To evaluate the performance of our neural network classification model, the probability of misclassification error and the area of the ROC curve for the model were estimated using a nonparametric approach based on a resubstitution method [36]. For the resubstitution method, the neural network classification model was trained using a pattern dataset, which included all of the available 830 clinical instances, and was tested to estimate the probability of misclassification error and the area of the ROC curve using the same pattern dataset. Generally, in this research, the resubstitution method resulted in optimistically unbiased estimates of the asymptotic probability of misclassification error and the area under the ROC curve since the pattern dataset, containing the 830 clinical instances, can be considered a relatively large dataset.