论文部分内容阅读
目的在美国放射学会推荐并在国内外广泛使用的乳腺报告和数据系统 BI-RADS基础上，通过量化影响评估分类的变量，建立适合国人的 BI-RADS评估判别，以准确客观进行乳腺影像诊断的评估分类。方法搜集过去10年中561例有完整影像学资料，及由病理或随访结果支持的 BI-RADS分类结果，参考NCCN2011诊断指南，对与乳腺癌诊断有关的各种变量进行量化，基于Bayes判别分析方法建立判别模型；分别进行根据影像学评估分类和 Bayes判别量化分类，评估两种方法下医生之间的诊断一致性。结果建立了完整的判别分析模型，其误判概率为1.25%。Bayes量化判别判别分类的诊断正确率高于单纯根据影像学分类时的正确率，以BI-RADS3、4类明显。分类结果的组间相关系数为0.987，软件分类结果的组间相关系数为0.999。结论通过大样本的统计学分析建立标准、合理的判别分析模型，能够有效的预测乳腺肿块的BI-RADS类型，减少BI-RADS 0类的召回率，提高BI-RADS1~5类病变的诊断准确率，并提高不同诊断者的一致性。“,”Objective Breast Imaging-Reporting and Data System (BI-RADS) is recommended by American Col ege of Radiology and currently, widely used in China. To build up a suitable and unbiased BI-RADS assessment system for the population in China, we classified the assessment of breast diagnostic imaging by quantization ef ects from dif erent assessment variables.Methods Medical image data were col ected from 561 patients in the past 10 years, which pathological data and fol ow-up results supported the classification by BI-RAD. Referring to clinic guideline of NCCN2011, we quantize dif erent variables of diagnosis of breast cancer and establish the discrimination model based on Bayes discriminant analysis. To assess the consistency of diagnostic results, we analyzed the assessments by imaging discriminant and Bayes discriminant, respectively. Results An entire discriminant analysis model is established and the error probability is 1.25%. The accuracy of classification by Bayes quantitative discrimination is higher than that by simply imaging discriminant. Among these, class 3 and 4 are clearer in BI-RADS. The intra-family correlation of classification result is 0.987, while the intra-family correlation by software id 0.999.Conclusion By establishing standards from statistical analysis of large samples and reasonable discriminant analysis model, we can ef ectively predict the BI-RADS type of breast tumor and reduce the recal rate of BI-RADS 0 type. We also can improve the diagnostic accuracy of BI-RADS 1~5 types and the consistency from dif erent diagnose.