experts (desperately in need for computer vision and machine learning methods to extract knowledge from already existing images) and the bioimaging community (that is to date largely disjoint from computer vision community).
What are the challenges that bioimage computing usually faces?
- To answer relevant biological questions a very high-level of accuracy is required (e.g. cell tracking in 3d movies of fruit fly embryo development is required to be at least 99.99% accurate per frame to yield 97% accurate overall tracks (cell lineages)).
- Whenever desired precision cannot be achieved by a fully automatic pipeline, proofreading and data curation must be conveniently possible by a user that is a biologist, but not a Computer vision specialist.
- Many bio-labs produce very large data sets (multiple TB per acquisition are not unusual) and developed solutions must be able to deal with them.
- Adaptive imaging strategies need often to run in ‘real-time’, meaning that there is just as much time for computation as the microscope leaves between image acquisitions.