Novel geometric deep learning methods for medical image analysis

Novel Geometric Deep Learning Methods for 3D Blood Vessel Analysis


Funded by:



chair MIA:

Daily supervisor:


Dr. Kak Khee Yeung (Amsterdam UMC)


Cardiovascular diseases are the global leading cause of mortality worldwide. Among these diseases are abdominal aortic aneurysms (AAAs), which are characterized by dilations of the vessel wall of the abdominal aorta. The most severe complication of this disorder is rupture of the aorta, which is in many cases fatal. Rupture can be averted using preventive surgeries, however, such interventions are not risk-free. Therefore, both rupture risk and risks associated with post-operative complications should be taken into consideration in AAA patient management.

Medical imaging using MRI, CT and ultrasound plays a key role in the management of patients with AAA. These images are used to extract metrics like the diameter of the aorta, which according to current clinical guidelines is a main indicator of rupture risk. However, such simple metrics are unlikely to tell the full story. Medical images likely contain more information for patient-specific risk prediction. In this project, we aim to bring this information to light by developing novel geometric deep learning techniques for 3D blood vessel analysis