Discovering interobserver variability in the cytodiagnosis of breast cancer using decision trees and Bayesian networks

摘要: We evaluate the performance of two decision tree procedures and four Bayesian network classifiers as potential support systems in cytodiagnosis breast cancer. In order to test their thoroughly, we use real-world databases containing 692 cases 322 collected by a single observer 19 observers, respectively. The results show that, general, there are considerable differences all tests (accuracy, sensitivity, specificity, PV+, PV- ROC) when specific classifier uses single-observer dataset compared those this same multiple-observer dataset. These suggest that different observers see things: problem known interobserver variability. graphically unveil such presenting structures trees networks resultant from running both databases.