Genetic advances over the past 10 years have led to the development of several targeted therapies for lung, breast and colon cancer. However, there are a number of factors that limit the optimal use of these innovations, including the high cost of the organizational process associated with molecular testing, and their late use in the patient's journey. Recently, the prospect of obtaining non-invasive, cost-effective and timely triggers for diagnostic & therapy has emerged from a discipline known as Radiomics.

This project aims at evaluating whether recent results in deep learning models, trained to exploit weak labels can serve to extract meaningful lesion localizations from image-level labels, either from individual scans or given a (longitudinal) sequence thereof. To this end, we will scale up existing models that have been shown to work on 2D images to a 3D context, studying labeling performance as the dataset size grows.