Künstliche Intelligenz in der Pathologie: Status quo und Zukunftsperspektiven

During the course of increasing individualization of oncological diagnosis and treatment, ever larger datasets of complex clinical, molecular, radiological, and histopathological (imaging) data are becoming available. At the same time, advances at a technical and algorithmic level in the field of artificial intelligence (AI) render it possible to utilize this medical big data reservoir for diagnostic purposes and to predict prognosis as well as treatment response. Despite an increasing number of promising applications providing a proof of concept, the widespread use of AI procedures in everyday pathological practice is still a distant prospect. A prerequisite for this is the digitalization of pathological specimens, e.g., using whole-slide imaging (WSI). Key points for a successful AI transformation of pathology are likely to depend on an increase in the accuracy, reproducibility, explainability, interpretability, and resulting reliability of AI-based systems, accompanied by an early involvement of pathologists in this complex implementation task. Based on the current state of research, this review sheds light on the current and future significance of artificial intelligence for (routine) histopathological diagnostic methods.