文献速递乳腺乳腺X线

计算机辅助检测(CAD)增强二维合成乳腺X线摄影的评估:与标准合成二维乳腺X线摄影及常规二维数字乳腺X线摄影的比较

Evaluation of a computer-aided detection (CAD)-enhanced 2D synthetic mammogram: comparison with standard synthetic 2D mammograms and conventional 2D digital mammography.
PubMed 文献 · Clin Radiol · 2018-06-30

在68例因软组织异常被召回的乳腺筛查病例中,比较CAD增强合成乳腺X线摄影、标准合成乳腺X线摄影与全视野数字乳腺X线摄影(FFDM)的诊断效能。CAD增强合成乳腺X线摄影的AUC为0.846,显著高于标准合成乳腺X线摄影的0.683(p=0.004)和FFDM的0.724(p=0.027);对全部68例,每幅图像平均增强3.2个区域,34例癌症中97.4%的病灶被正确增强,每幅图像平均2.1个假增强区域。结论认为CAD增强显著改善合成二维乳腺X线摄影的诊断性能,且诊断准确率优于常规二维FFDM。

要点

  • CAD增强合成乳腺X线摄影AUC 0.846,高于标准合成0.683(p=0.004)和FFDM 0.724(p=0.027)。
  • 34例癌症中97.4%病灶被正确增强,每幅图像平均2.1个假增强区域。
  • 全部68例每幅图像平均增强3.2个区域。

核心结论

  • CAD增强显著改善合成二维乳腺X线摄影的诊断性能,且诊断准确率优于常规二维FFDM。

研究设计与样本

  • 回顾性研究,纳入68例因软组织异常(肿块、实质变形、不对称致密)被召回的乳腺筛查病例。
  • 每例均有CAD增强合成乳腺X线摄影、标准合成乳腺X线摄影和FFDM三种图像。
  • 两名放射科医生在不知图像类型和最终评估结果的情况下,对每种乳腺X线摄影的斜位和头尾位投影进行阅片,共204对二维图像随机顺序呈现。
  • 采用五点量表评分(1=正常至5=恶性),阅片时不查看数字乳腺断层合成(DBT)断层图像。
  • 采用受试者工作特征(ROC)曲线分析。

关键影像表现/征象

  • CAD增强合成乳腺X线摄影对所有软组织异常均具有最高诊断准确率。
  • 对恶性病灶,敏感度不受肿瘤大小影响。

诊断标准与关键数值

  • CAD增强合成乳腺X线摄影AUC=0.846,标准合成乳腺X线摄影AUC=0.683,两者比较p=0.004。
  • CAD增强合成乳腺X线摄影AUC=0.846,常规二维FFDM AUC=0.724,两者比较p=0.027。
  • 全部68例中,每幅图像平均增强3.2个区域。
  • 34例癌症病例中,97.4%的病灶被正确增强,每幅图像平均2.1个假增强区域。

临床应用与阅片要点

  • 阅片时可参考CAD增强合成乳腺X线摄影对软组织异常的高诊断准确率。
  • 需注意每幅图像平均2.1个假增强区域,结合形态学评估判读。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

英文原文(摘要)· 点击展开

AIM: To evaluate the diagnostic performance of computer-aided detection (CAD)-enhanced synthetic mammograms in comparison with standard synthetic mammograms and full-field digital mammography (FFDM).

MATERIALS AND METHODS: A CAD-enhanced synthetic mammogram, a standard synthetic mammogram, and FFDM were available in 68 breast-screening cases recalled for soft-tissue abnormalities (masses, parenchymal deformities, and asymmetric densities). Two radiologists, blinded to image type and final assessment outcome, retrospectively read oblique and craniocaudal projections for each type of mammogram. The resulting 204 pairs of 2D images were presented in random order and scored on a five-point scale (1, normal to 5, malignant) without access to the Digital breast tomosynthesis (DBT) slices. Receiver operating characteristic (ROC) curve analysis was performed.

RESULTS: There were 34 biopsy-proven malignancies and 34 normal/benign cases. Diagnostic accuracy was significantly improved for the CAD-enhanced synthetic mammogram compared to the standard synthetic mammogram (area under the ROC curve [AUC]=0.846 and AUC=0.683 respectively, p=0.004) and compared to the conventional 2D FFDM (AUC=0.724, p=0.027). The CAD-enhanced synthetic mammogram had the highest diagnostic accuracy for all soft-tissue abnormalities, and for malignant lesions sensitivity was not affected by tumour size. For all 68 cases, there was an average of 3.2 areas enhanced per image. For the 34 cancer cases, 97.4% of lesions were correctly enhanced, with 2.1 false areas enhanced per image.

CONCLUSIONS: CAD enhancement significantly improves performance of synthetic 2D mammograms and also exhibits improved diagnostic accuracy compared to conventional 2D FFDM.

来源:Clin Radiol,查看原文。版权归原作者所有。

引用

  1. James JJ, Giannotti E, Chen Y, et al. Evaluation of a computer-aided detection (CAD)-enhanced 2D synthetic mammogram: comparison with standard synthetic 2D mammograms and conventional 2D digital mammography. Clin Radiol. 2018. PMID: 29970247. doi:10.1016/j.crad.2018.05.028

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。