Abstract
Purpose :
Color fundus images and angiography images are used by clinicians to diagnose retinal diseases and monitor lesions. However, using both images require co-registration to correlate findings. Here, we assess the accuracy of co-registration of color fundus images with fluorescein angiography (FA) images, indocyanine green angiography (ICGA) images, and fused images containing information from FA and ICGA.
Methods :
Color fundus, FA, and ICGA images were acquired using CLARUS™ 700 (ZEISS, Dublin, CA) device. 7 subjects were imaged using both eyes. Fused FA and ICGA images were created by averaging the individual images. The ground-truth registration was acquired by manually identifying 6 landmarks, usually vessel bifurcations and the optic nerve on all images. We used the automatic co-registration algorithm from CLARUS, based on an affine transformation, to register the color fundus image to the angiography images. The ground truth landmark locations on the color fundus image were converted according to the transformation matrix of the algorithm, shown in Fig. 1. Those transformed locations were compared with the angiography image ground truth landmark locations. All locations were corrected for stereographic projection. We defined the mean, standard deviation (STD), and root mean square error (RMSE) in units of degrees subtended in air at the pupil.
Results :
The RMSE for the registration of FA, ICGA, and fused FA and ICGA images with the color fundus image are shown in Table 1. The fused FA and ICGA images provided the highest accuracy in the registration algorithm. The ICGA images provided the second highest accuracy possibly due to the appearance of both superficial and choroidal vessels.
Conclusions :
We demonstrated that registering color fundus images with a fused FA and ICGA image improves the accuracy when compared to registering on FA or ICGA alone. This strengthens the notion that both superficial and choroidal vessels are useful in the registration process.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.