Computer Vision News - January 2020

2 Summary Medic l Imaging 8 Bernhard tells us that real-time modalities are a key focus of his work at the moment. In particular, ultrasound, because it’s an up-and-coming modality which can be enhanced with deep learning and has the most potential to liberate imaging, democratize expertise and be brought forward to the front line of care. His work from the machine learning perspective includes image classification and image enhancement linking to clinical pathways and decision- making. Application-wise, he is looking at foetal imaging, with some work on cardiac and other moving organs. In the last 12 months, with the medical Bernhard Kainz is a Senior Lecturer in the Department of Computing at Imperial College London. He was also an area chair at MICCAI 2019 in Shenzhen, playing a crucial role in ensuring the best papers were presentedat theconference following a high-quality review process. He speaks to us about progress and innovation in his field of work. with Bernhard Kainz Real Time Modalities field becoming more interested in this type of work, he has seen his peers begin to rethink the way they have been training and testing models for decades. Tackling domain shifts was a hot topic at MICCAI 2019 and it is having a big impact on the community. Over the next year, he hopes to see two new innovations on the horizon. One is a very general domain adaptation method for feature disentanglement ; proving that generic structural or disease features can be extracted from a set of images without considering the domain. He says there is a lot of work going on, but no convincing solution as