Medical Image Segmentation

Modern ML/DL algorithms achieve excellent results, even outperforming human experts in many use cases. For this, they need vast amounts of data, because they learn everything from scratch. This data is hard to acquire in medical contexts (due to e.g. privacy concerns or rare diseases).The goal of this project is to gain broad fundamental knowledge from multiple, diverse tasks (e.g. brain tumor and liver segmentation), which is then transferred to new tasks using just a few training examples.