Diagnosis of Different Types of Hyperbilirubinemia Using Artificial Neural Network

Abstract

Hyperbilirubinemia is a common condition in newborn children, as they are inapt to easily discharge bilirubin which induces its concentration in the blood and within other fluids in the body. This condition can be a serious signal for a more severe underlying disease. Its frequency is justification enough for the further development of early diagnostic techniques. The modernization of clinical diagnosis methods has paved the way for the application of artificial intelligence. This is often an enhancement in accuracy and affordability in comparison to traditional diagnosis methods and tools such as bilirubin meters. The main purpose of this paper is to show the possibility of using artificial neural networks (ANNs) in the classification of different types of hyperbilirubinemia. The neural network which was developed in this research is two-layer feedforward artificial neural network with 13 input parameters: gender, age, skin color, mucosa color, sclera color, urine color, stool color, bilirubin in urine, urobilinogen in urine, urobilinogen in a stool, total bilirubin level, conjugated bilirubin level, unconjugated bilirubin level, and liver enzyme (ALT and ALP) levels. This work outlines the testing of samples from 360 patients who are either healthy or have a syndrome associated with hyperbilirubinemia. The study successfully assesses the sensitivity and accuracy of this diagnostic approach and evaluates it against the data set from the afore mentioned sample.