Computer-Aided Diagnosis System for Breast Cancer Combining Digital Mammography and Genomics

摘要

This study investigated a computer-aided diagnosis system for breast cancer by combining the following three data sources: mammogram films, radiologist-interpreted BI-RADS descriptors, and proteomic profiles of blood sera. In this first year of the fellowship, we have collected calcification and mass data sets. To these data sets we have applied the following classification algorithms: Bayesian probit regression, linear discriminant analysis, artificial neural networks, as well as a novel method of decision fusion. For the calcification data set, the classifiers' performances under 100-fold cross validation were AUC = 0.73 for Bayesian probit regression, 0.68 +/- 0.01 for LDA, 0.76 +/- 0.01 for ANN, 0.85 +/- 0.01 for decision fusion . For the mass data set, the classifiers' performances under 100-fold cross validation were AUC = 0.94 for Bayesian probit regression, 0.93 +/- 0.01 for LDA, 0.93 +/- 0.01 for ANN, 0.94 +/- 0.01 for decision fusion. Decision fusion had a slight performance gain over the ANN and LDA (p = 0.02), but was comparable to Bayesian probit regression. Decision fusion significantly outperformed the other classifiers.