基于多模态超声及多因素分析结果构建的乳腺癌风险回归方程

Construction of a breast cancer risk regression equation based on multimodal ultrasound and multivariable analysis

摘要:

目的 探讨基于多模态超声及多因素分析结果构建乳腺癌的风险回归方程。方法 回顾性选取2023年8月—2024年12月浙江省人民医院毕节医院就诊的150例乳腺结节病变患者的临床资料。根据病变性质不同分为观察组（乳腺癌，52例）和对照组（乳腺良性病变，98例），患者均接受多模态超声检查。采用Lasso-logistic多因素模型分析乳腺癌变相关因素，构建风险回归方程，分析其判断乳腺癌变的预测价值。结果 观察组肿块边缘模糊占比、回声不均匀混合占比、毛刺征率、E-mean、E-max、CEUS结果恶性率、SMI血流分级Ⅱ级占比、CA153水平及CEA水平均高于对照组（P <0.05）。多因素一般Logistic回归分析，结果显示：有毛刺征［O^R=3.682（95% CI：1.274，10.641）］、E-mean高［O^R=4.128（95% CI：2.617，6.511）］、CEUS结果恶性［O^R=5.831（95% CI：3.645，9.328）］、SMI血流分级Ⅲ级［O^R=3.225（95% CI：1.239，8.394）］和CA153水平高［O^R=7.819（95% CI：1.073，56.977）］均为乳腺癌变的危险因素（P <0.05）。基于多因素一般Logistic回归分析结果，构建乳腺癌术后风险回归模型，方程为：Logit（P）＝2.112＋毛刺征×1.303＋E-mean×1.418＋CEUS结果×1.763＋SMI血流分级×1.171＋2.057×CA153。受试者工作特征曲线分析结果显示：该风险回归方程判断乳腺病变性质的曲线下面积为0.901，敏感性为82.7%（95% CI：0.753，0.914），特异性为76.5%（95% CI：0.683，0.847）。结论 乳腺癌与肿块是否有毛刺征、E-mean、CEUS结果、SMI血流分级及CA153水平相关，据此构建的风险回归方程有助于乳腺癌的筛查。

Abstract:

Objective To construct a breast cancer risk regression equation based on multimodal ultrasound and multivariable analysis.Methods The clinical data of 150 patients with breast nodules treated at Bijie Hospital of Zhejiang Provincial People's Hospital between August 2023 and December 2024 were retrospectively analyzed. According to lesion type, the patients were divided into an observation group (breast cancer, n = 52) and a control group (benign breast lesions, n = 98). All patients underwent multimodal ultrasonography. A Lasso-logistic multivariable model was used to analyze factors associated with breast cancer, construct a risk regression equation, and evaluate its predictive value for differentiating malignant from benign breast lesions.Results The observation group had higher proportions of indistinct mass margins, heterogeneous mixed echogenicity, spiculation, malignant contrast-enhanced ultrasound (CEUS) findings, and grade II superb microvascular imaging (SMI) blood flow, as well as higher E-mean, E-max, carbohydrate antigen 15-3 (CA15-3), and carcinoembryonic antigen levels, than the control group (P < 0.05). Multivariable logistic regression showed that spiculation [O^R = 3.682 (95% CI: 1.274, 10.641) ], higher E-mean [O^R = 4.128 (95% CI: 2.617, 6.511) ], malignant CEUS findings [O^R = 5.831 (95% CI: 3.645, 9.328) ], grade III SMI blood flow [O^R = 3.225 (95% CI: 1.239, 8.394) ], and higher CA15-3 [O^R = 7.819 (95% CI: 1.073, 56.977) ] were risk factors for breast cancer (P < 0.05). The resulting equation was Logit(P) = 2.112 + 1.303 × spiculation + 1.418 × E-mean + 1.763 × CEUS findings + 1.171 × SMI blood flow grade + 2.057 × CA15-3. The equation had an area under the receiver operating characteristic curve of 0.901, a sensitivity of 82.7% (95% CI: 0.753, 0.914), and a specificity of 76.5% (95% CI: 0.683, 0.847) for determining the nature of breast lesions.Conclusion Breast cancer is associated with spiculation, E-mean, CEUS findings, SMI blood flow grade, and CA15-3 level. A risk regression equation based on these factors may facilitate breast cancer screening.