Hertzberg T M, Tremblay G F, Lam C F
Division of Neurology, Children's Hospital of Philadelphia.
Comput Biomed Res. 1987 Oct;20(5):489-96. doi: 10.1016/0010-4809(87)90036-x.
Expert systems to assist in neurological diagnosis require a representation of anatomical relationships. In order to test one representational method, a prototype expert system was developed. It accepts patient signs of neurological dysfunction and identifies the site of nervous system injury. The system's knowledge base is contained in a semantic network which represents nervous system anatomy and the physical signs of injuries. When provided with an individual's physical signs, the network is searched by a simple algorithm; the anatomical locations which best explain the physical signs are the system's output. Medical expert systems which require anatomical reasoning can use a direct representation of spatial relationships to avoid the difficulties of encoding clinical associations in the form of If-Then rules.
