Brem Rachel F, Baum Janet, Lechner Mary, Kaplan Stuart, Souders Stuart, Naul L Gill, Hoffmeister Jeff
Department of Breast Imaging and Intervention, George Washington University Medical Center, 2150 Pennsylvania Ave. N.W., Washington, DC 20037, USA.
AJR Am J Roentgenol. 2003 Sep;181(3):687-93. doi: 10.2214/ajr.181.3.1810687.
Our study evaluated radiologist detection of breast cancer using a computer-aided detection system.
Three radiologists reviewed 377 screening mammograms interpreted as showing normal or benign findings 9-24 months before cancer diagnosis from 17 of the 18 participating centers. In 313 cases, study radiologists recommended additional mammographic evaluation. In 177 cases, the area warranting additional workup precisely correlated with the subsequently diagnosed cancer. These 177 missed cancers were evaluated with computer-aided detection. The proportion of radiologists identifying the missed cancers was used to determine radiologist sensitivity without computer-aided detection.
The study radiologists determined that 123 of the 377 missed cancer cases warranted workup. Therefore, 123 additional cancers cases could have been found. The calculated radiologist sensitivity without computer-aided detection was therefore 75.4% (377 / [377 + 123]). Similarly, using the performance of the system on the missed cancers, we estimated that 80 (65.0%) of these 123 missed cancer cases would have been identified with the use of computer-aided detection. Consequently, the estimated sensitivity of radiologists using computer-aided detection was 91.4% ([377 + 80] / [377 + 123])-resulting in a 21.2% ([91.4% / 75.4%] - 1) increase in radiologist sensitivity with computer-aided detection.
Use of the computer-aided detection system significantly improved the detection of breast cancer by increasing radiologist sensitivity by 21.2%. Therefore, for every 100,000 women with breast cancer identified without the use of computer-aided detection, an estimated additional 21,200 cancers would be found with the use of computer-aided detection.
我们的研究评估了放射科医生使用计算机辅助检测系统检测乳腺癌的情况。
三位放射科医生对来自18个参与中心中的17个中心的377例筛查乳腺X线照片进行了复查，这些照片在癌症诊断前9 - 24个月被解读为显示正常或良性结果。在313例病例中，参与研究的放射科医生建议进行额外的乳腺X线评估。在177例病例中，需要进一步检查的区域与随后诊断出的癌症精确相关。对这177例漏诊的癌症进行了计算机辅助检测评估。通过放射科医生识别出漏诊癌症的比例来确定无计算机辅助检测时放射科医生的敏感度。
参与研究的放射科医生确定，377例漏诊癌症病例中有123例需要进一步检查。因此，原本可以多发现123例癌症病例。由此计算出无计算机辅助检测时放射科医生的敏感度为75.4%（377 / [377 + 123]）。同样，根据该系统对漏诊癌症的检测表现，我们估计在这123例漏诊癌症病例中，使用计算机辅助检测可以识别出80例（65.0%）。因此，使用计算机辅助检测时放射科医生的估计敏感度为91.4%（[377 + 80] / [377 + 123]），这使得使用计算机辅助检测时放射科医生的敏感度提高了21.2%（[91.4% / 75.4%] - 1）。
使用计算机辅助检测系统通过将放射科医生的敏感度提高21.2%，显著改善了乳腺癌的检测。因此，对于每100,000例未使用计算机辅助检测而确诊的乳腺癌女性患者，使用计算机辅助检测估计可多发现约21,200例癌症。
