Assessment of bilharziasis history in outcome prediction of bladder cancer using a radial basis function neural network

Investigates the potential value of bilharziasis history in predicting the outcome progress of patients with bladder cancer using a radial basis function (RBF) neural network. The data set is described by eight input features: histology, tumour grade, lymph nodes status, bilharziasis history, stage, DNA ploidy, sex, and age interval. Two outcomes are of interest: recurrence of disease and death within five years of diagnosis. The total number of patients was 321, of whom 83.5% had been confirmed with bilharziasis history. Different feature subsets have been examined to improve the predictive accuracy and to assess the effect of bilharziasis. The highest predictive accuracy is 74.07% from the RBF network. The analysis shows that bilharziasis history is an important prognostic marker in the prediction