Non-invasive diagnosis of phenylketonuria by using artificial neural networking and nuclear magnetic resonance spectroscopy

Abstract

Introduction: Phenylketonuria (PKU) is a relatively common metabolic disease in world .The high incidence of the disease in our country is due to consanguineous marriages. The prevalence of this disease in Iran is reported to be about 1: 4000 to1: 8,000 births per year. Mental retardation, physical disabilities, neurological disorders are the clinical symptoms of the disease. Early diagnosis is very important to prevent the disabling consequences of the disease. The purpose of this study was to use a multi-layer neural network perceptron (MLP) to build a model for early detection and treatment of phenylketonuria patients.. Materials and Methods: Urine samples were obtained from healthy and PKU children. nuclear magnetic resonance spectroscopy was performed in NMR 400 MHz Bruker with the help of NOESY Protocol. Then peak resonance of each metabolite was identified, and modeling was done with multi-layer neural network perceptron. Results: The Model build in this study was able to classify the data in two groups of patient and healthy individuals successfully, with more than 90% sensitivity and 0.2% error rate with high predictive power Conclusion: Our results showed the high power capability of this technique to diagnose the Phenylketonuria with the help of NMR spectroscopy and artificial neural network