Computer Vision News - November 2020

2 Young Scientist Winner 6 Best of MICCAI 2020 Nils Eckstein is a PhD student at Funke Lab and is affiliated with ETH Zurich, the University of Zurich, and HHMI Janelia. His supervisor is Jan Funke. Last month, having presented his paper at MICCAI 2020, he had the honour of receiving a Young Scientist award in recognition of his talents. He speaks to us about his work and what the future holds. In recent years, electron microscopy (EM) has advanced so much that we are now able to acquire EM volumes of entire organisms or parts of organisms – the brain of the fruit fly Drosophila , for example. This is a huge dataset on which people are working on neuron segmentation and identification of structures of interest, such as microtubules, which are part of the cytoskeleton of a cell and important for a variety of cellular processes. This work proposes a method for tracking microtubules automatically in EM datasets . You could input a large part of the fruit fly brain and output all reconstructed microtubules . Until now, it has not been possible to automatically reconstruct these structures on such a large scale. Why has this not done before? Well, the first large-scale EM dataset only came out a couple of years ago and the resolution was not good enough. Also, microtubules are a difficult problem, so it took a long time to get this right. “Microtubules are very small and very long, and in reconstructing them it is very easy to make topological mistakes because they are close to the resolution limit of electron microscopy,” Nils explains. “They are around 24 nanometers in diameter. The resolution of electron microscopy pipelines, depending on which technique you use, is between 4 and 16 nanometers.” He tells us the motivation for this project was more from a connectomics perspective . The algorithms people use Microtubule Tracking in Electron Microscopy Volumes