目的 探讨人工智能（artificial intelligence，AI）在神经病学研究生腰椎穿刺术考核中的应用。 方法 将腰椎穿刺评分细化归类为机器可深入学习的语音类、场景监控类及动作评估类评分，在既往研究工作的基础上建立音频算法模块、监控算法模块以及动作评估算法模块，并对79名神经病学三年级专业型硕士研究生的腰椎穿刺术考核分别进行AI评分和考官评分，比较AI评分与考官评分的差异。 结果 在语音类评分上AI评分明显低于考官评分［（52.86±3.25）分 vs.（54.32±2.65）分，P＜0.05］，在场景监控类以及动作评估类评分上AI评分与考官评分差异无统计学意义［（11.24±0.55）分 vs.（11.15±0.62）分，P＞0.05；（28.09±11.37）分 vs.（28.34±1.21）分，P＞0.05］。 结论 AI技术在研究生腰椎穿刺术操作考核中的应用具有客观性、可行性，但仍需逐渐迭代优化机器AI评分系统的性能。

Objective To explore the application of artificial intelligence (AI) in the examination of lumbar puncture for postgraduate students in neurology. Methods The lumbar puncture scores were further classified into speech, scene monitoring and motion evaluation, and they would be learned by the machine. Based on previous research, audio algorithm module, monitoring algorithm module and motion evaluation algorithm module are established. AI scores of lumbar puncture examination were compared with examiners' scores for 79 postgraduate students in neurology. Results AI scores were significantly lower than examiners' scores in speech score (52.86±3.25 points vs. 54.32±2.65 points, P＜0.05), while there was no significant difference between AI and examiners' scores in scene monitoring and motion evaluation (11.24±0.55 points vs. 11.15±0.62 points, P＞0.05; 28.09±11.37 points vs. 28.34±1.21 points, P＞0.05). Conclusion The application of AI technology in the evaluation of lumbar puncture operation of postgraduate students is objective and feasible, but the performance of the machine AI scoring system needs to be gradually iteratively optimized.