超声误诊为乳腺癌临床特征与风险预测模型构建

Clinical Characteristics Ultrasound Misdiagnosis of BreastCancer and Construction of Risk Prediction Model

目的分析超声误诊为乳腺癌临床特征,并构建误诊风险预测模型。方法选取2021年1月至2023年12月行超声检查的乳腺肿块患者602例,按照误诊情况分为误诊组和诊断准确组,分析超声误诊为乳腺癌病例特征,并构建误诊风险预测模型。结果共有30例误诊为乳腺癌,误诊率为4.98%。诊断准确组与误诊组在病灶最大径、触诊、合并良性病灶、合并炎性病变、超声图像边缘、血流信号明显、腋下淋巴结回声、微钙化、病灶结构复杂多样方面比较差异有统计学意义(P<0.01)。经二元Logistic回归分析显示,病灶最大径(≥5 cm)、合并良性病灶、合并炎性病变、超声图像边缘(模糊)、血流信号明显、腋下淋巴结回声、微钙化、病灶结构复杂多样是超声误诊为乳腺癌的危险因素(P<0.05,P<0.01)。超声误诊为乳腺癌风险预测模型:2.623×病灶最大径(≥5 cm)+1.422×合并良性病灶+1.616×合并炎性病变+1.574×超声图像边缘(模糊)+1.134×血流信号明显+1.518×腋下淋巴结回声+2.027×微钙化+1.541×病灶结构复杂多样。Hosmer-Lemeshow拟合度检验显示,该模型拟合优度较好(χ^(2)=3.487,P=0.900);受试者工作特征曲线分析显示,预测模型预测超声误诊为乳腺癌风险的曲线下面积为0.921,约登指数为0.698,敏感度、特异度分别为83.3%、86.5%,95%CI为0.872,0.969,实际应用准确性为96.2%(579/602)。结论超声误诊为乳腺癌和多种因素有关,基于病灶最大径、合并良性病灶、合并炎性病变、超声图像边缘模糊、血流信号明显、腋下淋巴结回声、微钙化和病灶结构复杂多样危险因子构建的超声误诊为乳腺癌风险预测模型预测准确性较好。

Objective To analyze the clinical characteristics of ultrasound breast cancer misdiagnosed by ultrasound and to construct a risk prediction model for misdiagnosis.Methods A total of 602 patients with breast mass who underwent ultrasound examination from January 2021 to December 2023 were selected and divided into misdiagnosis group and accurate diagnosis group according to the presence of misdiagnosis.The characteristics of patients misdiagnosed were analyzed,and a risk prediction model for misdiagnosis was constructed.Results A total of 30 patients were misdiagnosed as breast cancer,and the misdiagnosis rate was 4.98%.There were significant diferences between the accurate diagnosis group and the misdiagnosis group with respect to the largest diameter of the lesion,palpation,combined benign lesion,combined inflammatory lesion,boundary of ultrasound image,obvious blood flow signal,axillary lymph node echo,microcalcification,and complex and diverse lesion structure(P<0.01).Binary Logistic regression analysis showed that the largest diameter of the lesions(≥5 cm),combined benign lesions,combined inflammatory lesion,boundary of ultrasound image(unclear),obvious blood flow signal,axillary lymph node echo,microcalcification,and complex and diverse lesion structure were risk factors for ultrasound misdi-agnosis of breast cancer(P<0.05,P<0.01).Risk prediction model of ultrasound misdiagnosis of breast cancer was as follows:2.623×the maxi-mum diameter of lesion(≥5 cm)+1.422×benign lesion+1.616×inflammatory lesion+1.574×boundary of ultrasound image(blurred)+1.134×obvious blood flow signal+1.518×axillary lymph node echo+2.027×microcalcification+1.541×complex and diverse lesion structure.Hosmer-Lemeshow fit test showed that the model had a good fit(χ^(2)=3.487,P=0.900).The analysis of receiver operating characteristic(ROC)curve showed that the area under the ROC curve(AUC)of the prediction model for preciting ultrasound misdiagnosis of breast cancer was 0.921,the Yodon index was 0.698,and the sensitivity and specificity were 83.3%and 86.5%,respectively;the 95%CI was 0.872 and 0.969,and the actual application accuracy was 96.2%(579/602).Conclusion Ultrasound misdiagnosis of breast cancer is related to a variety of factors.The risk prediction model of ultrasound misdiagnosis of breast cancer based on the maximum diameter of lesion,benign lesion,inflammatory lesion,unclear boundary of ultrasound image,obvious blood flow signal,axillary lymph node echo,microcalcifi-cation and complex lesion structure has good prediction accuracy.