首页|乳腺AI、超声联合MRI在乳腺结节BI-RADS分类中的效能

乳腺AI、超声联合MRI在乳腺结节BI-RADS分类中的效能

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
目的 分析乳腺人工智能(AI)与常规超声(US)联合磁共振成像(MRI)在乳腺结节乳腺影像报告和数据系统(BI-RADS)分类中的效能.方法 选取105例乳腺结节患者(140个乳腺结节),所有乳腺结节均行乳腺AI、US及MRI检查;采用BI-RADS对患者的乳腺结节进行良性和恶性的分类,以术后病理结果为标准,BI-RADS 4a及以下为良性组、BI-RADS 4b及以上为恶性组,对比3种检查方法对乳腺癌诊断的准确性、敏感性及特异性,计算受试者工作特征曲线下面积(AUC);根据US分类将55个US-BI-RADS 4a类结节作为研究对象,用AI-BI-RADS 4a作为截点值,采用AI-BI-RADS对该55个乳腺结节进行重新分类,分析调整后的AI-US-BI-RADS与US-BI-RADS、病理结果、MRI-BI-RADS诊断结果间的一致性和差异,比较AI、US和MRI对BI-RADS 4a类结节的诊断效能.结果 3种方法独立应用判断乳腺结节的良恶性时,MRI的准确性和敏感性最高,AI的特异性最高;在联合诊断乳腺结节的良恶性时,US+AI+MRI联合诊断乳腺癌的准确性、敏感性及特异性最高且AUC最大;以AI-BI-RADS 4a类作为截点,对55例US-BI-RADS 4a类乳腺结节进行重新分类,AI与US、病理结果及MRI检测结果均有较高的一致性(P＜0.01);AI调整后的乳腺BI-RADS分类准确性、敏感性、特异性均较调整分类前提高.结论 乳腺AI、US及MRI的联合使用可提高乳腺癌的诊断效能,AI-US-BI-RADS分类可较好的预测乳腺结节的良恶性.
Efficacy of AI in breast,ultrasound combined with MRI in BI-RADS classification for breast nodules
Objective To analyze the efficacy of artificial intelligence(AI)in breast and conventional ultrasound(US)combined with magnetic resonance imaging(MRI)in BI-RADS classification for breast nodules.Methods A total of 105 patients with breast nodules(140 breast nodules)were selected.All breast nodules were examined by Al,US and MRI.Breast nodules were classified into benign and malignant according to breast images and data system.Based on postoperative pathological results,BI-RADS 4a and below were classified as benign group,and BI-RADS 4b and above were classified as malignant group.The accuracy,sensitivity and specificity of the three examination methods for breast cancer diagnosis were compared.Area under the receiver operating characteristic curve(AUC)was calculated.According to US classification,55 US-BI-RADS 4a nodules were taken as research objects.AI-BI-RADS 4a was used as cut-off value to reclassify these 55 breast nodules using AI-BI-RADS.The consistency and differences were analyzed between the adjusted AI-US-BI-RADS and US-BI-RADS,pathological results and MRI-BI-RADS diagnostic results.Diagnostic efficacy was compared among AI,US,and MRI for BI-RADS 4a nodules.Results When three methods were independently applied to determine the benign or malignant nature of breast nodules,MRI had the highest accuracy and sensitivity,while AI had the highest specificity.In the joint diagnosis of benign and malignant breast nodules,US+AI+MRI had the highest accuracy,sensitivity specificity and AUC in joint diagnosing breast cancer.When AI-BI-RADS 4a was used as a cut-off value to reclassify 55 US-BI-RADS 4a breast nodules,there was high consistency between AI and US,pathological results and MRI detection results.The accuracy,sensitivity and specificity of breast BI-RADS classification after AI adjustment were improved when compared to before adjustment.Conclusion The combined application of AI,US and MRI can improve the diagnostic efficiency of breast cancer,and AI-US-BI-RADS classification can provide better prediction for benign and malignant nature of breast nodules.

ultrasoundartificial intelligenceconventional ultrasound combined with magnetic resonance imagingBI-RADS classificationcombined diagnosisefficiency

罗富欢、施丽英、谢瑾、赵丽娜、张蓓、陈霞、胡小丽

展开 >

贵州医科大学附属医院超声中心,贵州贵阳 550001

超声 人工智能 常规超声联合磁共振成像 BI-RADS分类 联合诊断 效能

贵州省研究生科研基金贵州省教育厅青年科技人才成长项目贵州省科技厅基金

黔教合YJSKYJJ[2020]144黔教合KY字[2022]242黔科合基础-ZK[2022]一般403

2024

10.19367/j.cnki.2096-8388.2024.06.013

贵州医科大学学报
贵阳医学院

贵州医科大学学报

CSTPCD
影响因子：0.827
ISSN：2096-8388
年,卷(期)：2024.49(6)
  • 4