人工智能辅助诊断技术在低剂量胸部CT检查阅片中的应用效果  

Application effects of artificial intelligence-assisted diagnosis technology in low-dose chest CT image reading

作　　者：赵世龙[1] 燕晶晶 胡冰 汝晓双[1] ZHAO Shilong;YAN Jingjing;HU Bing;RU Xiaoshuang(Department of Radiology of Affiliated Zhongshan Hospital of Dalian University,Dalian 116000 Liaoning,China)

机构地区：[1]大连大学附属中山医院放射科,辽宁大连116000

出　　处：《中国民康医学》2021年第21期125-126,共2页Medical Journal of Chinese People’s Health

摘　　要：目的:观察人工智能(AI)辅助诊断技术在低剂量胸部CT检查阅片中的应用效果。方法:选取130例行低剂量胸部CT检查的患者为研究对象,先由放射科医师阅片,给出诊断结果,再结合AI辅助诊断技术阅片结果,由放射科医师重新给出诊断结果。以病理检查为诊断金标准,比较不同阅片方式中,不同病理类型结节(磨玻璃样结节、实性结节)检出率、不同直径(直径<0.5 cm与直径0.5~2.0 cm)结节检出率和结节总检出率。结果:经病理检查,130例患者共检出结节439个,其中磨玻璃样结节127个,实性结节312个;直径<0.5 cm 257个,0.5~2.0 cm 182个。AI辅助医师阅片的磨玻璃样结节检出率和总检出率明显高于单纯医师阅片,差异有统计学意义(P<0.05);AI辅助医师阅片的直径<0.5 cm结节检出率为93.77%,明显高于单纯医师阅片的88.72%,差异有统计学意义(P<0.05)。结论:AI辅助诊断技术应用于低剂量胸部CT检查阅片可提高磨玻璃样结节检出率、结节总检出率和直径<0.5 cm结节检出率,效果优于单纯医师阅片诊断效果。Objective:To observe application effects of artificial intelligence(AI)-assisted diagnosis technology in low-dose chest CT image reading Methods:130 patients who underwent low-dose chest CT examination were selected as the research objects.First,the radiologist read the images and gave the diagnosis results.Combined with the results of AI-assisted diagnosis technology,the radiologist then gave new diagnosis results.Using the pathological examination results as the diagnostic gold standard,the detection rates of different pathological nodules(ground glass nodules,solid nodules),the detection rates of the nodules of different sizes(diameter<0.5 cm and diameter 0.5-2.0 cm)and the total detection rate were compared the two image reading methods.Results:The pathological examination showed that a total of 439 nodules were detected in 130 patients,of which 127 were ground-glass nodules and 312 were solid nodules;and 257 were less than 0.5 cm in diameter and 182 were 0.5-2.0 cm in diameter.The detection rate and total detection rate of ground-glass nodules in AI-assisted physician reading were significantly higher than those of simple physician reading,and the differences were statistically significant(P<0.05).The detection rate of nodules with the diameter<0.5 cm in the AI-assisted physician reading was 93.77%,which was significantly higher than 88.72%of the simple physician reading,and the difference was statistically significant(P<0.05).Conclusions:AIassisted diagnosis technology applied to the low-dose chest CT image reading can improve the detection rate and the total detection rate of ground glass nodules as well as the detection rate of nodules with the diameter<0.5 cm.Moreover,it is superior to simple physician reading.

关 键 词：人工智能辅助诊断技术 低剂量 胸部CT 阅片 诊断 结节 检出率 

分 类 号：R445[医药卫生—影像医学与核医学]