43 Computer Vision News DCA in MI academia and industry can come together to share a range of viewpoints and perspectives on using data to solve medical imaging problems and, ultimately, improve patient care and health. “When I was a grad school student, I always tried to focus on innovation,” Shuoqi recalls. “The emphasis was on developing novel models and algorithms to optimize certain metrics. In industry, we’re looking to start from the data and develop robust solutions. We have loads and loads of data and want to ensure our method is the simplest and works in as many cases as possible. The contrast between those two fields is what we want to bring to the workshop.” The workshop’s agenda is being finalized, but attendees can expect solution-focused content on data selection, data synthesis, learning with limited and imperfect data, and data verification, as well as some of the persistent challenges that come with working with data for machine learning, like data scarcity, annotation costs, domain shift, and bias. In medical imaging, these issues are compounded by privacy concerns and quality standards. “In the medical domain, we’ve been thinking about these challenges a lot,” Dominik reveals. “I think this should be a more prominent topic in the main computer vision area. We hope we can interest people in it and get insights in both directions.” There will be a range of paper presentations, panels, and keynote speakers. The organizers hope the Synthetic data can potentially alleviate issues related to data scarcity and privacy, but it also poses new challenges. From public synthetic datasets for surgery: Top Bottom
RkJQdWJsaXNoZXIy NTc3NzU=