The potential contribution of a computer-aided detection system for lung nodule detection in multidetector row computed tomography
The potential contribution of a computer-aided detection system for lung nodule detection in multidetector row computed tomography
复制标题
DOI：
10.1097/00004424-200411000-00001
复制
发表时间：
2004-11-01
影响因子：
6.7
通讯作者：
Kim, YT
中科院分区：
文献类型：
--
作者：
Lee, JW;Goo, JM;Kim, YT
目的：我们试图评估的潜在好处的计算机辅助检测（CAD）系统检测肺结节在多排CT（MDCT）scannes.Methods：一个CAD系统被开发用于检测肺结节的MDCT扫描，并应用于从15例患者获得的数据。两名胸部放射科医生一致建立了参考标准。根据结核的大小及其与周围结构的关系（结核类型）对结核进行分类。使用卡方（2）分析评价了经验丰富的胸部放射科医生和无用户交互的CAD系统之间的灵敏度差异。结果：共309个结节被确定为参考标准。CAD系统的敏感性（81%）与放射科医师的敏感性（85%）无显著差异（P > 0.05）。CAD系统对≤ 5 mm结节及孤立结节的检出率均高于放射科医师（83% vs.75%，P > 0.05; 93% vs.76%，P < 0.001）。放射科医师对> 5 mm结节和与其他结构相连的结节的检出灵敏度高于CAD系统（98% vs. 79%，P < 0.001; 91% vs. 71%，P < 0.001）。每次CT检查有28.8个CAD假阳性结果。结论：本研究开发的CAD系统在结节大小和结节类型方面以与放射科医生不同的方式执行结节检测任务，这表明CAD系统可以发挥补充作用放射科医生从大CT数据集中检测结节。
Objectives: We sought to evaluate the potential benefits of a computer-aided detection (CAD) system for detecting lung nodules in multidetector row CT (MDCT) scans.Methods: A CAD system was developed for detecting lung nodules on MDCT scans and was applied to the data obtained from 15 patients. Two chest radiologists in consensus established the reference standard. The nodules were categorized according to their size and their relationship to the surrounding structures (nodule type). The differences in the sensitivities between an experienced chest radiologist and a CAD system without user interaction were evaluated using a chi(2) analysis. The differences in the sensitivities also were compared in terms of the nodule size and the nodule type.Results: A total of 309 nodules were identified as the reference standard. The sensitivity of a CAD system (81%) was not significantly different from that of a radiologist (85%; P > 0.05). The sensitivities of the CAD system for detecting nodules less than or equal to 5 mm in diameter as well as detecting isolated nodules were higher than those of a radiologist (83% vs. 75%, P > 0.05; 93% vs. 76%, P < 0.001). The sensitivities of a radiologist for detecting nodules > 5 mm and the nodules attached to other structures were higher than those of a CAD system (98% vs. 79%, P < 0.001; 91% vs. 71%, P < 0.001). There were 28.8 false-positive results of CAD per CT study.Conclusion: The CAD system developed in this study performed the nodule detection task in different ways to that of a radiologist in terms of the nodule size and the nodule type, which suggests that the CAD system can play a complementary role to a radiologist in detecting nodules from large CT data sets.