The goal of this project was to develop automatic tools for discovery of new imaging biomarkers using deep learning that can improve prediction of AMD progression. We aimed to develop novel algorithms to combine structural information contained in multimodal and longitudinal retinal images and quantify a rich set of imaging biomarkers that can describe representative disease characteristics correlated with progression.  The results ere meant to be integrated in a prediction model together with non-imaging biomarkers, such as genetic and environmental markers, to predict the risk of progression to late AMD. This model would give clinicians the possibility to identify patients at high-risk of progression to provide them with active surveillance and personalized therapy to prevent blindness.