MIDL Vision 2021

images and still see the tumor. However, we don’t want to see individual features like the tracer signal and the tumor because in our target image where we do the registration this structure doesn’t exist, and the network can’t guess correctly. That’s why we use the population average 2p image as the target domain rather than the individual 2p images with the tracer. ” The team have been working on this for two years and are finding that there are no longer any technical issues. It may very occasionally fail when an image is damaged, but generally, it just works! The RIKEN Center for Brain Science is located in Saitama Prefecture, which borders Tokyo to the south, and is the biggest RIKEN campus. RIKEN is a research institute renowned for high-quality research in a variety of scientific fields. The Center for Brain Science covers a wide range of research related to brain science, from computational neural science to behavior studies . “ My favorite part of this work is how interdisciplinary it is, ” Henrik shares with us. “ Also, that it is used in the real world. People really depend on our output. Our colleagues who are mostly neuroscientists acquire the images, we do the image processing, then they interpret the results and want to include them in their work, so it must really work! ” To learn more about Henrik’s paper, you can view his short presentation video [E5] anytime and visit his discussion session during Image Registration / Synthesis at 13:45- 14:30 (UCT+2) 17 VISION MIDL A population average Nissl and population average backlit image, respectively (based on all our registered images) aligned to our population average 2p microscopy image. Henrik Skibbe

RkJQdWJsaXNoZXIy NTc3NzU=