MICCAI 2019 Tuesday

MICCAI 2019 DAILY 10 Women in Science Mathilde Bateson is a PhD student working in computer vision and medical imaging at Ecole Technologique Superieure in Montreal, Canada. Tell us about your work. What do you do? I work in computer vision with applications in medical imaging. I’m very motivated by integrating interesting and principed mathematical intuitions into deep learning to make models more understandable, more robust, and better in general. I think medical imaging applications are challenging and important. Are you more of a mathematician or medical imagist? Definitely somewhere in the middle, which sometimes is not the easiest. You can have the trap of being bad at both. [laughs] The goal is to be good at both so there is so much to learn and so much to remember at each moment. I am definitely interested in understanding why deep learning works, and not just doing applications, but also to really try and gain better knowledge for myself. You have done work on medical imaging. I’m here to present a paper on segmentation of the vertebrae. It could be applicable to any kind of organ. Actually, what I’m working on right now, my next application, will be on the heart, the segmentation of the whole heart. How do you choose an organ to study? I think the main trend right now is to use available, good quality datasets that you can compare. If a very good group has published a very good paper, you want to have the same baseline to have them as a competitor. You want to compete with their work so you should work on the same dataset, for instance. Then maybe add various datasets. I also try to integrate prior knowledge. Prior knowledge can be anatomical and things like that. This means you can choose wisely your organ or your application. For example,