人工智能应用于青光眼临床筛查及卫生效益分析

目的:探讨应用人工智能系统对青光眼患者的检出率、筛查的时间成本和经济成本,总结分析其应用于青光眼筛查的优越性与可行性。方法:回顾性研究。收集2019-2020年使用腾讯觅影眼底照片辅助诊断系统进行青光眼筛查的53700例受检者,2012-2018年使用北京同仁医院眼科研究所远程眼科会诊中心的493221例患者,邯郸眼病研究中接受基础眼病筛查的6716名受检者资料。评价人工智能筛查系统对青光眼的检出率,并采用Pearson x~2检验,比较人工智能筛查系统与远程阅片平台、基础眼病筛查对于青光眼检出率的差异,并对三种筛查方式的时间成本和经济成本进行比较分析。结果:临床应用表明,其成本效益的可靠数据及主要指标:检出率、x~2值、时间成本、经济成本等效益显著。人工智能筛查系统对青光眼的检出率为6.56%,远程眼科会诊平台的检出率为2.78%,邯郸眼病研究的检出率为1.70%,人工智能筛查系统与另外两种模式对青光眼诊断检出率的差异有统计学意义(P0.05)。而且使用人工智能筛查系统的患者其医疗时间成本和经济成本花费较低。结论:可行性分析表明,相较于远程眼科会诊平台和基础眼病筛查,人工智能筛查系统对青光眼的检出率较高,同时患者的时间成本和经济成本都得到了大幅度降低,这为青光眼筛查提供了一种有效而快捷的方式,值得临床推广。

Application and Benefit Analysis of Artificial Intelligence in Clinical Screening of Glaucoma

ABSTRACT Objective: To explore the detection rate, time cost and economic cost of the application of artificial intelligence system in glaucoma screening, and to summarize and analyze the advantages and feasibility of its application in glaucoma screening. Methods: Retrospective study. 53700 patients using Tencent Miying fundus photography diagnosis assistant system during 2019 to 2020, 493221 patients from the Center of Teleophthalmology Consultation in Beijing Tongren Eye Center from 2012 to 2018, and 6716 patients collected in Handan Eye Study were included. The detection rate of glaucoma by artificial intelligence screening system was evaluated, and compared to that of the Center of Teleophthalmology Consultation and Handan Eye Study. The difference in detection rates was analyzed by Pearson x² test. The time and economic cost of these three screening methods was also compared and analyzed. Results: The clinical application showed that the reliable data and main indexes of cost-benefit are improved significantly, including detection rate, x² value, time cost and economic cost. The detection rate of the artificial intelligence screening system for glaucoma was 6.56 %, and was 2.78 % and 1.70 % for the Center of Teleophthalmology Consultation and Handan Eye Study, respectively. The difference between the artificial intelligence screening system and the other two modes was statistically significant(P<0.05). In addition, patients using artificial intelligence screening system spend less time and economic cost. Conclusion: The feasibility analysis showed that compared with the teleophthalmology consultation platform and basic ophthalmic screening, the detection rate of the artificial intelligence screening system for glaucoma is higher, and the time and economic cost of patients have been reduced, which provided an effective and quick way for glaucoma screening.