In the en face images, cross-sections of the vessels in the sclera were delineated as dark dots or line segments, indicating that the blood cells and flow could not be detected with this SS-OCT system. Meanwhile superficial conjunctival vessels were detected as low-intensity tubes with a high-intensity line corresponding to the reflection from blood cells at the center of the tube. However, in the video of the en face images played from the level of the episclera to the interface between the sclera and the ciliary body, dark cross-sections of the vessels were detected as dark, moving objects against the bright sclera, although the blood flow itself could not be seen (
Supplementary Video S1). Therefore, the map of the scleral vessels (vasculature map) was constructed as a trajectory of dark, moving objects by using the motion contrast enhancement technique, which was originally developed for capillary visualization, assisted by the use of adaptive optics scanning laser ophthalmoscopy without a contrast dye (
Fig. 2).
13–15 First, 202 frames spanning from the episclera to the interface between the sclera and ciliary body were extracted, and after application of a Gaussian blur filter, 201 division images were calculated by dividing the pixels between sequential frames as [
Dj(
x,
y) =
Ij(
x,
y)/
Ij+1(
x,
y)], where
Ij(
x,
y) represents the intensity of frame
j. Then, the division images were divided into three equal stacks of frames, and for each stack, the variance of the pixels among all of the division images at each
x–y position was calculated to visualize the contrast-enhanced vessel images. As a consequence, three vascular images of different depths were obtained, and the images were merged in different colors (
Fig. 3E, superficial layer network, blue; intermediate layer network, green; deep layer, red) in order to visualize the connection between the layers. Digital image processing was performed by using ImageJ software (National Institutes of Health, Bethesda, MD; in the public domain:
http://rsb.info.nih.gov/ij/index.html), and a series of ImageJ commands was performed automatically by using a macro (
Supplementary Fig. S1).