Artificially intelligent eyes, built on machine and deep learning technologies, can empower our capability of analysing patients’ images. By revealing information invisible at our eyes, we can build decision aids that help our clinicians to provide more effective treatment, while reducing side effects. The power of these decision aids is to be based on patient tumour biologically unique properties, referred to as biomarkers. To fully translate this technology into the clinic we need to overcome barriers related to the reliability of image-derived biomarkers, trustiness in AI algorithms and privacy-related issues that hamper the validation of the biomarkers. This thesis developed methodologies to solve the presented issues, defining a road map for the responsible usage of quantitative imaging into the clinic as decision support system for better patient care.