gms | German Medical Science

Motivation: Arden Syntax for Medical Logic Module (MLM) was designed for writing and sharing medical rules [1]. Several researchers have developed Arden engines to compile and execute MLMs [2], [3], [4]. The mentioned Arden engines translate the MLMs term by term into another program language. In the last few years, building the medical knowledge packages (MKPs) with hierarchical and nested MLMs is a popular research topic [5]. For processing the hierarchical and nested MLMs in MKPs, the mentioned Arden engines appear to be not efficient. Designing an Arden engine which can match the hierarchical and nested MLMs efficiently is our objective.

Method: Compare the mentioned Arden engines with business rule engines.

Results: Our comparison showed that the so-called RETE algorithm [6] is an efficient efficient pattern matching algorithm for interpreting production rules. This algorithm has been used in many rule based systems, including Drools, CLIPS, and iLogic, to speed up the rule matching. From these experiences, we think that it is possible to integrate the RETE algorithm with an Arden engine. Such Arden engine should be particular suitable for hierarchical and nested MLMs in MKPs. Although RETE algorithm is implicitly used in EGADSS framework [3], but it is not an Arden engine using RETE algorithm. The reason is that EGADSS converts the MLMs into CLIPS at first, and then the CLIPS will be converted into JAVA using RETE algorithm. We think that an Arden engine should convert MLMs into a goal program language using RETE algorithm directly.

Discussion: RETE algorithm is an effective way to accelerate rule matching, but its effectiveness is dependent on the scale of the knowledge base and the hardware system. However, we will implement an Arden engine including the RETE algorithm and evaluate its effectiveness.