Cardiomyocyte fiber reconstruction with deep learning Cardiomyocyte fiber reconstruction with deep learning

The orientation of cardiomyocyte fibres is key for the electromechanical behaviour of the heart. Unfortunately, it cannot be measured in vivo with the current imaging techniques. For this reason, researchers have found several ways to include cardiomyocyte fiber information into their models of the heart so that they behave in a realistic way. One way is to perform long acquisitions of ex-vivo data and then build statistical atlases of the obtained data that are then warped to each new heart. A more successful alternative have been the development of rule-based models (RBM) that assigns a vector to each node of the heart mesh with the cardiomyocyte fiber orientation, which is obtained with mathematical models mimicking ex-vivo data. The main goal of this project is to explore deep learning (i.e. geometrical deep learning, graph-based) strategies to make this process faster, based on the learning of the relation between the cardiac anatomical structures and the cardiomyocyte orientation provided by exvivo data and RBM models

Supervisor: Oscar Camara