Background: There is growing evidence that fat surrounding the heart is associated with an increased risk of cardiovascular disease . Magnetic Resonance Imaging (MRI) is a non-invasive modality free of ionizing radiation which is increasingly used to assess cardiovascular disease. A recent MRI technique using water-fat separation reconstruction allows for estimation of the pericardial and epicardial fat volumes . However, current fat segmentation methods using MRI require manual interaction, are operator dependent, and the delineation of the epicardium is sometimes challenging.
Aim: This project will explore the use of convolutional neural networks for automated and robust segmentation and quantification of pericardial and epicardial fat in MRI data. The implemented algorithm will be developed based on conventional fat MRI techniques and evaluated in collaboration with clinical partners. A novel MRI approach which allows time-resolved, volumetric water-fat MRI of the heart will also be explored if time permits. It is hypothesized that the new technique which provides richer MRI data will facilitate segmentation of epicardial and endocardial fat.
Prerequisites: The development of algorithms will be performed in Python or Matlab. Candidates with a biomedical or computer science background are strongly encouraged to apply.
. Dey D, Nakazato R, Li D, Berman DS (2012) Epicardial and thoracic fat—noninvasive measurement and clinical implications. Cardiovasc Diagn Ther 2(2):85–93
. Homsi R, Meier‐Schroers M, Gieseke J, Dabir D, Luetkens JA, Kuetting DL, Naehle CP, Marx C, Schild HH, Thomas DK, Sprinkart AM. 3D‐Dixon MRI based volumetry of peri‐ and epicardial fat. Int J Cardiovasc Imaging. 2016;32:291–299.