| 摘要: |
| 目的 探讨乳腺癌改良根治术后胸壁复发结节的超声图像特征，并建立预测模型。
方法 回顾性分析我院102例乳腺癌改良根治术后患者胸壁新发结节的超声图像特征，其中复发结节57例，良性结节45例，通过回归分析得出有鉴别意义的特征性指标，建立Logistic回归模型。
结果 乳腺癌改良根治术后胸壁复发结节超声图像常表现为单发或多发、累及多层组织结构、边界不清、形态不规则、少-中量血流信号的低回声实性结节。多因素回归分析显示胸壁病灶最大径、数目、边界、血流信号、术后间隔时间是乳腺癌术后胸壁结节恶性风险的独立预测因素（均P<0.05）。建立的Logistic回归模型判断术后胸壁结节为复发病灶的风险灵敏度为91.2%、特异度为82.2%、准确度为87.3%、曲线下面积为0.935。
结论 利用超声图像特征建立的风险预测模型能有效鉴别乳腺癌改良根治术后胸壁结节的良恶性，提高超声诊断准确率。 |
| 关键词: 超声检查 乳腺癌 胸壁复发 预测模型 |
| DOI： |
| 投稿时间：2022-05-20修订日期：2022-06-29 |
| 基金项目:佛山市“十四五”医学高水平重点专科建设项目 |
|
| Establishment a prediction model of chest wall recurrent nodules after modified radical mastectomy using ultrasound characteristics |
| He Huizhen,Huang Weijun,Ye Jieyi,He Yanping |
| (The First People’S Hospital of Foshan) |
| Abstract: |
| Objective To explore the ultrasound image features of chest wall recurrent nodules after modified radical mastectomy and to establish a prediction model. Methods Ultrasound image features of chest wall nodules of 102 patients(57 patients with recurrent nodules and 45 patients with benign nodules) after mastectomy were retrospectively analyzed. The ultrasound features were screened using regression analysis, and then the screened features were used to build a regression model. Results Chest wall recurrent nodules after mastectomy often show single or multiple hypoechoic solid masses with multilayer tissue structure implication, poorly defined borders, irregular morphology, and low to medium blood flow signals. Multifactorial analysis showed that maximum diameter, number, borders, blood flow signal and postoperative intervals were independent predictors of the risk of malignant nodules（P<0.05）. The logistic regression model showed that the risk sensitivity, specificity and accuracy was 91.2%, 82.2% and 87.3%, with the area under curve 0.935. Conclusions Prediction model based on ultrasound features can effectively evaluate the malignant risk of chest wall nodules after modified radical mastectomy. |
| Key words: Ultrasonography Breast cancer Chest wall recurrence Prediction model |
