Cardiovascular diseases are the most common cause of death in EU member countries. To reduce their burden and improve quality of life, it is critical for healthcare procedures to become more personalized. This requires the integration of information from multiple types of medical data (clinical, imaging, functional, molecular, and genomic) and from multiple centers. However, integrating multimodal data raises problems regarding the dimensions and characteristics of the various modalities, as well as situations where one or more of the modalities can be unavailable across sites. This proposal aims to overcome the challenges brought by combining multiple
modalities in missing settings, through the development of robust single modality encoding strategies as well as a novel universal multimodality integrator that uses transformers to handle the missing information. Towards these goals we propose to integrate the NextGen consortium, who is already working towards the integration of multiple modalities, bringing new capacity to the NextGen consortium and contributing to improving its tools.