Computer Vision News - November 2020

“The accuracy of an automatic algorithm might not always match the target accuracy that we are actually interested in,” she explains. “To circumvent that when the fully automatic algorithm fails, we have a fallback with a manual interaction that is straightforward, easy to use, and does not take a lot of effort from the physician. We can potentially also further automize this in the future. ” Having access to all the data that you need for this could be a challenge, but a fruitful collaboration with Sahlgrenska University Hospital in Sweden resolved this concern. Bringing computer scientists and physicians together to work on a project like this is no mean feat. Siemens Healthineers was also a collaborator, including co-authors Marcus Pfister and Markus Kowarschik, who, along with her advisor, Andreas Maier , Katharina declares it was great to work with. These relationships gave her the opportunity to experiment with the algorithms on real clinical prototypes – following all the appropriate guidelines, of course. What computer visions techniques were involved? “In general, X-ray systems or C-arm systems can be seen as just another version of the typical pinhole camera model,” Katharina tells us. “The whole set-up that we have in terms of epipolar geometry , making sure that we use the camera geometry, the projection 3 Katharina Breininger 15 Best of MICCAI 2020