MICCAI 2016 Daily - Wednesday

MICCAI Daily: What is challenging in doing that? Iulia: Biomedical image analysis is not an easy field. Also it’s not like mechanical engineering. There’s a lot of irregularities and shape analysis, all sorts of things like that. Perhaps one of the difficulties is that we don’t have an algorithm that automatically segments the ventricles from MRI sequences at the moment. MICCAI Daily: Why don’t you have it? Iulia: People are working on it, but I think it’s quite a challenge to come up with an accurate enough algorithm. MICCAI Daily: How do you solve these challenges from an image analysis point of view? Iulia: One of the challenges is that we don’t actually have a ground truth which means a golden standard to assess how extended the scar is after somebody had a myocardial infarction or heart attack. It is very difficult to come up with an algorithm to compare against if the clinicians can’t agree among themselves on the extent of the damage. MICCAI Daily: Is there a feature you would like to add to your work? Iulia: I would very much like to have much larger data sets to work with so if I could have perhaps, 2 million data sets, I could develop more intelligent machine learning algorithm to independently segment the scars. MICCAI Daily: Do you hope to bring this into clinical tests? Iulia: Yes, we actually have a very close collaboration at the University of Oxford with John Radcliffe Hospital which is a research hospital in Oxford. It’s a very big, regional hospital. Our work is a very close collaboration with the clinical work. In fact, we’re part of multidisciplinary teams to try and sort out these problems. Presentation 19 MICCAI Daily: Wednesday