文献速递白内障与屈光手术

实时人工智能在白内障手术中的应用

Application of real-time artificial intelligence to cataract surgery.
PubMed 文献 · Br J Ophthalmol · 2025-11-20

研究分两阶段:第一阶段分析Moorfields眼科医院成人白内障手术视频192段(2020年3月至2021年3月),基于国际眼科理事会手术能力评估表工具将手术分段;第二阶段招募74例术中使用三块辅助屏幕实时显示阶段检测信息的超声乳化病例,与26例无屏幕病例比较。实时人工智能使顾问医生手术时长显著缩短。

要点

  • 第一阶段分析192段白内障手术视频,顾问医生(n=68)平均手术时长11.18分钟,高级学员(n=59)17.54分钟,初级学员(n=65)21.36分钟。
  • 第一阶段以高级学员和顾问医生病例的中位指标值分别确定效率基准。
  • 第二阶段实时人工智能使顾问医生手术时长显著缩短;研究结论认为实时人工智能有改善手术效率和手术团队培训的潜力。

核心结论

  • 研究首次展示使用完全独立的人工智能平台分析白内障手术效率指标。
  • 实时人工智能有改善手术效率和手术团队培训的潜力。

研究设计与样本

  • 第一阶段为手术视频分析:采集Moorfields眼科医院成人白内障手术录像,通过Touch Surgery Ecosystem开发人工智能,基于国际眼科理事会-眼科手术能力评估表工具将手术分为不同阶段。
  • 第二阶段为实时手术研究:在证实人工智能第一阶段功能后,另招募一组超声乳化患者进入实时手术研究组,在手术室部署三块辅助屏幕实时显示阶段检测和阶段相关信息。
  • 第一阶段于2020年3月至2021年3月间分析192段视频;第二阶段比较74例部署屏幕病例与26例未部署屏幕病例的效率指标。

主要发现

  • 第一阶段:顾问医生(n=68)、高级学员(n=59)和初级学员(n=65)的平均手术时长分别为11.18、17.54和21.36分钟。
  • 第一阶段:分别以高级学员和顾问医生病例的中位指标值确定效率基准。
  • 第二阶段:使用实时人工智能时,顾问医生手术时长显著缩短。

关键数值与推荐

  • 第一阶段基于国际眼科理事会-眼科手术能力评估表工具将手术分段。
  • 第一阶段分别以高级学员和顾问医生病例的中位指标值确定效率基准。
  • 第二阶段比较74例部署屏幕病例与26例未部署屏幕病例的效率指标。

临床怎么用

  • 原文未据此给出推荐。

局限性

  • 原文未报告局限性。

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/AIMS: Artificial intelligence (AI) in Ophthalmology has yet to be applied to real-time cataract surgery. This work explores a new AI tool, developed for phacoemulsification, and evaluates its potential uses.First, our study aimed to demonstrate the use of AI in phase recognition and analysis of phacoemulsification. Second, to evaluate the application of real-time AI to live cataract surgery. METHODS: Phase I: surgical video recordings of adult patients undergoing cataract surgery at Moorfields Eye Hospital were captured. The AI, via Touch Surgery Ecosystem, was developed and used to segment surgery into phases based on the International Council of Ophthalmology-Ophthalmology Surgical Competency Assessment Rubric tool. PHASE II: having demonstrated the AI's functionality in phase I, a further group of phacoemulsification patients was recruited into a live surgery study arm. Three auxiliary screens were deployed in the operating theatres, displaying phase detection and phase relevant information in real time. RESULTS: Phase I: 192 videos were analysed between March 2020 and March 2021. Average case duration for consultants (n=68), advanced trainees (n=59) and junior trainees (n=65) was 11.18, 17.54 and 21.36 min, respectively. Efficiency benchmarks were determined using the median metric values for advanced trainee and consultant cases, respectively. PHASE II: efficiency metrics for 74 cases with screen deployment and 26 without were compared. With real-time AI, consultant surgeons had a significant decrease in case duration. CONCLUSIONS: We demonstrate the first use of a fully independent AI platform for analysing efficiency metrics in cataract surgery. Real-time AI has the potential to improve operative efficiency and surgical team training.

来源:Br J Ophthalmol,查看原文。版权归原作者所有。

引用

  1. Shah N, Wawrzynski J, Hussain R, et al. Application of real-time artificial intelligence to cataract surgery. Br J Ophthalmol. 2025. PMID: 40664490. doi:10.1136/bjo-2024-326111

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。