Coronary vessel analysis in CTA
The goal of this project is to develop automatic and semi-automatic tools that help in the diagnosis and follow-up of
coronary artery disease. We investigate learning-based approaches that allow the identification of vascular lesions in CT
images. Since machine learning techniques require of highly accurate label sets, our main goal is to reduce the dependence on
such labels by using semi-supervised or unsupervised methods.
In coronary artery disease diagnosis, cardiologists have to explore at least two different cardiac timepoints of the
CT acquisition to validate possible lesions. Such an operation is time consuming. Alternatively, the different
volumes can be registered to avoid visual matching. However, the latter can be even more time demanding.
We seek for the development of fast registration methods that tackle only the registration of potential lesions instead
of the entire volumes. Issues such as the aperture problem make this task challenging.