Re: effectiveness of computer-aided detection in community mammography practice.
Re: effectiveness of computer-aided detection in community mammography practice.
复制标题
回复：计算机辅助检测在社区乳房X线摄影实践中的有效性。
DOI：
10.1093/jnci/djr492
复制
发表时间：
2012
期刊：
影响因子：
--
通讯作者：
Metz,CharlesE
中科院分区：
文献类型：
--
作者：
Nishikawa,RobertM;Giger,MaryellenL;Jiang,Yulei;Metz,CharlesE
衡量计算机辅助检测 (CAD) 在乳房 X 线摄影筛查中的临床效益具有挑战性，部分原因是用于量化乳房 X 线摄影筛查有效性的指标并不总是以直接的方式应用于有 CAD 和无 CAD 筛查的比较。 Fenton 等人最近在《期刊》上发表的一篇文章对此进行了阐述。(1) 关于在社区环境中将 CAD 引入乳房 X 光检查的效果。 Fenton 等人 (1) 的主要结论之一是 CAD 增加了导管原位癌 (DCIS) 的检出率。芬顿等人。暗示这是一个不希望的结果，因为 DCIS 可能导致过度诊断。然而，他们的数据显示，在乳房 X 光检查中包含 CAD 的女性中，DCIS 检出率增加，但伴随着浸润性乳腺癌检出率的下降。这一结果与 CAD 的目标一致，CAD 的使用应该能够检测出 DCIS，而 DCIS 可能最终发展为浸润性癌症。因此，通过 CAD 增加 DCIS 检测不应被视为过度诊断，而应视为早期检测，这是筛查乳房 X 光检查的一个重要目标。也许更重要的是，通过 CAD 检测 DCIS（在 CAD 与无 CAD 比较中）与 DCIS 的乳房 X 光检查（在筛查与无筛查比较中）不同。 CAD 检测到的所有癌症均可通过乳房 X 光检查检测到；因此，如果放射科医生没有检测到 DCIS（无论是否使用 CAD），最终都会在未来的筛查检查中被检测到。因此，正如作者所暗示的那样，CAD 不会导致潜在的更多过度诊断，而是会导致患有 CAD 检测的 DCIS 的女性得到更早的治疗，即使不使用 CAD，她们也会得到治疗。缺点是患有惰性导管原位癌的女性将提前一两年接受治疗。好处是，一些接受额外治疗的女性将患有 DCIS，如果没有 CAD，这些女性可能会发展为浸润性癌症。Fenton 等人 (1) 使用公认的标准方法来计算筛查乳房 X 光检查的敏感性。当使用 CAD 时，敏感性计算的分子和分母都适当地包括仅借助 CAD 检测到的癌症。然而，当将 CAD 与无 CAD 进行比较时，该方法会产生偏差，因为在没有 CAD 的情况下，筛查人群中癌症的估计数量（任何敏感性计算中的分母）仅包括通过筛查检测到的癌症和间隔癌症，即在常规筛查之间的时间间隔内检测到的癌症。尽管其他一些癌症可能被遗漏，但在没有 CAD 的情况下，它们并未计入敏感性计算的分母中，因为无法确定它们的数量。 CAD 是否是一种有效的临床工具是一个尚未得到解答的重要问题。必须特别注意评估和解释结果所使用的方法。 Fenton 等人的文章(1) 概述了 CAD 在引入该技术时的效果。 CAD 系统性能的改进、放射科医生在使用 CAD 方面获得了更多经验以及全场数字乳腺 X 线摄影的引入极大地改变了这一现状。
Measuring the clinical benefits of computeraided detection (CAD) in mammographic screening is challenging, in part because metrics used to quantify the effectiveness of mammographic screening do not always apply in a straightforward way to comparisons of screening with and without CAD. This was illustrated in a recent article in the Journal by Fenton et al.(1) on the effect of introducing CAD to mammographic screening in the community setting. One of the principal conclusions of Fenton et al.(1) was that CAD increases detection of ductal carcinoma in situ (DCIS). Fenton et al. imply that this is an undesired result because DCIS may lead to overdiagnosis. However, their data showed that the increased rate of detection of DCIS in women whose mammograms included CAD was accompanied by a decrease in the rate of detection of invasive breast cancers. This result is consistent with the goal of CAD, the use of which should lead to the detection of DCIS that might ultimately progress to invasive cancer. Thus, the increased detection of DCIS by CAD should not be considered overdiagnosis, but rather earlier detection, which is an important goal of screening mammography. More importantly, perhaps, the detection of DCIS by CAD (in a CAD vs no CAD comparison) is not the same as mammographic detection of DCIS (in a screening vs no screening comparison). All cancers detected by CAD are mammographically detectable; hence, if the radiologists does not detect the DCIS (with or without using CAD), it will eventually be detected at a future screening exam. Therefore, CAD will not lead to potentially more overdiagnoses, as the authors imply, but rather to earlier treatment of women with CAD-detected DCIS who would be treated even if CAD were not used. The downside is that women with indolent DCIS will be treated one or two years earlier. The upside is that some of the additional treated women will have DCIS that would have progressed to invasive cancer if it were not for CAD.Fenton et al.(1) used the accepted standard method to calculate sensitivity of screening mammography. When CAD was used, both the numerator and the denominator of the sensitivity calculation appropriately included cancers that were detected only with the aid of CAD. However, that method introduces bias when CAD is compared with no CAD because without CAD, the estimated number of cancers in the screened population (the denominator in any calculation of sensitivity) included only cancers detected by screening and interval cancers, those cancers that were detected in the interval between routine screening. Although some other cancers are likely to have been missed, they were not counted in the denominator of the sensitivity calculation without CAD because there was no way of determining their number. Whether CAD is an effective clinical tool is an important question that remains unanswered. Careful attention must be given to the methodology used in the evaluation and interpretation of results. The article by Fenton et al.(1) presents a snapshot of the effect of CAD at the introduction of the technology. Improvements in the performance of CAD systems, radiologists who have gained more experience in using CAD, and the introduction of fullfield digital mammography have greatly changed the landscape.