Fig. 1

Workflow of AI in histopathology and clinical genomics. In this simplified workflow, a tissue of a solid tumor is harvested via surgery or biopsy. One part is sequenced in the genomics facility to obtain molecular data about, for instance, RNA, epigenetics, or mutations, while another part is sent to the pathology department. There, tumor slices are captured on glass slides and stained with hematoxylin and eosin (H&E). Images of these glass slides can then be taken. Tabular and image data are used to train models, e.g., neural networks to provide a prediction. In this review, we describe six distinct medical application tasks (Diagnosis, Grading, Subtyping, Mutation, Response, and Survival) for these models