MSc Thesis: Cardiac MRI Segmentation using Morphometric Informed Multimodal Self-Supervised Models


Medical datasets and especially biobanks, often contain extensive tabular data with rich clinical information in addition to images. In practice, clinicians typically have less data, both in terms of diversity and scale, but still, wish to deploy deep learning solutions. Combined with increasing medical dataset sizes and expensive annotation costs, the necessity for unsupervised methods that can be pretrain multimodally and predict unimodally has risen.

We have developed a self-supervised framework that incorporates both tabular and imaging data during pretraining but requires only images during testing. We found that the training dynamics are strongly influenced by morphometric features, i.e. features that describe shape and size, and that these features greatly improve the performance on relevant downstream tasks. We expect segmentation to be a task that benefits from such morphometric features.

Your goal will be to use the aforementioned multimodal framework to improve the performance of cardiac MR segmentation using UK Biobank cardiac MR data. You will explore the influence of the morphometric features on downstream performance and how different encoder architectures change the pretraining and finetuning dynamics. You will investigate the influence of different imaging planes (Short Axis, Long Axis 2 channel, Long Axis 4 channel) on the pretraining and downstream segmentation performance.

To accomplish this work successfully, we expect you to have:

  • Strong coding skills and familiarity with Pytorch and Numpy
  • Basic knowledge of segmentation
  • Basic knowledge of contrastive self-supervised learning, especially SimCLR
  • A strong spirit of independent work and desire to solve interesting research questions

What we offer

  • An exciting research project with many possibilities to bring in your own ideas.
  • Close supervision and access to state-of-the-art computer hardware.
  • The chance to work in a team of experts in image processing, deep learning, biomedical engineering and medicine.
  • Support in bringing your finished project to publication

How to apply:

Send an email to with your CV and transcript. We promise to get back to you within a couple of days.

Paul Hager
Paul Hager
PhD Student

My main research interest is how to effectively integrate biological, genetic and lifestyle information into medical imaging deep learning models.