Jacobs Retina Center, University of California, San Diego, CA, USA.
Viterbi Family Department of Ophthalmology and Shiley Eye Institute, University of California, San Diego, CA, USA.
Eye (Lond). 2024 Apr;38(6):1189-1195. doi: 10.1038/s41433-023-02868-3. Epub 2023 Dec 19.
This study aimed to compare a new Artificial Intelligence (AI) method to conventional mathematical warping in accurately overlaying peripheral retinal vessels from two different imaging devices: confocal scanning laser ophthalmoscope (cSLO) wide-field images and SLO ultra-wide field images.
Images were captured using the Heidelberg Spectralis 55-degree field-of-view and Optos ultra-wide field. The conventional mathematical warping was performed using Random Sample Consensus-Sample and Consensus sets (RANSAC-SC). This was compared to an AI alignment algorithm based on a one-way forward registration procedure consisting of full Convolutional Neural Networks (CNNs) with Outlier Rejection (OR CNN), as well as an iterative 3D camera pose optimization process (OR CNN + Distortion Correction [DC]). Images were provided in a checkerboard pattern, and peripheral vessels were graded in four quadrants based on alignment to the adjacent box.
A total of 660 boxes were analysed from 55 eyes. Dice scores were compared between the three methods (RANSAC-SC/OR CNN/OR CNN + DC): 0.3341/0.4665/4784 for fold 1-2 and 0.3315/0.4494/4596 for fold 2-1 in composite images. The images composed using the OR CNN + DC have a median rating of 4 (out of 5) versus 2 using RANSAC-SC. The odds of getting a higher grading level are 4.8 times higher using our OR CNN + DC than RANSAC-SC (p < 0.0001).
Peripheral retinal vessel alignment performed better using our AI algorithm than RANSAC-SC. This may help improve co-localizing retinal anatomy and pathology with our algorithm.
本研究旨在比较一种新的人工智能（AI）方法与传统的数学变形方法，以准确叠加两种不同成像设备（共焦激光扫描检眼镜（cSLO）宽视场图像和 SLO 超广角图像）的周边视网膜血管。
使用 Heidelberg Spectralis 55 度视野和 Optos 超广角采集图像。采用随机样本一致-样本集和一致集（RANSAC-SC）进行传统的数学变形。将其与基于单向正向配准过程的 AI 配准算法进行比较，该过程包括全卷积神经网络（CNN）和异常值拒绝（OR CNN），以及迭代 3D 相机姿态优化过程（OR CNN+失真校正[DC]）。图像以棋盘格模式提供，根据与相邻框的对齐情况，在四个象限中对周边血管进行分级。
共分析了 55 只眼中的 660 个方框。比较了三种方法（RANSAC-SC/OR CNN/OR CNN+DC）的 Dice 评分：复合图像中折叠 1-2 的 0.3341/0.4665/4784 和折叠 2-1 的 0.3315/0.4494/4596。使用 OR CNN+DC 组合的图像中位数评分为 4（满分 5 分），而使用 RANSAC-SC 的评分为 2。使用我们的 OR CNN+DC 比 RANSAC-SC 获得更高分级水平的几率高 4.8 倍（p<0.0001）。
与 RANSAC-SC 相比，我们的 AI 算法在周边视网膜血管配准方面表现更好。这可能有助于我们的算法更好地定位视网膜解剖结构和病理。
Zotero 插件
