VERIFICATION OF AUTOMATA MODELS FOR CLINICAL DECISION SUPPORT SYSTEMS

Abstract

In the treatment of any disease, the doctor must strictly follow the clinical protocol. However, this document cannot be fully considered as a graph diagram of the algorithm of the doctor’s actions, since most of the transition rules in it cannot be implemented without the involvement of the doctor’s knowledge. So, based on the clinical protocol, it is necessary to create a clinical decision support system (CDSS). Existing CDSSs do not allow describing the clinical protocol as a sequence of actions. This drawback can be eliminated by using the automata model as a CDSS inference engine, since the terminology of the treatment process itself leads to a natural thought - to model this process in states, i.e. use an automata approach. At the same time, it is vital to ensure the extensibility of the automata model, which will allow among other things for taking into account when choosing drugs any possible adverse reactions to drugs already prescribed. The article discusses the process of development and verification of an automata model as an inference engine for CDSS. The process, based on the Automata approach and performed offline, uses a sequence of refinements, which allows you to get a more detailed model that reflects the typical options for the doctor. In addition, with each replacement or addition of the drug, online testing of the model is performed, during which compatibility with the already prescribed drugs and the patient's history is checked. The developed process is applied to the construction and verification of CDSS for the management of patients with multiple sclerosis. to automata