Abdominal Pain Estimation in Childhood based on Artificial Neural Network Classification

Artificial Neural Networks (ANNs) are particular implementations of AI (artificial intelligence) systems. ANNs have established themselves as powerful tools in clinical practice whenever disease prognosis is based upon the statistical analysis of a set of similar cases characterized by specific clinical data that describe the physical condition of the patient. This study examines the implementation of an ANN architecture for the estimation of abdominal pain in children, which is a critical factor in deciding upon performing a surgical operation of the abdomen. To our knowledge, this approach is the first computational intelligent method based on ANNs that deals with abdominal pain prediction. The proposed ANN implements a multilayer perceprton (MLP) architecture featuring an input layer of 16 nodes, a hidden layer of 5 neurons and an output layer of a single neuron. The decision between applying conservative treatment or performing a surgical operation depending upon the particular exhibits of each case is reached automatically by the output of the proposed ANN estimator of abdominal pain. The proposed ANN attains a percentage of 97% of successful prognosis in cases that belong to the testing set. All pathological cases belonging to the training set are classified correctly. The proposed method may be used as a software tool that assists surgeons in making a diagnosis speeding up thereby the entire examination procedure in emergency