Prototypical Case Mining from Medical Literature

This article addresses the task of mining cases from biomedical literature to automatically build an initial case base for a case-based reasoning (CBR) system. This research takes place within the Mémoire project, which has for goal to provide a framework to facilitate building CBR systems in biology and medicine. By analyzing medical literature, the ProCaseMiner system mines for medical concepts such as diseases, signs and symptoms, laboratory tests, and treatment plans all connected together in a given medical domain. It then organizes these concepts in a structure called a case. This case mining component provides a definite help to start-up the creation of a biomedical CBR system case base, composed of both concrete cases and prototypical cases. This article validates the approach by presenting a comparison between the prototypical cases learnt from stem-cell transplantation domain with those created by a team of experts in the domain.