目的 比较放射科医师在人工智能（artificial intelligence，AI）骨龄评价系统辅助前后对儿童左手X线摄影的骨龄评估效能。 方法 回顾性分析在我院就诊的300例患儿左手X线平片。骨龄评测采用中华-05骨龄评定标准，两位低年资医师（医师1及医师2）分别在有无AI辅助下分别记录左手各骨质的骨龄发育等级，并记录时间。以两位高年资放射医师分别在有无AI系统辅助下评估结果的均值为参考标准，计算骨龄测评的准确率、均方根误差（root mean square error，RMSE）、测评时间。 结果 无AI辅助下，医师1及医师2分析差值在6个月及12个月诊断准确率分别为77.3%和83%、88.7%和93.7%，RMSE值分别为9和8。在AI辅助下，医师1及医师2分析差值在6个月及12个月诊断准确率分别为88.7%和90.3%、97%和97.3%，RMSE值分别为6和6；差异均具有统计学意义。无AI辅助下，实验组医师和标准组医师，平均评测耗时分别为124.79 s和89.13 s；有AI辅助下实验组医师和标准组医师，平均评测耗时分别为86.10 s和63.87 s，在AI辅助下平均评测耗时均有较大幅度减少（P=0.000）。 结论 AI辅助骨龄评价系统可显著提高医师工作效率，减少阅片时间。

Objective To compare the bone age assessment efficiency of radiologists for children by left hand radiography before and after the implementation of the artificial intelligence（AI）-assisted bone age assessment system. Methods We conducted a retrospective analysis of left hand X-ray plain films of 300 children treated in our hospital. The China-05 standards were used to assess bone age. The bone age development grade of each bone in the left hand was assessed by two junior physicians（physician 1 and physician 2，experimental group） with and without the assistance of the AI system，and the time was recorded. The accuracy，root mean square error（RMSE），and time of bone age assessment were calculated with the mean values of assessment results of two senior radiologists（control group） with and without the assistance of the AI system as the reference standards. Results Without the assistance of the AI system，the diagnostic accuracy rates of physician 1 and physician 2 were 77.3%/83% and 88.7%/93.7% at month 6 and month 12，respectively，and the RMSE values were 9 and 8，respectively. With the assistance of the AI system，the diagnostic accuracy rates of physician 1 and physician 2 were 88.7%/90.3% and 97%/97.3% at month 6 and month 12，respectively，and the RMSE values were 6 and 6，respectively，showing significant differences. Without the assistance of the AI system，the mean assessment time of physicians in the experimental and control groups was 124.79 s and 89.13 s，respectively. With the assistance of the AI system，the mean assessment time of physicians in the experimental and control groups was 86.10 s and 63.87 s，respectively. By utilizing AI，the mean assessment time was significantly reduced（P<0.001）. Conclusion The AI-assisted bone age assessment system can significantly improve physicians’ work efficiency and reduce the film reading time.