Unsupervised Deep Domain Adaptation in Cardiac Imaging

Abstract: In this talk, we will discuss two different applications that require Unsupervised Domain Adaptation. Firstly, in congenital heart defects, MRI data is scarce in comparison to other diseases. We developed an approach that completely eliminates the necessity of labels in the target domain during training to solve a voxel-based task. Secondly, we show how adversarial domain transfer can efficiently be used to improve the realism of minimally-invasive surgical training simulators,which has been also  posed as a challenge (AdaptOR) at MICCAI 2021 conference.

‍

Presenter: Sandy Engelhardt, Ph.D., is an Assistant Professor at Heidelberg University Hospital and leads the Working Group‘Artificial Intelligence in Cardiovascular Medicine’. Her dissertation“Computer-assisted Quantitative Mitral Valve Surgery” won the BVM-Award 2017 for the best PhD thesis in the German Image Processing Community.

The goal of her newly established working group is to leverage methods from the field of Artificial Intelligence for the analysis of heterogenous data collected from patients with cardiovascular diseases. The group especially focuses on exploiting the potential of multimodal time-resolved cardiac images, such as MRI, Echocardiography, CT and Endoscopyfor objective decision support in diagnosis and treatment. Beyond that,they  continuously work in the direction of increasing the safety of surgical and interventional cardiovascular procedures. For example, building customized surgical training modules and intraoperative assistance systems are an integral part of their research topics.

‍
It is her strong belief that research can only thrive through collaboration,hence they follow a translational approach and work very closely together with clinical partners. This enables them to address highly relevant clinical questions at the interface of cardiac surgery, cardiac intervention and cardiology.

‍