Mining Diagnostic Rules of Breast Tumor on Ultrasound Image Using Cost-sensitive RuleFit Method

摘要：

In the medical diagnosis,the false negativeprediction is more serious than the false positiveprediction.We introduce the cost-sensitive ruleensemble method(RuleFit)to breast ultrasound whichcan induce the interpretable scoring rules formalignancy assessment,and can be applied to tune thesensitivity and specificity of the predictive model byvarying the cost weights of misclassification.TheGentleCost boosting algorithm is proposed to generatethe decision tree ensemble.Then,we use the modifiedRuleFit method with the cost-weighted loss function toselect and fit the rules decomposing from the treeensemble.Experiments results on a breast ultrasoundimage dataset(168 cases)with the varying costweights demonstrate that the final rule ensemblecontain only 22(among total 600 decomposed rules)rules with the comparable performance to the treeensemble.The examples of the rule ensemble for breastultrasound and its interpretation are also illustrated

作者: Wei Yang Su Zhang Yazhu Chen Yaqing Chen Wenying Li Hongtao Lu

作者单位: Department of Biomedical Engineering,Shanghai Jiao Tong University,Shanghai,China Shanghai Sixth Hospital,Shanghai Jiao Tong University,Shanghai,China Department of Computer Science,Shanghai Jiao Tong University,Shanghai,China

会议类型: 国际会议

会议名称: 2008 3rd International Conference on Intelligent System and Knowledge Engineering(第三届智能系统与知识工程国际会议)(ISKE 2008)

会议地点: 厦门

会议语种:英文

页码: 354-359

在线出版日期: 2008-11-17（万方平台首次上网日期，不代表论文的发表时间）