首页|人工智能对医师诊断新鲜肋骨骨折的应用价值

人工智能对医师诊断新鲜肋骨骨折的应用价值

扫码查看
目的 探讨人工智能(AI)软件对放射科住院医师、主治医师在新鲜肋骨骨折病灶检出率的差异及其应用AI前后的一致性评价,评估应用AI后各级医师诊断新鲜肋骨骨折效能的提高情况。方法 收集因急性胸部外伤行胸部CT扫描病例300例,其中确诊为肋骨骨折的152例病例。将6位医师分为住院医师组和主治医师组两组,3人/组,对随机分配的300例CT图像独立阅片。在间隔为4周的洗脱期后,医师结合AI第2次阅片。采用卡方检验比较两组医师对新鲜肋骨骨折病灶及不同类型病灶检出率的差异,并评价各组应用AI前后的一致性、敏感度、特异度差异。结果 应用AI后住院医师、主治医师对所有新鲜肋骨骨折、完全性骨折、不完全性骨折的检出率均高于医师单独阅片,差异有统计学意义(P<0。001)。住院医师+AI、主治医师+AI对全部肋骨骨折、不完全性骨折的Kappa值与Phi系数均明显提高,不完全性骨折提高幅度最显著。住院医师+AI、主治医师+AI检出新鲜肋骨骨折敏感度与住院医师、主治医师单独阅片的差异有统计学意义(P<0。001),特异度的差异无统计学意义。结论 AI可有效提高不同级别医师对新鲜肋骨骨折的检出效能,并提高不同级别医师之间的一致性、敏感度。
Application value of artificial intelligence in the diagnosis of fresh rib fractures by physicians
Objective To investigate the difference of detection rate of fresh rib fracture lesions between radiology residents and attending physicians by artificial intelligence(AI)software and the consistency evaluation before and after application of AI,and to evaluate the improvement of efficiency of doctors at all levels in diagnosing fresh rib fractures after applying AI.Methods A total of 300 patients with acute chest trauma underwent chest CT scan,152 of which were confirmed to be rib fracture.6 physicians were divided into resident physician group and attending physician group,with three physicians in each.300 randomly assigned CT images were reviewed independently.After the washout interval of 4 weeks,the physicians combined with AI read the film for second time.Chi-square test was used to compare the difference in the detection rate of fresh rib fracture lesions and different types of lesions between the two groups,and to evaluate the difference in consistency,sensitivity and specificity before and after the application of AI in each group.Results After applying AI,the detection rate of all fresh rib fractures,complete fractures and incomplete fractures by residents and attending physicians was higher than that by physicians alone,and the difference was statistically significant(P<0.001).The Kappa values and Phi coefficients of all rib fractures and incomplete fractures were significantly improved by residents+AI and attending physicians+AI,and the improvement of incomplete fractures was the most significant.The sensitivity of fresh rib fractures detected by residents+AI,attending physicians+AI were significantly different from that by residents and attending physicians alone(P<0.001),there was no significant difference in specificity.Conclusion AI can effectively improve the detection efficiency of fresh rib fractures among physicians of different levels,and improve the consistency and sensitivity among physicians of different levels.

artificial intelligencefresh rib fracturecomputer tomography

李玲、郑爽爽、柳丽

展开 >

首都医科大学附属复兴医院放射科,北京 100038

人工智能 新鲜肋骨骨折 计算机断层扫描

2024

分子影像学杂志
南方医科大学

分子影像学杂志

CSTPCD
ISSN:1674-4500
年,卷(期):2024.47(3)
  • 10