Computer Vision News - January 2022

47 Daniel Rueckert challenges and what you really want to do with the images you acquire. In the 14 months you have been at TU Munich, can you pinpoint one thing you found in your research that made it worth making the move? You mean worse, or worthwhile? Yeah, worthwhile. Oh, I thought you said it is bad! No, worth, with T-H. I am generally a positive guy, so if in doubt, take the positive! It is good you clarified it! [ we both laugh ] We have developed algorithms which help us acquire better quality images faster. One of the worthwhile things has been seeing the difference this makes in clinical practice. I have learned a huge amount about what radiologists do in clinical practice, how they operate, their daily routine, and what their workflow looks like. algorithms work as they are supposed to work, and whether they are reliable and useful for them. Nassir Navab toldme some time ago ago how important it was for him that his students be in touch with physicians constantly to know what they need because they should be the main input for their work. Exactly. Nassir and I work very closely together. It has been great coming to TU Munich and having somebody like Nassir as a colleague. He is very right, of course. You want to get feedback from clinicians, but you also need clinicians to define what the right problem is you should work on. Otherwise, you go ahead and develop something, then they tell you nobody ever wanted to solve that problem in the first place! Fortunately, that does not happen very often. I am also very happy I now have postdocs in my group who are medical doctors and radiologists. Your readers will know medical imaging often has different challenges than standard computer vision tasks. It is very important to understand those