Neural networks significantly improve cancer staging accuracy

Summary form only given. Survival prediction is important in cancer because it determines therapy, matches patients for clinical trials, and provides patient information. Is a backpropagation neural network more accurate at predicting survival, in breast cancer, than the current staging system? For over thirty years, cancer outcome prediction has been based on the pTNM staging system. There are two problems with this system: (1) it is not very accurate, and (2) its accuracy cannot be improved because predictive variables cannot be added to the model without increasing the model´s complexity to the point where it is no longer useful to the clinician. The Surveillance, Epidemiology, and End Results data set, for the years 1977 - 1985, is used to compare the predictive accuracy of the current pTNM stage system to a backpropagation neural network, for five-year breast cancer survival. The c-index is the measure of accuracy. The pTNM stage system has a c-index of .69, while the backpropagation neural network, using the same three variables, has a c-index of .73 (SE for both models is less than .01). Using the same variables, the backpropagation neural network is significantly more accurate at predicting five year survival, in breast cancer, than the current pTNM stage system