AI对急诊外伤肋骨骨折诊断效能的应用研究  

Performance evaluation of artificial intelligence(AI)detection system in the diagnosis of emergency traumatic rib fractures

作　　者：赵艳红[1] 张晓文 苏治祥 张涛[1] 曹永佩 宋静[1] 哈若水[1] ZHAO Yanhong;ZHANG Xiaowen;SU Zhixiang;ZHANG Tao;CAO Yongpei;SONG Jing;HA Ruoshui(Medical Imaging Center,People′s Hospital of Ningxia Hui Autonomous Region,Yinchuan 750002,Ningxia,China)

机构地区：[1]宁夏回族自治区人民医院医学影像中心,宁夏银川750002

出　　处：《宁夏医学杂志》2024年第7期607-609,共3页Ningxia Medical Journal

基　　金：北方民族大学一般科研项目(2022XYZYX03)。

摘　　要：目的探讨CT不同重建算法图像对人工智能(AI)辅助肋骨骨折检测效能的影响及AI辅助诊断系统对放射科住院医师诊断急诊外伤肋骨骨折诊断效能的影响。方法收集急诊胸部外伤病史的患者50例,所有患者均利用256层螺旋CT进行胸部CT成像,使用Standard、Lung、Bone和Soft 4种不同kernel重建算法进行图像重建,计算并比较AI在不同重建算法胸部CT图像下对肋骨骨折识别灵敏度、特异度及准确度的差异。结果AI在Standard、Lung、Bone和Soft 4种算法下对骨折检出的灵敏度分别为75.74%、86.63%、89.60%和69.31%,Lung和Bone算法下AI对肋骨骨折检出的灵敏度高于Standard和Soft(P<0.05)。AI在Standard、Lung、Bone和Soft 4种算法下对骨折检出的特异度分别为93.88%、93.17%、93.78%和94.18%,4种算法比较差异无统计学意义(P>0.05)。AI在4种算法下的准确度分别为90.82%、92.07%,93.07%和89.98%,4种算法比较差异有统计学意义(P<0.05)。放射科住院医师不使用AI和使用AI诊断肋骨骨折的灵敏度分别为82.18%、96.53%,特异度分别为95.08%、94.78%,准确度分别为92.90%、95.08%,使用AI后对肋骨骨折诊断的灵敏度及准确度均明显提高(P<0.05),而特异度比较差异无统计学意义(P>0.05)。放射科住院医师不使用AI诊断所用的时间为(240.79±63.20)s,使用AI所用的时间为(105.26±57.20)s,两个时间比较差异有统计学意义(P<0.05)。结论放射科住院医师利用AI能够明显缩短急诊外伤肋骨骨折诊断所用时间,提高工作效率,而且能够提高其对于急诊外伤肋骨骨折诊断的准确性。AI在帮助放射科住院医师更准确、更快速识别肋骨骨折中具有重要的价值。Objective To evaluate the impact of different CT reconstruction algorithms on the efficiency of artificial intelligence(AI)detection system for rib fracture diagnosis and to explore the impact of AI assisted diagnostic systems on the diagnostic efficiency of radiology resident physicians in diagnosing emergency traumatic rib fractures.Methods A total of 50 patients with emergent chest trauma underwent chest CT examination(Revolution CT,GE Healthcare)were retrospectively collected.All patients were reconstructed with Standard/Lung/Bone/Soft algorithms.The sensitivity,specificity and accuracy of rib fracture detection under different reconstruction algorithms by AI were compared.Results The sensitivities of AI for rib fractures detection under the four algorithms of Standard,Lung,Bone and Soft were 75.74%,86.63%,89.60%and 69.31%,respectively.The sensitivity of AI for rib fractures detection under the Lung and Bone algorithms was higher than that of Standard and Soft(P<0.05).The specificities of AI for rib fractures detection under Standard,Lung,Bone and Soft algorithms were 93.88%、93.17%、93.78%and 94.18%,respectively,with no statistical significance(P>0.05).The accuracies of the four algorithms were 90.82%、92.07%、93.07%and 89.98%,respectively,and the difference was statistically significant(P<0.05).The sensitivities of residents in the diagnosis of rib fracture without AI and with AI were 82.18%and 96.53%,the specificities were 95.08%and 94.78%,and the accuracies were 92.90%and 95.08%,respectively.The sensitivity and accuracy of doctors in the diagnosis of rib fracture were significantly improved after using AI(P<0.05).There was no significant difference in specificity(P>0.05).The reading time using AI(105.26±57.20 s)was greatly reduced compared with that without using AI(240.79±63.20)s(P<0.05).Conclusion The AI-assisted diagnostic system significantly improved rib fracture detection sensitivity and reduced diagnostic time.AI technology is of great value in assisting residents physicians to identify rib f

关 键 词：人工智能 深度学习 肋骨骨折 诊断效能 重建算法 

分 类 号：R445[医药卫生—影像医学与核医学]