Estimation of LDL cholesterol by artificial neural networks

Artificial neural networks (ANNs) have become modeling tools that have found extensive acceptance and they have frequently used in applications in many disciplines for solving complex problems. Different ANN structures are valuable models, which are used in the medical field for the development of decision support systems. In this study, three multilayer neural networks trained with different algorithms were used for estimation of LDL Cholesterol and the most efficient training algorithm was determined. Gradient descent with momentum and adaptive learning rate backpropagation, Levenberg-Marquardt backpropagation, Bayesian regulation backpropagation were the studied three training algorithms. The multilayer neural networks were trained and tested with subject records from the database. Performance indicators and statistical measures were used for evaluating the multilayer neural networks and the results demonstrated that the Bayesian regulation backpropagation algorithm was the most efficient training algorithm for estimation of LDL Cholesterol.