实习医师以共同阅片和第二阅片者模式使用计算机辅助检测系统在低剂量CT中的应用研究

A computer-aided detection(CAD)system on low-dose chest CT images in concurrent-reader and second-reader modes:influence on interns

目的:评估使用计算机辅助检测(CAD)系统辅助在低剂量胸部CT图像中,以第二阅片者模式和共同阅片模式对实习医生肺结节检出的敏感性和读片时间的影响。方法:前瞻性纳入60例胸外恶性肿瘤患者,接受2次连续低剂量胸部CT扫描。3名实习生以盲法和随机顺序独立地对低剂量胸部CT图像进行三次阅片,一次使用共同阅片模式,第二次不使用CAD,第三次使用第二阅片模式。记录真阳性率(TP),假阳性率(FP)和阅读时间。以3名高年资胸部放射科医师(>20年)检出的肺结节(包括对CAD检出结节认可)的一致意见为真结节参照标准。结果:共同阅片模式(259s)的阅读时间明显短于不使用CAD的阅读时间(341s,P<0.01)和第二阅片模式(363s;P<0.01)。低剂量胸部CT,CAD作为共同阅片者(CR)的敏感性(67.36%)显著高于不使用CAD的敏感性(40.83%,P<0.001),而与CAD作为第二阅片者(SR)的敏感性(68.34%:P>0.05)无统计学差异。CAD作为CR的FP率(5.31/CT)高于无CAD的FP率(3.28/CT,P<0.0001),而与CAD作为SR的FP率(5.23/CT,P>0.05)无统计学差异。使用CAD在共同阅片模式下的FOM明显高于不使用CAD的FOM(分别为0.59和0.44,P<0.001),使用CAD在第二阅片模式的平均FOM值为0.60,与共同阅片模式无统计学意义(P=0.675)。结论:在低剂量CT图像,使用CAD显著提高实习医师对包括小结节在内的肺结节诊断敏感性;并且以共同读片模式使用CAD显著地缩短了实习医师的阅片时间。

Objective:To assess the value of a computer-aided detection(CAD)system in second-reader(SR)and concurrent-reader(CR)models for the detection of pulmonary nodules and time efficiency for interns on low-dose chest CT images.Methods:A total of 60 patients with extra-thoracic malignancies,who underwent low-dose CT scan,were prospectively included.3 interns independently evaluated images of 60 patients in a blinded and randomized manner thrice,once with CR twice without CAD,and third with SR in the images of low-dose chest CT.True positive rates(TP),false positive rates(FP)and reading times were recorded.Free-response receiver operating characteristic analysis was also used for the statistical analysis.The consensus reading of CT images by three senior chest radiologists(>20 years)and those obtained using the CAD system served as the reference standard.Results:Reading times were significantly shorter in CR mode(259s)than those without CAD(341s,P<0.01)and in SR mode(363s;P<0.01).Sensitivity of reading with CAD as CR(67.36%)was significantly higher than reading without CAD(40.83%,P<0.001)and not statistically different in CAD as SR(68.34%:P>0.05)in the low-dose setting.The FP rates using CAD as CR(5.31/CT)were higher than without CAD(3.28/CT,P<0.0001),while not statistically different in CAD as SR(5.23/CT,P>0.05).The FOM for the 3 interns in the CR mode was statistically signifcant better than that without CAD(0.59 vs 0.44,respectively;P<0.001),and the average FOM value in the SR mode was 0.60,giving not statistically different between tha CR and SR modes(P=0.675).Conclusion:CAD benefited interns in the low-dose settings including small nodules,which both signifcantly improved diagnostic performance and shortened the reading times.