比较欧堡眼底图像结合人工智能分析、OCT及OCTA对垂直杯盘比一致性分析

Comparison of the Consistency of the Vertical Cup/Disc Ratio Analyzed by Daytona Combined with Artificial Intelligence, OCT and OCTA

目的:评估欧堡Daytona眼底图像结合人工智能分析(artifical intelligence, AI)、蔡司高分辨率光学相干断层扫描仪(ZEISS Cirrus high density coherence tomography, Zeiss Cirrus HD-OCT)及光学相干断层扫描血流成像(optical coherence tomography angiography, OCTA)三种不同影像学技术所分析并计算出的垂直杯盘比(vertical cup/disc ratio, VC/D)大小的差异性及一致性,以指导临床应用。方法:前瞻性,横断面研究。纳入2021年9月~2022年1月在暨南大学附属深圳爱尔眼科医院就诊的患者25例(48只眼)进行OCT、OCTA及欧堡眼底图像联合AI分析检查并计算出视盘参数,对比三种设备测量VC/D的一致性、差异性。结果:OCT、OCTA及欧堡眼底图像联合AI分析三种技术测量VC/D值两两比较存在差异(P 欧堡眼底图像联合AI分析 > OCTA;经Bland-Altman分析:欧堡眼底照相联合AI分析测量所得VC/D值与OCTA的一致性较其余两组更为接近。其中欧堡眼底图像联合AI分析组与OCTA组测VC/D值95%的LoA:−0.2765~0.3981,6.25%点位于95%的LoA之外,一致性差;欧堡联合AI组与OCT组测VC/D值95%的LoA:−0.2712~0.1838,8.3%点位于95%的LoA之外,一致性差;OCTA组与OCT组测VC/D值95%的LoA:−0.1046~0.12286,8.3%点位于95%的LoA之外,一致性差。结论:三种不同的检查设备对VC/D的测量之间存在差异,一致性差,不同设备之间的VC/D值不可互换。

Objective: To evaluate the difference and consistency of vertical cup/disc ratio analyzed and calculated by Daytona fundus images combined with artificial intelligence analysis (AI), Zeiss Cirrus high resolution optical coherence tomography (Zeiss Cirrus HD-OCT) and optical coherence tomography angiography (OCTA), so as to guide clinical application. Methods: A prospective, cross-sectional study. A total of 25 patients (48 eyes) who were treated in Shenzhen Eye Hospital, Jinan University from September 2021 to January 2022 were included. They all were underwent examination of OCT, OCTA, and Daytona fundus images combined with AI, the optic disc parameters were calculated. The consistency and difference of vertical cup-to-disc ratio measured by three kinds of equipment were compared. Results: There were significant differences in the pairwise comparisons of the VC/D ratio measured by the three techniques of OCT, OCTA and Daytona fundus images combined with AI (P Daytona combined AI > OCTA. The consistency of the VC/D values measured by Daytona fundus images combined with AI analysis group and OCTA group was closer than that of the other two groups. The Bland-Altman plots showed that the 95% of the LoA were −0.2765 ~ 0.3981 between the Daytona fundus images combined with AI group and the OCTA group, 6.25% of the points were outside the 95% LoA, and the consistency was poor;the 95% of the LoA were −0.2712 ~ 0.1838 for Daytona fundus images combined with AI group and OCT group , 8.3 % of the points were outside the 95% LoA, poor consistency;the 95%of the LoA were −0.1046 ~ 0.12286 for OCTA group and OCT group ,8.3 % of the points were outside the 95% LoA, poor consistency. Conclusion: There are differences in the VC/D measurements of the three different devices, the consistency is poor, and the VC/D values between different devices are not interchangeable.