An Intelligent Rule-based System for Status Epilepticus Prognostication

Status epilepticus is one of the most common emergency neurological conditions with high morbidity and mortality.

The aim of this study is to propose an intelligent approach to determine prognosis and the most common causes and outcomes based on clinical symptoms.

In this descriptive-analytic study, a perceptron artificial neural network was used to predict the outcome of patients with status epilepticus on discharge. But this method, which is understandable, is known as black boxes. Therefore, some rules were extracted from it in this study. The case study of this paper is data of Nemazee hospital patients.

The proposed model was prognosticated with 70% accuracy, while Bayesian network and Random Forest approaches have 51% and 46% accuracy. According to the results, recovery and mortality groups had often used phenytoin and anesthetic drugs as seizure controlling drug, respectively. Moreover, drug withdrawal and cerebral infarction were known as the most common etiology for recovery and mortality groups, respectively and there was a relationship between age and outcome, like in previous studies.

To identify some factors affecting the outcome such as withdrawal, their effects either can be avoided or can use sensitive treatment for patients with poor prognosis.