Computer Vision News 22 Best Paper MIUA 2023 This paper proposes a novel approach to generalized image registration that leverages learned shape encodings. Traditional image registration methods often demand tailored solutions for specific tasks. However, this innovative approach uses medical image segmentations to perform registration across different domains and datasets, promising a solution capable of addressing diverse challenges. The researchers needed to find a common representation between different datasets. They grappled with the task of training a model on one instance of data and then inferring it on something completely different. “The idea to leverage other advances in medical image analysis, like segmentations, lets us use a bridge between these different problems,” Christoph explains. “We can use advantages from one field to solve or find some kind of solution in another field. That’s pretty nice, in my opinion.” Christoph Großbröhmer is a PhD student at the University of Lübeck under the supervision of Associate Professor Mattias Heinrich. Fresh from winning the Best Paper prize at the 27th UK Conference on Medical Image Understanding and Analysis (MIUA), Christoph joins us with Mattias to discuss his award-winning work. Generalized 3D Medical Image Registration with Learned Shape Encodings Christoph Großbröhmer Mattias Heinrich
RkJQdWJsaXNoZXIy NTc3NzU=