Artificial neural networks for estimating glomerular filtration rate by urinary dipstick for type 2 diabetic patients
Journal
Biomedical Engineering - Applications, Basis and Communications
Journal Volume
28
Journal Issue
3
Date Issued
2016
Author(s)
Abstract
Background: The result of a standard urinary dipstick from a patient with diabetes mellitus type 2 can be used to predict the estimated glomerular filtration rate (eGFR). We designed a multilayer perceptron (MLP) to investigate the possibility and optimal number of variables for the prediction. Methods: A total of 299 volunteers with diabetes mellitus type 2 were included. The blood and urine samples from volunteers were analyzed for blood sugar, glycated hemoglobin, serum creatinine, and urine chemistry. The urine chemistry was examined by a standard urinary dipstick. Volunteer age and gender and six test items of the dipstick were set as eight variables for this study. The eight variables were grouped and examined for the optimal combination. The eight variables from 232 of 299 volunteers were used to train an MLP for the optimal variables. The performance of trained MLP was validated by the data from 69 of 232 volunteers. Results: The optimal combination for variables was the six test items of the dipstick and volunteer age. The area under the curve (0.928), accuracy (0.879), sensitivity (0.83), and specificity (0.88) of the trained MLP were examined. Conclusions: The results demonstrate the eGFR prediction potential of the results of a urinary dipstick using this method. ? 2016 National Taiwan University.
Subjects
Chronic kidney disease; Diabetes mellitus type 2; Estimated glomerular filtration rate; Multilayer perceptron; Urinary dipstick
SDGs
Other Subjects
Blood; Chemical analysis; Forecasting; Multilayer neural networks; Multilayers; Neural networks; Area under the curves; Blood and urine samples; Chronic kidney disease; Diabetes mellitus type 2; Glomerular filtration rate; Glycated hemoglobins; Multi layer perceptron; Urinary dipstick; Body fluids; creatinine; glucose; hemoglobin A1c; adult; area under the curve; Article; artificial neural network; blood sampling; chronic kidney disease; controlled study; creatinine blood level; diabetic patient; diagnostic accuracy; diagnostic test accuracy study; female; glomerulus filtration rate; glucose blood level; hemoglobin blood level; human; kidney function test; major clinical study; male; middle aged; non insulin dependent diabetes mellitus; perceptron; point of care testing; polymerase chain reaction; sensitivity and specificity; urinalysis; urinary dipstick; urine chemistry
Publisher
World Scientific Publishing Co. Pte Ltd
Type
journal article