Implementation of artificial neural network for identification of acute appendicitis

Abstract

At present, the morbidity of appendicitis is very high, this certainly makes appendicitis a priority health issue because it has a large impact on public health. The purpose of this study is to obtain a more accurate diagnostic method so that it is expected to be used in reducing morbidity and mortality rated due to inappropriate and rapid handling of appendicitis. Research on the identification of acute appendicitis using Artificial Neural Networks has been conducted. In this study used 154 medical records of inpatients with acute appendicitis and not acute appendicitis from Airlangga University Hospital. The identification program is made by training and testing the program using Backpropagation Neural Network algorithm. The training conducted showed that there was a significant influence between the maximum iteration, the MSE limit, the number of hidden neurons, and the learning rate used on the final weights of the resulting network, thus greatly affecting the percentage of testing. The training and testing of the program carried out with various combinations of input parameters, shows that the best parameters that can be used in this program are: 1000 times the maximum iteration, MSE limit of 0.000001, learning rate of 0.1, and hidden neurons of 5, which with these parameters, the program succeeded in achieving accurate identification of test data up to 98.4375%. Based on the results that have been obtained, this method is expected to be useful for the progress of the diagnostic system, especially in cases of acute appendicitis.