摘要
目的探讨基于深度学习的计算机辅助诊断系统(DL-CAD)在提高放射医师对急性肋骨骨折诊断效能方面的作用。方法回顾性分析214例急性胸部钝挫伤患者的CT图像。两名放射实习医师和两名主治医师以盲法和随机的方式独立评估所有患者的CT图像,并于1个月后在DL-CAD的辅助下重新评估。以两名资深放射医师的一致诊断作为参考标准,比较前后两次阅片的诊断敏感度、特异度、阳性预测值、诊断信心和阅片时间。结果214例患者共有680处急性肋骨骨折。实习医师在DL-CAD辅助下前后两次的诊断敏感度(68.82%vs.91.76%,P<0.05)、阳性预测值(84.50%vs.93.17%,P<0.05)均显著提升,与主治医师独立或在DL-CAD辅助下的诊断敏感度和阳性预测值相比无显著性差异(P>0.05)。DL-CAD显著提高诊断医师的诊断信心(P<0.05),并显著降低医师的阅片时间[实习医师:(99.48±21.69)s vs.(46.40±26.40)s,P<0.05;主治医师:(65.96±17.08)s vs.(43.54±23.54)s,P<0.05]。结论DL-CAD能够降低急性肋骨骨折的阅片时间,同时提升放射医师的诊断信心以及诊断质量,提高不同经验医师的诊断一致性。
Objective To evaluate the value of a deep learning-based computer-aided diagnosis system(DL-CAD)in improving the diagnostic performance of acute rib fractures in patients with chest trauma.Methods CT images of 214 patients with acute blunt chest trauma were retrospectively analyzed by two interns and two attending radiologists independently and re-analyzed with the assistance of a DL-CAD one month later,using a blinded and randomized manner.The consistent diagnoses of two other senior thoracic radiologists were regarded as a reference standard.The diagnostic sensitivity,specificity,positive predictive value,diagnostic confidence and the mean reading time with and without assistance were compared.Results There were 680 rib fracture lesions among the 214 patients.The diagnostic sensitivity and positive predictive value of interns was significantly improved from 68.82%,84.50%to 91.76%,93.17%with the assistance of DL-CAD respectively(P<0.05),and was comparative to those of attendings aided by DL-CAD or not(P<0.05).In addition,when radiologists were assisted by DL-CAD,the mean reading time was significantly reduced and on the other hand their diagnostic confidence was significantly increased(P<0.05).Conclusion Assistance of DL-CAD improves the diagnostic performance of acute rib fracture in chest trauma patients,by increasing the diagnostic confidence,sensitivity and positive predictive value of radiologists,shortening the reading time and advancing the diagnostic consistency of radiologists with different experience.
作者
谭辉
田占雨
潘宁
于勇
于楠
段海峰
樊秋菊
薛育
李悦
TAN Hui;TIAN Zhanyu;PAN Ning(Department of Radiology,Affiliated Hospital of Shaanxi University of Chinese Medicine,Xianyang,Shaanxi Province 712000,P.R.China)
出处
《临床放射学杂志》
CSCD
北大核心
2020年第12期2493-2497,共5页
Journal of Clinical Radiology
基金
陕西中医药大学学科创新团队建设项目(编号:2019-QN09)
关键词
肋骨骨折
深度学习
计算机辅助诊断
诊断信心
阅片时间
Rib fracture
Deep learning
Computer aided diagnostic
Diagnostic performance
Reading time
