The System for Neurological Analysis of Patient Symptoms: An Integrated Artificial Intelligence Prototype

J.L. Sponsler, G. Goltukhchyan, F. Van Scoy, and J. Culberson (USA)


Expert system, neurology, stroke, neuroanatomy, naturallanguage processing


Objectives: To create a stroke expert system prototype that reads a patient text file, computes a disease location, and presents brain graphics showing the lesion. Methods: Coded in Common Lisp, The Natural Language Module reads a text file, parses, and produces parse trees. An Information Extraction Module employs rules to extract data from parse trees and to store these into a database. A Neuro-Anatomic Atlas has been created to represent clinically relevant nervous system structures. A Localization Module employs rules and methods to map from a clinical finding to an anatomic structure. The Graphics Module presents representations of brain structures corresponding to the localization. A 3D Brain Model has been created by scanning a specimen and is a part of this module. Results: Text files representing stroke syndromes are analyzed by the system which produces localization hypotheses and graphical representations of the structures. Formal controlled trials of program behavior are planned. Conclusions: The prototype demonstrates feasibility and should, as new knowledge is incorporated, evolve into a utility for diagnosis, instruction, and experimentation. Formal testing will be required in later stages of the project.

Important Links:

Go Back