In medicine, the high volume of available data and the expanded number oftreatment options have rendered it increasingly complex to determine theappropriate therapy for a specific patient. Precision medicine is a promisingand emerging approach to tailor disease prevention and treatment by takinginto account individual patient characteristics. Medical imaging constitutesa rich and valuable source of patient information, but due to its unstructurednature, it requires prior processing to extract useful parameters. Currently,this analysis of imaging data often requires manual image interpretationwhich proves too cumbersome for clinical routine. Computer-aided diagnosis(CAD) systems can support physicians by performing fast, objective andreproducible image analyses and by extracting parameters that allow formore personalised disease assessment and response prediction, allowing tosupport therapeutic decisions. In this work, we have investigated toolsfor CAD and clinical decision support (CDS) systems for two pathologies,malignant melanoma and coronavirus disease 2019 (COVID-19).Malignant melanoma is the most lethal form of skin cancer. However, survivalchances have increased since recent advances in immune checkpoint inhibitorsand targeted therapies. Treatment planning and monitoring are generallyperformed using combined positron emission tomography/computed tomo-graphy (PET/CT) with fluorine-18 fluorodeoxyglucose ([18F]FDG) and regu-lar testing of blood values. Though guidelines like immune Positron EmissionTomography Response Criteria In Solid Tumors (iPERCIST) recommendquantitative analysis and comparison of target lesions, their applicationis currently not feasible in routine clinical practice. We developed a fullyautomated system for lesion detection and segmentation on whole-body[18F]FDG PET/CT which provides a fast, objective and reproducible wayto extract information on the tumour load from the imaging data.Immunotherapy has shown to improve the prognosis of patients suffering fromadvanced melanoma with respect to the previous generation of treatments.Still, a considerable part of this population demonstrates progressive disease.If patients with a poor prognosis can be identified before the start of therapy,a more aggressive treatment pathway could be considered to improve thesurvival chances. Previous clinical research had identified several image-based and non-imaging predictors, which were used as a starting point.The contributions of this work are threefold. First, we demonstrated thefeasibility of using automatically derived imaging features in survival analysisthrough a comparative study with the manual method. Next, we analysed thepredictive value of organ-specific tumour loads that can be used to includeadditional features in future clinical research. As a third contribution,multivariate Cox regression models were developed to predict one- and two-year overall and progression-free survival for metastatic melanoma patientswhen treated with pembrolizumab, an immune checkpoint inhibitor.COVID-19 has presented great challenges for the medical sector worldwide.During the pandemic, intensive care units were overwhelmed and properresource allocation became problematic. COVID-19 is an infectious diseasecaused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Patients can be asymptomatic or experience mild symptoms, while othersdevelop very severe illness. During the periods of high prevalence, there wasan urgent need for computer-aided systems to assist in diagnosis, treatmentrecommendation and resource allocation. In a large research collaboration,artificial intelligence (AI) tools were developed to alleviate the situation.We contributed to the development of automated methods for the segmenta-tion of lung lesions suggestive of COVID-19 on CT by investigating deeplearning segmentation approaches. The resulting methods allow to obtainautomated lung severity scores and can aid radiologists in quantifying theextent of the lesions. In addition, we developed a prognosis prediction modelto estimate the short-term severity, defined as intubation or death withinone month of CT image acquisition. Our model utilised age, gender andimage information, all of which can be extracted from the Digital Imagingand Communications in Medicine (DICOM) image format, enabling conveni-ent deployment in a clinical setting. Its performance was validated in thecontext of an international challenge and proved robust through evaluationon different, multicentre datasets.In this work, we developed tools for CAD and CDS systems that can helpin the shift towards precision medicine. We demonstrated their potentialin oncological care for malignant melanoma and in the management ofCOVID-19.