实时可视化交互人工智能技术在细胞病理学教育中的实践探索
本研究纳入45名病理住院医师,实验组20人采用教师-人工智能-住院医师三方协作教学模式,对照组25人接受常规教学,均进行8周教学周期。结果显示实验组灰区细胞识别一致率术后达78.65%,显著高于干预前64.38%和对照组66.84%(t=8.962,p<.001),诊断速度也显著加快(3.05±0.52分钟/例 vs 干预前5.92±0.85分钟/例和对照组5.63±0.79分钟/例;t=14.821,p<.001)。结论认为该技术能有效提升住院医师细胞病理诊断技能,值得推广。
要点
- 实验组灰区细胞识别一致率78.65%,显著高于对照组66.84%(p<.001)。
- 实验组诊断速度3.05±0.52分钟/例,显著快于对照组5.63±0.79分钟/例(p<.001)。
- 问卷调查显示20人中19人认为新模式非常必要,15人认为显著提升诊断能力。
核心结论
- 实时可视化交互人工智能技术能有效提升住院医师的细胞病理诊断技能,值得在病理教育中推广。
研究设计与样本
- 纳入45名病理住院医师,实验组20人采用教师-人工智能-住院医师三方协作教学模式,对照组25人接受常规教学,均进行8周教学周期。
主要发现
- 实验组灰区细胞识别干预后一致率78.65%,显著高于干预前64.38%和对照组66.84%(t=8.962,p<.001)。
- 实验组诊断速度平均3.05±0.52分钟/例,显著快于干预前5.92±0.85分钟/例和对照组5.63±0.79分钟/例(t=14.821,p<.001)。
- 对照组干预前后无统计学显著变化(p>.05)。
- 实验组问卷中19/20人认为新模式非常必要,15/20人认为显著提升诊断能力。
- 半结构化访谈显示该模式增强诊断能力、促进个性化学习并缓解学习焦虑。
临床怎么用
- 可将实时可视化交互人工智能技术整合到细胞病理住院医师培训中,以提升灰区细胞识别能力和诊断速度。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND: Proficiency in cytopathologic diagnosis depends heavily on extensive hands-on practice and immediate error correction. Traditional teaching models, however, are constrained by limited practice opportunities and delayed feedback, which fails to meet the core skill-development needs of residents. METHODS: In total, 45 pathology residents were enrolled and assigned to two groups. The experimental group (n = 20) adopted a tripartite teacher-artificial intelligence-resident collaborative teaching model, whereas the control group (n = 25) received conventional instruction. Both groups underwent an identical 8-week teaching cycle. RESULTS: The questionnaire results from the experimental group indicated that 19 of 20 residents (9%) deemed the new model highly necessary, and 15 of 20 (75%) believed it significantly improved their diagnostic competence. Semistructured interviews further revealed that the model enhanced diagnostic ability, facilitated personalized learning, and alleviated learning anxiety. For objective metrics, the experimental group demonstrated a significantly higher postintervention concordance rate for gray-zone cell identification (78.65%) compared with both their preintervention baseline (64.38%) and the contemporaneous control group (66.84%; t = 8.962; p < .001). In addition, the experimental group exhibited a markedly faster diagnostic speed (mean ± standard deviation, 3.05 ± 0.52 minutes per case) compared with their preintervention performance (5.92 ± 0.85 minutes per case) and the control group (5.63 ± 0.79 minutes per case; t = 14.821; p < .001). No statistically significant changes were observed in the control group (p > .05). CONCLUSIONS: This study demonstrates that artificial intelligence technology integrated with real-time visual interaction effectively improves the cytopathologic diagnostic skills of residents and merits wider promotion in pathology education.
引用
- Ren W, Rong L, Wang Q, et al. Practical exploration of real-time visual interactive artificial intelligence technology in cytopathology education. Cancer Cytopathol. 2026. PMID: 42145147. doi:10.1002/cncy.70097
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。