Artificial intelligence can play an important role in healthcare by analysing medical images to provide diagnostic decision support to medical professionals. This thesis explores novel computer-aided diagnosis methods, focusing on two key aspects that are vital for clinical adoption: accuracy and interpretability.