MICCAI 2016 Daily - Wednesday

MICCAI Daily: How will you solve this problem? Odyssée: We need the position to delineate all of the images manually, but it’s really long and complex. Also there is existing software which designs simulated data, phantoms of blood vessels. For example, VascuSynth can provide this data, but there is also synthetic data which is not as complex and difficult to segment as the real data. There is also a lot that MICCAI challenge that can provide. For example, I think of 2012 Grand Challenge on blood vessels from cardiac images, which can provide data and some ground truth along with the images. MICCAI Daily: What are the steps of this brain project? Odyssée: The brain project work in four steps. The first step is the segmentation of the blood vessel. The second step is to model the vascular tree. We need to mesh the vascular tree and establish a true model. Then other people will simulate the blood flow inside this model. With the simulation of the blood flow and with the model resulting from the segmentation, we can simulate a real MRA acquisition. MICCAI Daily: If you could add one feature to your model, which would it be? Odyssée: A real hard thing that I would like to have for segmentation would be a connectivity feature. A hard thing to do when you segment blood vessels is to reconnect the vessel, especially when there is a lot of noise in the image. In general you will end up with blood vessels that are disconnected. You don’t have fully connected vascular tree. One of the things that I would like to work on in the future is to have a connectivity feature, to be able to connect all of the blood vessels. MICCAI Daily: What is needed to succeed in that? Odyssée: We are already working on directional connectivity. We want to have, at each point, the direction of the blood vessel and when we have that we can regularize or try to reconnect the vessels. Presentation 21 MICCAI Daily: Wednesday