Neural networks for measuring cancer outcomes

Harry B. Burke, Philip H. Goodman, David B. Rosen

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations


Background. The TNM staging system has been used since the early 1960's to predict breast cancer patient outcome. In an attempt to increase prognostic accuracy, many putative prognostic factors have been identified. Because the TNM stage model can not accommodate these new factors, the proliferation of factors in breast cancer has led to clinical confusion. What is required is a new computerized prognostic system that can test putative prognostic factors and integrate the predictive factors with the TNM variables in order to increase prognostic accuracy. Methods. Using the area under the curve (AUC) of the receiver operating characteristic, I compare the accuracy of the following predictive models: pTNM staging system, principal component analysis, classification and regression trees, logistic regression, cascade correlation neural network, conjugate gradient descent neural network, probabilistic neural network, and backpropagation neural network. Results. The pTNM staging system's accuracy is. 720. Logistic regression (LR) and both the probabilistic neural network (PNN) and the backpropagation neural network (BPNN) are significantly more accurate than the pTNM staging system, using just the TNM variables (.762, .759, and .768, respectively). Adding variables further increases the prediction accuracy of LR and both PNN and BPNN (.776, .777, .779, respectively). Adding the new prognostic factors p53 and HER-2/neu increases the backpropagation neural network's accuracy to .850. These results generalize across breast cancer data sets and to a colorectal cancer data set. Conclusions. Computerized prediction systems are more accurate than the current look-up table system. The backpropagation neural network is consistently more accurate than the best conventional statistical models. Artificial neural networks are able to combine prognostic factors to further improve prognostic accuracy. Artificial neural networks are robust across data bases and cancer sites, they can perform as well as the best traditional prediction methods, and they can capture the power of nonmonotonic predictors and discover complex interactions.

Original languageEnglish
Title of host publicationConference Proceedings - 10th Anniv., IMTC 1994
Subtitle of host publicationAdvanced Technologies in I and M. 1994 IEEE Instrumentation and Measurement Technology Conference
PublisherInstitute of Electrical and Electronics Engineers Inc.
Number of pages3
ISBN (Electronic)0780318803, 9780780318809
StatePublished - 1994
Event1994 IEEE Instrumentation and Measurement Technology Conference, IMTC 1994 - Hamamatsu, Japan
Duration: 10 May 199412 May 1994

Publication series

NameConference Proceedings - 10th Anniv., IMTC 1994: Advanced Technologies in I and M. 1994 IEEE Instrumentation and Measurement Technology Conference


Conference1994 IEEE Instrumentation and Measurement Technology Conference, IMTC 1994


Dive into the research topics of 'Neural networks for measuring cancer outcomes'. Together they form a unique fingerprint.

Cite this