MICCAI 2020 Daily - Tuesday

3 Nazim Haouchine 7 that will estimate the brain shift and that will align and register the brain .” Another important aspect of this technique is that it only uses one image, which turns it into an ill-posed problem, but makes it appealing for future use in the operating room. Nazim solves the problem with a graph- based registration using physics-based modeling and a graph in 2D extracted from cortical vessels . Those cortical vessels are extracted in the OR using a convolutional neural network. Deep learning is very successful nowadays in generating robust segmentation. He built the first segmentation technique on cortical vessels so that it can be used and generalized later. Nazim previously worked on a similar technique related to the liver at Inria Strasbourg , where he brought together computer vision, computer graphics, physics, and space simulation. “When I came here to the Brigham they had this problem, and it is a well-known problem, but we have taken a fresh view on it and I wanted to bring to it some of what I had learned before,” he tells us. “We tried to do something different from others using simple methods and existing data without adding hardware. It makes a really nice pipeline. The collaboration between neurosurgeons and scientists makes me very proud actually. We are on the edge of trying this in the OR in the real world.” DAILY Tu e s d a y