Medical Image Segmentation


Medical Image Segmentation

Open Master Thesis

Contact person: Fabian Bongratz, Anne-Marie Rickmann, Christian Wachinger


Medical Image Segmentation is the process of partitioning medical scans into different structures. Medical scans, like MR or CT scans give information about the morphology of the scanned body part. For monitoring disease progression like e.g. Alzheimers disease, measurements like volume of certain structures are needed, for which a segmentation is needed. Since manual labeling of whole volume scans takes a long time, there is a clinical need for automatic segmentation methods. Current state of the art segmentation methods are usually fully convolutional neural networks based on the popular U-net architecture. This project will focus on applying deep neural networks for the task of segmentation of MRI or CT images.

Since the field of medical image segmentation is quite broad, the project can be adapted to the students interests and ideas.


Some areas we are interested in:

    • Brain parcellation
    • Memory efficient 3D convolutional neural networks
    • Uncertainty estimation for segmentation



Required skills:

  • good python programming skills
  • machine learning experience
  • Experience in medical image analysis and deep learning (PyTorch, Tensorflow etc.) is desirable.