基于BI-RADS描述的乳腺肿瘤判别

Determination of Breast Cancer Based on BI-RADS

目的探讨BI-RADS在乳腺肿瘤检查中的应用价值。方法先是预处理大量的历史数据集含BIRADS特征和对应的病理结果,然后基于这些数据集进行决策树算法建模,最后采用留一法(leave-one-out cross validation)验证模型的性能。结果对实验中收集的361例(良性182例,恶性179例)进行留一法验证,实验结果准确率为91%,敏感性为87%,特异性为95%,受试者工作特征曲线下面积为95%。结论基于BI-RADS描述的乳腺肿瘤判别方法十分有效,有望为临床诊断提供有价值的参考意见。

Objective To assess the value of BI-RADS（Breast Imaging-Reporting and Data System ）in the examination of the breast cancer. Methods First to deal with a lot of historical data including the feature of BI-RADS and the corresponding pathological findings,Then to model the decision tree algorithm based on these datas.Finally to use the “leave-one-out cross validation”to verify the performance of the model. Results the pathological examination （ revealed 1 82 benign and 1 79 malignant lesions ） was verified .The accuracy,sensitivity and specificity of the result were 91%,87% and 95%.Area under the ROC curve was 95%. Conclusion The discrimination method of breast cancer based on BI-RADS is very effective,it is expected to provide valuable reference for clinical diagnosis.