Computer-aided detection of nodule in low-dose CT screening for lung cancer
10.3760/cma.j.issn.1005-1201.2012.07.009
- VernacularTitle:计算机辅助检测系统在低剂量肺癌筛查结节检出中的应用价值
- Author:
Wei TANG
;
Jianwei WANG
;
Ning WU
;
Yao HUANG
;
Qiang CAI
;
Shijun ZHAO
;
Xiaojuan XU
- Publication Type:Journal Article
- Keywords:
Multiple pulmonary nodule;
Radiation dosage;
Tomography,X-ray computed
- From:
Chinese Journal of Radiology
2012;46(7):619-623
- CountryChina
- Language:Chinese
-
Abstract:
ObjectiveTo evaluate the performance of computer-aided detection (CAD) system for detection of pulmonary nodules in 64-slice low-dose CT screening and to investigate whether CAD can improve the performance of radiologists in detecting pulmonary nodules.MethodsOne hundred low-dose screening CT examinations were randomly selected from the database containing 578 consecutive cases between Jun 2007 and Jun 2008.All the examinations were performed on a 64-MSCT scanner with the exposure of 120 kVp,30 or 40 mA,or automatic exposure control.Before the study started,the screening reports had been made with double reading by two radiologists.All the selected images were analyzed with the lung VCAR software from GE Healthcare with a nodule diameter threshold 3.0 mm.All discrepancies between the screening reports and the CAD results were reviewed and the true non-calcified nodules were determined in consensus by two experienced chest radiologists.Detected nodules were classified by density,size and location.The performance of the double reading and the CAD system were compared and analyzed statistically.McNemar-Bowker test was used for the statistical analysis.ResultsA total of 257 true noncalcified nodules were determined in all 100 low-dose screening CT examinations.The detection rate of CAD system was 91.1％ (234/257),with the missed rate of 8.9％ (23/257).Twenty three nodules were missed by CAD,in which 10 were solid with the diameter ranged from 2.4 to 6.0 mm,and 13 were nonsolid with the diameter ranged from 2.1 to 8.6 mm.Of the 23 nodules,17 were located in the outer zones of lungs and 6 in the inner zones.The double reading showed a detection rate of 59.1％ ( 152/257 )and a missed rate of 40.9％ ( 105/257),which was significantly lower than CAD.The diameter of all the 105 missed nodules by radiologists were ranged from 2.4 to 11.8 mm,in which 94 nodules were solid,10 were partly solid and 1 was nonsolid,with 69 located in outer zones of lungs and 36 in the inner zones. Conclusions The capability of the CAD system for detecting non-calcified pulmonary nodules is superior to double reading in low-dose screening CT examination,especially for the nodules located in the inner zone of the lung.When lung VCAR is used,nonsolid pulmonary nodules are more easily missed so that it should be paid more attentions by radiologists.
