基于超声特征决策树模型对乳腺结节良恶性肿瘤的鉴别诊断价值研究  

Study on the Value of Decision Tree Model Based on Ultrasonic Characteristics in Differential Diagnosis of Benign and Malignant Breast Nodules

作　　者：刘诚 林陶玉 侯吉学[1] Liu Cheng;Lin Taoyu;Hou Jixue(The First Affiliated Hospital of Shihezi University,Shihezi,Xinjiang 832008,China;The People's Hospital of Suzhou New District,Suzhou,Jiangsu 215129,China)

机构地区：[1]石河子大学第一附属医院,新疆石河子市832008 [2]苏州高新区人民医院,江苏省苏州市215129

出　　处：《中国超声医学杂志》2025年第3期270-274,共5页Chinese Journal of Ultrasound in Medicine

基　　金：国家自然科学基金(No.82260105)。

摘　　要：目的 基于常规超声(CUS)联合超声造影(CEUS)特征构建决策树模型,探讨其鉴别乳腺结节良恶性肿瘤的诊断效能和应用价值。方法 采用病例对照研究方法,收集66例乳腺结节恶性肿瘤患者和65例乳腺结节良性肿瘤患者的CUS和CEUS资料,采用χ~2检验、Fisher确切概率法、Mann-Whitney U检验、Lasso回归筛选变量,利用CHAID方法构建决策树模型,运用受试者工作特征(ROC)曲线、决策曲线(DCA)和临床影响曲线(CIC)对模型进行内部验证及临床获益分析。结果 决策树模型显示,对乳腺结节良恶性肿瘤产生重要影响的变量依次为增强范围(χ^(2)=74.895,P<0.001)、血流(χ^(2)=15.402,P<0.001)、放射状汇聚增强范围(χ^(2)=5.624,P<0.05)。决策树模型诊断乳腺恶性肿瘤的ROC曲线下面积为0.942(95%CI:0.904~0.979,P<0.001)。DCA和CIC结果显示,阈概率在4%~95%时,决策树模型预测乳腺恶性肿瘤有较高的净获益;风险阈值达到23%后,预测乳腺恶性肿瘤的人数与实际人数趋于一致。结论 基于CUS联合CEUS超声特征构建的决策树模型可以作为鉴别诊断乳腺结节良恶性肿瘤的可靠工具。Objective To construct a decision tree model based on ultrasound features of conventional ultrasound(CUS)and contrast-enhanced ultrasound(CEUS),and to explore its diagnostic efficacy and clinical value of distin-guishing benign from malignant breast lesions.Methods A case-control study was performed to collect the data of CUS and CEUS in 66 patients with malignant breast nodules and 65 patients with benign breast nodules.Variables were screened by X?test,Fisher's exact probability method,Mann-Whitney U test,and Lasso regression.A decision tree model was constructed by the CHAID method.Receiver operating characteristic(ROC)curve,decision curve analysis(DCA),and clinical impact curve(CIC)were used for the model to verify the internal verification and analyze the clinical benefit.Results The decision tree model showed that the important variables influencing the benign and malignant breast lesions were enhancement range(x^(2)=74.895,P<0.001),blood flow(x^(2)=15.402,P<0.001)and focal enhancement(x^(2)=5.624,P<0.05).The area under the ROC curve(AUC)of the decision tree model for diag-nosing breast malignancy was 0.942(95%CI:0.904-0.979,P<0.001).The DCA and CIC results showed that when the threshold probability was between 4% and 95%,the decision tree model had a high net benefit;when the risk threshold reached 23%,the number of breast malignancies predicted by the model tended to be consistent with the actual number.Conclusions The decision tree model based on ultrasound features of CUS and CEUS can be used as a reliable tool for differentiating benign and malignant breast tumors.

关 键 词：常规超声 超声造影 乳腺结节 肿瘤 决策树 

分 类 号：R445.1[医药卫生—影像医学与核医学] R737.9[医药卫生—诊断学]