论文部分内容阅读
目的:探讨计算机辅助检测系统(computer-aided detection,CAD)提高低剂量CT筛查者非钙化肺结节检出率方法的优化和临床应用价值。方法:抽取100例接受低剂量CT筛查的高危人群进行研究,分别使用3种方法阅读图像。方法 A:仅依靠CAD辅助阅片;方法 B:根据CAD辅助自动检出结节结果,由影像诊断医师通过分析CAD自动检出的目标结构(结节或非结节结构),以判定CAD自动检出的目标结果是否为真性非钙化肺结节;方法 C:在方法 B基础上,影像诊断医师通过薄层横断面图像阅片,分析并记录CAD结合影像医生薄层阅片共同检测结果。记录每种方法检出的每个患者结节总数、结节大小、结节位置和结节密度情况。最终以2名高年资影像诊断主任医师共同拟定的结节作为真结节参照标准。计算三种方法对非钙化肺结节的检出率、假阴性率和检出假结节总数,用χ2检验比较三种方法对非钙化肺结节检出率差异。结果:根据参照标准共检出287个真结节。方法 A共检出结节总数336个,其中真结节238个,方法 B共检出结节总数249个,真结节238个,方法 C共检出结节总数285个,真结节数274个。方法 C非钙化肺结节检出率95.50%明显高于方法 A(χ2=23.434,P<0.001)和方法 B(χ2=23.434,P<0.001)对非钙化结节检出率82.90%;方法 C及方法 B检出假阳性的非钙化肺结节数(11个)明显少于方法 A(98个)。结论:影像医生薄层阅片联合修正CAD结果的筛查方式明显提高LDCT筛查者非钙化肺结节检出率并降低假阳结节检出,可以作为高危人群LDCT筛查肺结节的首选方法。
Objective: To investigate the optimization and clinical value of computer-aided detection (CAD) in improving the detection rate of non-calcified pulmonary nodules in low-dose CT screening. Methods: A total of 100 high-risk patients undergoing low-dose CT screening were selected for study. Three methods were used to read the images. Method A: rely only on CAD-assisted reading; Method B: automatic detection of nodules based on CAD-assisted by the image diagnostic physician by analyzing CAD automatically detected target structure (nodules or non-nodular structure) to determine the CAD automatic Method C: On the basis of Method B, the imaging diagnostician analyzed and recorded the results of the joint examination of the thin-section images of the doctor with CAD in combination with the thin-section image reading. The total number of nodules, nodule size, nodule location, and nodule density for each patient were recorded for each method. Finally, two high-risk image diagnosis chief physician co-developed nodules as the true nodule reference standard. Calculate the detection rate, false negative rate and the total number of detected false nodules in the three methods for non-calcified pulmonary nodules, and compare the detection rate of non-calcified pulmonary nodules byχ2 test. Results: According to the reference standard a total of 287 true nodules were detected. A total of 336 nodules were detected in Method A, including 238 true nodules. Total number of nodules was detected in Method B, and total number of true nodules was 238 in Method B. A total of 285 nodules were detected in Method C, 274. The positive rate of non-calcified pulmonary nodules in method C was 95.50%, which was significantly higher than that of method A (χ2 = 23.434, P <0.001) and method B (χ2 = 23.434, C and Method B detected false positive non-calcified pulmonary nodules (11) was significantly less than Method A (98). CONCLUSION: The combination of thin-section imaging and modified CAD results screening method of imaging doctors can significantly improve the detection rate of non-calcified pulmonary nodules and reduce the detection of false-positive nodules in LDCT screening. It can be used as a screening method for LDCT in high-risk groups Preferred method.