ONTOLOGY DRIVEN DECISION SUPPORT FOR EARLY DIAGNOSTIC RECOMMENDATIONS

Purpose:

Studies have demonstrated how providing early diagnosis to clinicians can improve diagnostic accuracy. It is also necessary to present the diagnostic recommendations in an appropriate ranked list to support the clinician in decision making. However the techniques and methods that can be used to generate the diagnostic recommendations throughout the clinical information gathering process need further research and development. In addition we need tools that can provide semantically interoperable diagnostic recommendations using current web standards.

Method(s):

Electronic Medical Record (EMR) systems are the main interface through which clinicians access patient data. Diagnostic Clinical Decision Support Systems (CDSS) need to access this patient data from the EMR at various stages of the clinical encounter in order to provide diagnostic recommendations that can guide further information gathering or help formulate a differential diagnosis. 

We have developed a novel ontology driven diagnostic decision support system that can remotely access patient data from a test EMR platform, and provide a ranked diagnostic recommendation list with weights reflecting the degree of confidence in the diagnosis. A dental diagnostic decision support system was developed using OWL (Web Ontology language) and SPARQL (SPARQL Protocol and RDF Query Language) rules. The diagnostic criteria were encoded using OWL and SPARQL inference rules. Apache Jena  was used as the platform to develop the recommendation engine.

OpenEMR was used as the test EMR platform. Dental Information Model (DIM) V.1.0 was used as the basis for creating forms within OpenEMR. DIM separates the information collected during a dental encounter into various processes and sub-processes. This includes processes like Chief Complaint, Extra Oral Examination, Intraoral Examination, Radiographic examination followed by final Diagnosis. The patient data was recorded in a MySQL database. The data was then converted to RDF (Resource Description Format) using the D2RQ platform. D2RQ also provides a SPARQL endpoint that allows the decision support system to access the patient data.

Result(s):

The results were displayed in a web page that would provide a ranked list of the diagnosis after querying a SPARQL endpoint with the decision model and patient data.   

Conclusion(s):

Our diagnostic recommender tool demonstrates the ability of the ontology driven decision support system to support information gathering and potentially reduce diagnostic error with the help of timely diagnostic recommendations.