Computer aided methods are an integral part of the medical diagnosis process. They support radiologists and other medical doctors with functionalities such as segmentation, registration, or quantification. Expert systems aim for a broad functionality based on an extensive knowledge base. They perform diverse tasks, ranging from assessing single, e.g. pulmonary functions to supporting intensive care personal. However, these expert systems are not as broadly acceptanced as envisioned by their developers, because the medical expert has to intensly curate the knowledge base, which then in turn become subject to interobserver differences.
Modern deep learning systems promise to address this downside of contemporary expert systems, because they do not require a curated knowledge base. In addition, such systems promise to enable new breakthroughs, which have not been imagined before, due to the high accuracies obtained.
Yet, the application of artificial intelligence to the field of medicine is a challenge in itself. The first step is to combine the benefits of knowledge- and learning-based systems to allow for state of the art accuracy in medical diagnosis. Second, medical doctors have to be put into position to understand resulting diagnoses and predictions. Thus we have to consider the embeddedness of AI output in the medical decision process and its ethical implications.
This spring school event introduces and addresses related research questions with key note input of experts of the domain of artificial intelligence, namely explainable artificial intelligence, learning, and visualisation. We address the following topics: