Department of Imaging Physics, University of Texas M. D. Anderson Cancer Center, Houston, TX 77030, USA.
Radiology. 2010 Oct;257(1):40-6. doi: 10.1148/radiol.10092170. Epub 2010 Aug 2.
To prospectively determine the interpretation time associated with computer-aided detection (CAD) and to analyze how CAD affected radiologists' decisions and their level of confidence in their interpretations of digital screening mammograms.
An Institutional Review Board exemption was obtained, and patient consent was waived in this HIPAA compliant study. The participating radiologists gave informed consent. Five radiologists were prospectively studied as they interpreted 267 clinical digital screening mammograms. Interpretation times, recall decisions, and confidence levels were recorded without CAD and then with CAD. Software was used for linear regression fitting of interpretation times. P values less than .05 were considered to indicate statistically significant differences.
Mean interpretation time without CAD was 118 seconds ± 4.2 (standard error of the mean). Mean time for reviewing CAD images was 23 seconds ± 1.5. CAD identified additional findings in five cases, increased confidence in 38 cases, and decreased confidence in 21 cases. Interpretation time without CAD increased with the number of mammographic views (P < .0001). Mean times for interpretation without CAD and review of the CAD images both increased with the number of CAD marks (P < .0001). The interpreting radiologist was a significant variable for all interpretation times (P < .0001). Interpretation time with CAD increased by 3.2 seconds (95% confidence interval: 1.8, 4.6) for each calcification cluster marked and by 7.3 seconds (95% confidence interval: 4.7, 9.9) for each mass marked.
The additional time required to review CAD images represented a 19% increase in the mean interpretation time without CAD. CAD requires a considerable time investment for digital screening mammography but may provide less measureable benefits in terms of confidence of the radiologists.
前瞻性地确定计算机辅助检测（CAD）的解释时间，并分析 CAD 如何影响放射科医生的决策及其对数字筛查乳房 X 线照片解释的信心水平。
本研究获得了机构审查委员会豁免，并在符合 HIPAA 规定的情况下放弃了患者同意。参与的放射科医生获得了知情同意。在没有 CAD 和使用 CAD 的情况下，前瞻性地研究了 5 名放射科医生对 267 例临床数字筛查乳房 X 线照片的解释。记录了解释时间、召回决策和信心水平，没有 CAD 和使用 CAD。使用软件对解释时间进行线性回归拟合。P 值小于 0.05 被认为表示具有统计学意义的差异。
没有 CAD 的平均解释时间为 118 秒±4.2（均数的标准误差）。审查 CAD 图像的平均时间为 23 秒±1.5。CAD 在 5 例中发现了额外的发现，在 38 例中增加了信心，在 21 例中降低了信心。没有 CAD 的解释时间随乳腺 X 线照片视图数量的增加而增加（P<0.0001）。没有 CAD 的解释时间和 CAD 图像的审查时间都随 CAD 标记数量的增加而增加（P<0.0001）。解释时间是所有解释时间的显著变量（P<0.0001）。每个钙化簇标记增加 3.2 秒（95%置信区间：1.8，4.6），每个肿块标记增加 7.3 秒（95%置信区间：4.7，9.9），这会导致使用 CAD 时的解释时间增加。
审查 CAD 图像所需的额外时间使没有 CAD 的平均解释时间增加了 19%。CAD 为数字筛查乳房 X 线摄影需要大量的时间投入，但在放射科医生的信心方面可能提供的可衡量的益处较少。
