ARTIFICIAL NEURAL NETWORKS IN PEDIATRIC UROLOGY - PREDICTION OF SONOGRAPHIC OUTCOME FOLLOWING PYELOPLASTY

Purpose: Computerized artificial neural networks are analogous to biol ogical neuronal systems. Since they may be trained to recognize the re levance of complex patterns in data, neural networks may be useful for decision making in the multifactorial management of ureteropelvic jun ction obstruction. We determine the ability of a customized neural net work to predict sonographic outcome after pyeloplasty in children with ureteropelvic junction obstruction. Materials and Methods: A data set was constructed with 242 demographic, clinical, radiological and surg ical elements. We analyzed the available retrospective data in 100 con secutive children who underwent unilateral pyeloplasty for ureteropelv ic junction obstruction chosen from all 144 surgically treated for ure teropelvic junction obstruction between 1993 and 1995. One radiologist reviewed all film data and provided a final sonographic outcome desig nation in each case. We wrote a set of computer programs to construct a neural network. A composite 4-layer network was built with output no des representing 4 possible sonographic outcomes. The 100 patient data set was randomly divided into 84 training and 16 testing examples. Re sults: The neural network correctly predicted all 5 of 5 significantly improved, 7 of 7 improved, 2 of 2 same and 2 of 2 worse sonogram resu lts after pyeloplasty. Therefore, sensitivity and specificity were 100 % for all 4 outcomes. Linear regression analysis of the data yielded i nferior sensitivity and specificity values (52 to 94%), confirming tha t ureteropelvic junction obstruction is a nonlinear data analysis prob lem. Conclusions: The 100% accuracy, sensitivity and specificity of ou r neural network in this pilot study provide evidence of the value of the neural computational approach for the modern exploration and model ing of the clinical problem of pediatric ureteropelvic junction obstru ction.