An Approach to Automated Extraction of Diagnostic Rules From the Text of Clinical Guidelines for Decision Support Systems

Currently, there is a large amount of accumulated medical knowledge about various diseases, formalized in the form of clinical guidelines. For general practitioners, it is difficult to remember several dozen documents, due to information overloaded. To solve this problem, medical decision support systems (DSS) are used. Those DSS based on digitized clinical guidelines, which are used to help doctors make more accurate diagnoses and prescribe treatment to patients. Periodic updating of clinical practice guidelines raises an additional problem with timely updates to the rules in the DSS. Natural language analysis methods were used to extract information from clinical guidelines. As a result of the analysis of natural language analysis methods, the dependency grammar method was chosen as the most suitable for the Russian language. In conclusion we has been built a prototype of a program for extracting information from the text of clinical guidelines. This program allows extracting of "if-then" rules in the "treatment" Chapter, whose performance has been tested on clinical guidelines for various diseases.