Understanding the correlation and causal relationship between cardiovascular diseases and the heart shape is of utmost importance to improve diagnosis methodologies.
The MRI images are a common heart diagnostic tool that allow visualizing slices of the heart at different spatial levels, temporal steps and heart orientations. Several methods exist to reconstruct a 3D model of the heart for each time step yielding a 4D reconstruction, much less explored is the automatic discovery of correlations and causality relationships between geometric features of the reconstructed shape and cardiovascular diseases. This project aims to form a bridge between cutting-edge techniques in deep learning, geometric processing and heart imaging to improve the current diagnostic tools. An ongoing collaboration with the radiology team from the department of experimental medicine in the Umberto I hospital will provide domain expertise and MRI scans of at least 100 patients with and without ventricular segmentations.
Cardiovascular diseases are the leading cause of death in most developed countries, it is of utmost importance to improve the diagnosis tool to improve the prevention and early treatments.
The field of medical imaging is progressing at a fast pace, however, the most advanced techniques in geometry processing and deep learning remain underexplored in heart and ventricular analysis, despite the possibility to greatly improve the current state of the art in diagnostic tools, that continue to rely mainly on volumetric features of the ventricles.
This project aims to build a bridge between cutting-edge deep learning, geometry processing and heart 4D MRI imaging, to apply state of the art techniques when analysing the heart and the ventricles.
We strongly believe that applying state of the art techniques in the learning and geometry processing community to heart and ventricles, could greatly improve the heart diagnostic tools currently available.