Portions of this document were adapted in part from our paper:
Nepple KG, Knudson MN, Austin JC, Wald M, Makhlouf AA, Niederberger CS, Cooper CS: Adding Renal Scan Data Improves the Accuracy of a Computational Model to Predict Vesicoureteral Reflux Resolution: The Journal of Urology, Volume 180, Issue 4 (October 2008) pages 1648-52.
The reader is strongly encouraged to read this paper prior to using the network, as it contains a more specific description of the model.
Previous studies suggest that an abnormal renal scan may be a predictor of the failure of vesicoureteral reflux to resolve. We investigated whether the addition of renal scan data would improve the accuracy of our previously published computational model. Medical records and renal scans were reviewed on 161 children with primary reflux.
In addition to the 9 input variables from our prior model, we added renal scan data on decreased relative renal function (40% or less in the refluxing kidney) and renal scars. Resolution outcome was evaluated 2 years after diagnosis. Data sets were prepared for outcomes, and randomized into a modeling set of 111 and a cross-validation set of 50. The model was constructed using "neUROn2++", a suite of C++ programs we designed to implement neural computational and statistical algorithms. The results were that a logistic regression model had the best fit with an ROC area of 0.945 for predicting reflux resolution in the 2-year model (see Table). This was improved compared to our previous model without renal scan data.
| Method | ROC Area1 |
| Logistic Regression | |
| Radial Support Vector Model | |
| Linear Support Vector Model | |
| Linear Discriminant Function Analysis | |
| 3 Hidden Node Neural Network |
1Receiver Operating Characteristic Curve area. Numbers approach 1.0 as accuracy improves (a value of 1.0 would indicate sensitivity and specificity both 1.0). Areas were computed using the statistical method described by Wickens: Wickens TD, Elementary signal detection theory, New York: Oxford University Press, 2002.