Adaptive modelling of creatinine concentration in human blood

Abstract

The need and importance of modelling and evaluation of renal function has grown with the increasing need for kidney replacement, design of artificial kidney and proper diagnosis and treatment of kidney disorders. This paper explores a fuzzy model of the creatinine concentration (C r ) in human blood using adaptive Neurofuzzy Inference System (NFIS). This model aims to predict blood creatinine using blood urea, sodium and potassium without a need for actual measurement of C r in blood. The creatinine level in blood is an efficient tool for evaluation and assessment of renal function. The proposed model was validated by comparing the predicted creatinine values from the model with the experimental creatinine values. The obtained results showed that the developed model is capable of estimating the blood creatinine with a Root-Mean Square Error (RMSE) of 0.76 and also the predicted C r values agree closely with the experimentally measured C r values. Practically, this implies that the developed neurofuzzy model is adequate one for prediction of creatinine concentration in human blood and consequently will lead to a proper evaluation of renal function. He is an Associate Professor at the Biomedical Engineering Department in Jordan University of Science and Technology. His research interests include design of intelligent medical systems, mobile-phone-based medical systems, fuzzy and neuro-fuzzy-based systems, intelligent medical sensors, intelligent therapeutic systems and remote monitoring of physiological signals. He has two granted patents and published more than 30 research papers in international journals and conference proceedings.