论文部分内容阅读
目的评价计算机辅助检测(computer-aided detection,CAD)系统在数字化胸片肺结节检出中的应用价值及其对放射科医师的辅助作用。方法选取数字化胸片328例。由2名专家组医师应用 IQQAT~(TM)-Chest 系统阅读具有结节样阴影的胸片,2人意见达成一致后标记结节的位置和大小并保存标记结果,将标记结果作为金标准来评估 CAD 系统检测肺结节的能力。由8名不同年资的放射科医师首先独立阅读具有结节样阴影的数字 X 线摄影(DR)胸片并保存诊断结果,然后再应用CAD 系统阅读胸片,将最终结果存入 CAD 系统。应用受试者操作特征曲线(ROC)和配对 t 检验来分析放射科医师应用 CAD 系统前后在肺结节检测能力上的差异。结果在100例 DR 胸片中,金标准结节151个,CAD 系统肺结节检测敏感性为78.1%(118.0/151),低年资放射科医师不用和应用 CAD系统时,肺结节的检测敏感性分别为62.4%(94.2/151)和77.4%(116.8/151),ROC 曲线下面积分别为0.769 和0.836,二者之间的差异具有统计学意义(P<0.01);高年资放射科医师不用和应用CAD 系统时,肺结节的检测敏感性分别为73.8%(111.5/151)和76.2%(115.0/151),ROC 曲线下面积分别为0.820和0.827,二者之间的差异无统计学意义(P>0.05)。结论 CAD 系统能够辅助放射科医师提高肺小结节的检测敏感性,对低年资医师的帮助更大。
Objective To evaluate the value of computer-aided detection (CAD) system in the detection of digital chest pulmonary nodules and its aid to radiologists. Methods 328 cases of digital chest radiography. Two expert panelists used the IQQAT ~ (TM) -Chest system to read chest radiographs with nodular shadows. After two people agreed on the location and size of the nodules, the results of the labeling were saved and the results of the labeling were used as gold standard Assess CAD system’s ability to detect lung nodules. Eight radiosurgeons of different age groups first independently read digital radiography (DR) chest radiographs with nodular shadows and save the diagnostic results. The chest radiographs were then read using a CAD system and the final results were stored in a CAD system. The receiver operating characteristic curve (ROC) and paired t-test were used to analyze the difference in lung nodule detection before and after radiologist applied CAD system. Results In 100 DR chest radiographs, there were 151 gold standard nodules and 78.1% (118.0 / 151) sensitivity of CAD system in detecting pulmonary nodules. The low-grade radiologists did not need and applied CAD system, The sensitivity of detection was 62.4% (94.2 / 151) and 77.4% (116.8 / 151), respectively. The areas under the ROC curve were 0.769 and 0.836, respectively. The differences between the two methods were statistically significant (P <0.01) The radiosensitivity of pulmonary nodules was 73.8% (111.5 / 151) and 76.2% (115.0 / 151), respectively. The areas under the ROC curve were 0.820 and 0.827 The difference was not statistically significant (P> 0.05). Conclusion The CAD system can assist radiologists to improve the detection sensitivity of pulmonary nodules and is more helpful to lower-grade physicians.