基于三种预测模型的乳腺超声BI-RADS分类3~5类结节恶性危险分层量化评分研究

Study on quantitative scoring of malignant risk stratification of breast ultrasound BI-RADS Category 3-5 nodules based on three prediction models

目的 探讨更加客观、准确、规范、简单易推广的乳腺结节量化评估分类标准。方法 回顾性分析2019年1月至2023年2月在深圳市人民医院及深圳市龙华区妇幼保健院行超声检查并有明确病理结果的644个乳腺结节,应用SPSS统计学软件随机分为训练组(n=439)和测试组(n=205)。收集乳腺结节的超声声像特征、年龄、病理结果等资料,分别建立基于计数法、权重法和回归公式法的量化评分预测模型,并绘制受试者工作特征(ROC)曲线,以ROC曲线下面积(AUC)和测试试验对标病理结果分析各预测模型的诊断效能,并与BI-RADS分类的诊断效能进行比较,筛选出兼具有效性和特异度的最优模型。结果 训练组的439个乳腺结节中,良性结节328个,恶性结节111个。基于计数法、权重法、回归公式法的评分标准均为客观地评估乳腺结节的分类标准,诊断效能均优于传统BI-RADS分类。基于权重法和回归公式法的预测模型对良恶性乳腺结节的诊断价值高于计数法,基于权重法和回归公式法的预测模型对良恶性乳腺结节的诊断价值差异无统计学意义(P> 0.05)。结论 基于权重法建立BI-RADS分类量化评分标准更客观实用,易于临床推广。

Objective To explore more objective,accurate,standardized,simple and easy-to-popularize classification criteria for quantitative evaluation of breast nodules.Methods A total of 644 breast nodules with definite pathological results at Shenzhen People’s Hospital and Shenzhen Longhua District Maternal and Child Health Hospital from January 2019 to February 2023 was retrospectively analyzed and randomly divided into the training group(n=439)and the test group(n=205)by SPSS software.The ultrasonic image features,patient age and pathological results of breast nodules were collected,and quantitative scoring prediction models based on counting method,weight method and regression formula method were established respectively,and the receiver operating characteristic(ROC)curve was drawn.The diagnostic efficiency of each prediction model was analyzed by the area under curve(AUC)of ROC and the pathological results of test,and then it was compared with the diagnostic efficiency of BI-RADS classification to select the optimal model with both high efficiency and specificity.Results Among the 439 breast nodules in the training group,there were 328 benign nodules and 111 malignant nodules.The scoring criteria based on counting method,weight method and regression formula method were all classification criteria for objectively evaluating breast nodules,with diagnostic efficiency better than that of traditional BI-RADS classification.The predictive model based on weight method and regression formula method had higher diagnostic value for benign and malignant breast nodules than that based on counting method,and there was no statistical difference between the former two models(P>0.05).Conclusion The quantitative scoring criteria of BI-RADS classification established based on weight method is more objective and practical,and easy to be popularized in clinical practice.