| 李葚煦,吴静云,孔迅,陈路增.超声智能诊断系统联合钼靶X线检查鉴别乳腺良、恶性非肿块型病变[J].中国介入影像与治疗学,2025,22(5):336-340 |
| 超声智能诊断系统联合钼靶X线检查鉴别乳腺良、恶性非肿块型病变 |
| Ultrasound intelligent diagnostic system combined with mammography for differentiating benign and malignant non-mass breast lesions |
| 投稿时间：2024-11-11 修订日期：2025-05-06 |
| DOI：10.13929/j.issn.1672-8475.2025.05.008 |
| 中文关键词: 乳腺疾病 超声检查 人工智能 乳腺X线摄影术 |
| 英文关键词:breast diseases ultrasonography artificial intelligence mammography |
| 基金项目: |
|
| 摘要点击次数: 56 |
| 全文下载次数: 33 |
| 中文摘要: |
| 目的 观察超声智能诊断系统联合钼靶X线检查鉴别乳腺良、恶性非肿块型病变（NMBL）的价值。方法 回顾性纳入107例NMBL，包括64个恶性（恶性组）、43个良性（良性组）病变。比较组间临床、常规超声、超声智能诊断系统[人工智能（AI）系统]及钼靶X线检查资料，行logistic回归分析，绘制受试者工作特征 （ROC）曲线，计算曲线下面积（AUC），评估AI系统联合钼靶X线检查鉴别良、恶性NMBL的效能。结果 组间病灶最大径、腋窝肿大淋巴结率、钼靶X线片显示可疑恶性钙化率、AI系统恶性风险及AI系统乳腺影像报告和数据系统（BI-RADS）分类差异均有统计学意义（P均＜0.05）。基于AI系统恶性风险获得AI系统二分类。单一钼靶X线片显示可疑恶性钙化、AI系统BI-RADS分类及AI系统二分类鉴别良、恶性NMBL的AUC分别为0.840、0.810及0.817；钼靶X线片显示可疑恶性钙化分别与AI系统BI-RADAS分类及AI系统二分类联合鉴别良、恶性NMBL的AUC均为0.856，与单一AI系统BI-RADAS分类/AI系统二分类差异均有统计学意义（P均＜0.05）而与单一钼靶X线片显示可疑恶性钙化差异无统计学意义（P均＞0.05）。针对基于年龄、病灶最大径、腋窝肿大淋巴结及钼靶X线片显示可疑恶性钙化联合AI系统恶性风险构建的模型1、联合AI系统BI-RADS分类构建的模型2及联合AI系统二分类构建的模型3的logistic回归分析结果显示，钼靶X线片显示可疑恶性钙化、AI系统恶性风险、AI系统BI-RADS分类及AI系统二分类均为恶性NMBL的独立危险因素（P均＜0.05），上述3个模型鉴别诊断良、恶性NMBL的AUC依次为0.966、0.964及0.957。结论 超声智能诊断系统联合钼靶X线检查有助于鉴别良、恶性NMBL；联合临床指标或有助于提高其诊断效能。 |
| 英文摘要: |
| Objective To explore the value of ultrasound intelligent diagnostic system combined with mammography for differentiating benign and malignant non-mass breast lesions （NMBL）. Methods Totally 107 patients with NMBL were retrospectively enrolled， including 64 cases of malignant （malignant group） and 43 cases of benign lesions （benign group）. Clinical， routine ultrasound， ultrasound intelligent diagnostic system （artificial intelligence [AI] system） and mammography data were compared between groups. Logistic regression analysis was performed， receiver operating characteristic （ROC） curves were drawn， the areas under the curves （AUC） were calculated， and the efficacy of AI system combined with mammography for differentiating benign and malignant NMBL was analyzed. Results Significant differences of the maximum diameter of lesion， ratio of axillary lymph node enlargement and suspected malignant calcification on mammography， as well as of AI system malignancy risk and AI system breast imaging reporting and data system （BI-RADS） classification were found between groups （all P＜0.05）. AI system binary classification was obtained based on AI system malignancy risk. The AUC of suspected malignant calcification on mammography， AI system BI-RADS classification and AI system binary classification for differential diagnosis of benign and malignant NMBL was 0.840， 0.810 and 0.817， respectively， while of suspected malignant calcification on mammography combined with AI system BI-RADS classification or AI system binary classification were both 0.856， higher than that of AI system BI-RADS classification/AI system binary classification alone （both P＜0.05） but not significantly different with that of suspected malignant calcification on mammography alone （both P＞0.05）. Logistic regression analysis of age， the maximum diameter of lesion， axillary lymph node enlargement and suspected malignant calcification on mammography combined with AI system malignancy risk （model 1）， AI system BI-RADS classification （model 2） or AI system binary classification （model 3） showed that suspected malignant calcification on mammography， AI system malignancy risk， AI system BI-RADS classification and AI system binary classification were all independent risk factors of malignant NMBL （all P＜0.05）， and AUC of model 1， 2 and 3 for differentiating benign and malignant NMBL was 0.966， 0.964 and 0.957， respectively. Conclusion Ultrasound intelligent diagnostic system combined with mammography was helpful for differentiating benign and malignant NMBL. Combining with clinical indicators might improve diagnostic efficacy. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
