Neural networks significantly improve cancer staging accuracy

Cancer outcome prediction has been largely based on the pTNM staging system but this system presents two problems: (1) low accuracy rate and (2) no room for accuracy improvement because addition of predictive variables increases complexity. In this light, a comparison was made on the predictive accuracy of the current pTNM stage to a backpropagation neural network, for five-year breast cancer survival. This undertaking used the c-index as the measure of accuracy, which ranges from 0.5 (chance) to 1.0 (perfect prediction). Under the same variables, the pTNM stage system had a c-index of 0.69, while the backpropagation neural network scored 0.73, indicating better prediction accuracy.