文献速递胸部技术与安全DRCT

放射科工作人员在人工智能用于肺癌快速分诊实施过程中的观念演变

Evolution of radiology staff perspectives during artificial intelligence (AI) implementation for expedited lung cancer triage.
PubMed 文献 · Clin Radiol · 2024-09-21

研究通过三个时间点(实施前、实施后1个月和7个月)的问卷调查,评估放射科工作人员对用于胸部X线分诊、标记疑似肺癌以加速同日胸部CT检查的人工智能工具的看法。共收到45/195、29/195和53/195份回复,回复率分别为23.1%、14.9%和27.2%。工作人员最初期待AI节省时间(50.8%),实施后转向关注患者随访更快(51.7%);对AI自主使用的反对始终占多数(53.3-72.5%),但若自己作为患者,对个人使用AI的接受度从实施前31.1%升至实施后47.2%。

要点

  • 多数受访者始终不同意AI可自主使用(各期53.3-72.5%)。
  • 工作人员对AI的个人使用接受度从实施前31.1%升至实施后47.2%。
  • 实施后主要期待从节省时间转向患者随访更快(51.7%)。

核心结论

  • 人工智能在放射科的成功整合需要工作人员积极参与并解决其关切,将最初的兴奋与抵触转化为建设性适应;持续反馈对于优化AI部署策略至关重要。

研究设计与样本

  • 对某机构所有放射科工作人员在AI实施前、实施后1个月和7个月三个时间点发放问卷调查;共收到45/195、29/195和53/195份回复,回复率分别为23.1%、14.9%和27.2%。

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)。

英文原文(摘要)· 点击展开

AIMS: To investigate radiology staff perceptions of an AI tool for chest radiography triage, flagging findings suspicious for lung cancer to expedite same-day CT chest examination studies.

MATERIALS AND METHODS: Surveys were distributed to all radiology staff at three time points: at pre-implementation, one month and also seven months post-implementation of artificial intelligence (AI). Survey questions captured feedback on AI use and patient impact.

RESULTS: Survey response rates at the three time periods were 23.1% (45/195), 14.9% (29/195) and 27.2% (53/195), respectively. Most respondents initially anticipated AI to be time-saving for the department and patient (50.8%), but this shifted to faster follow-up care for patients after AI implementation (51.7%). From the free text comments, early apprehension about job role changes evolved into frustration regarding technical integration challenges after implementation. This later transitioned to a more balanced view of recognised patient benefits versus minor ongoing logistical issues by the late post-implementation stage. There was majority disagreement across all survey periods that AI could be considered to be used autonomously (53.3-72.5%), yet acceptance grew for personal AI usage if staff were to be patients themselves (from 31.1% pre-implementation to 47.2% post-implementation).

CONCLUSION: Successful AI integration in radiology demands active staff engagement, addressing concerns to transform initial mixed excitement and resistance into constructive adaptation. Continual feedback is vital for refining AI deployment strategies, ensuring its beneficial and sustainable incorporation into clinical care pathways.

来源:Clin Radiol,查看原文。版权归原作者所有。

引用

  1. Togher D, Dean G, Moon J, et al. Evolution of radiology staff perspectives during artificial intelligence (AI) implementation for expedited lung cancer triage. Clin Radiol. 2024. PMID: 39443240. doi:10.1016/j.crad.2024.09.010

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。