Three-dimensional reconstruction of blood vessels in the rabbit eye by X-ray phase contrast imaging

Abstract

Background: A clear understanding of the blood vessels in the eye is helpful in the diagnosis and treatment of ophthalmic diseases, such as glaucoma. Conventional techniques such as micro-CT imaging and histology are not sufficiently accurate to identify the vessels in the eye, because their diameter is just a few microns. The newly developed medical imaging technology, X-ray phase-contrast imaging (XPCI), is able to distinguish the structure of the vessels in the eye. In this study, XPCI was used to identify the internal structure of the blood vessels in the eye.

Methods: After injection with barium sulfate via the ear border artery, an anesthetized rabbit was killed and its eye was fixed in vitro in 10% formalin solution. We acquired images using XPCI at the Shanghai Synchrotron Radiation Facility. The datasets were converted into slices by filtered back-projection (FBP). An angiographic score was obtained as a parameter to quantify the density of the blood vessels. A three-dimensional (3D) model of the blood vessels was then established using Amira 5.2 software.

Results: With XPCI, blood vessels in the rabbit eye as small as 18 μm in diameter and a sixth of the long posterior ciliary artery could be clearly distinguished. In the 3D model, we obtained the level 4 branch structure of vessels in the fundus. The diameters of the arteria centralis retinae and its branches are about 200 μm, 110 μm, 95 μm, 80 μm and 40 μm. The diameters of the circulus arteriosus iridis major and its branches are about 210 μm, 70 μm and 30 μm. Analysis of vessel density using the angiographic score showed that the blood vessels had maximum density in the fundus and minimum density in the area anterior to the equator (scores 0.27 ± 0.029 and 0.16 ± 0.032, respectively). We performed quantitative angiographic analysis of the blood vessels to further investigate the density of the vessels.

Conclusions: XPCI provided a feasible means to determine the structure of the blood vessels in the eye. We were able to determine the diameters and morphological characteristics of the vessels from both 2D images and the 3D model. By analyzing the images, we obtained measurements of the density distribution of the microvasculature, and this approach may provide valuable reference information prior to glaucoma filtration surgery.

Figure 1

Setup of the X-ray Imaging and biomedical application beamline unit, Shanghai Synchrotron Radiation Facility. The stage has a six-joint control system that allows placement of the sample in the center of the field of view of the CCD. During scanning, the sample is placed on the turntable and the turntable is rotated around its cylindrical axis by 180° in steps of 1°.

Figure 2

Algorithm flowchart of the program used to convert the projected images into slices. The sinograms were computed by FBP based on the energy, distance and number of projected images. Because of the nonuniform thickness of the tissues of the eye, the gray value of the background has not been agreed. So we adjusted the gray value of background to remove the background noise.

Figure 3

One slice of the blood vessel database. Projected images were converted to slices to establish a blood vessel database, and then the 3D model of the blood vessels could be obtained. The white points are cross section of the blood vessels, and the circle outside the eye is polystyrene used to fix the eye.

Figure 4

Measurement of the blood vessels. Left is a projected image of group CT 2. The number of pixels contained in the blood vessels was 7.8. Each pixel was 9 μm, so this blood vessel was about 70 μm in diameter. The diameter of the visualized vessels was quantified using Image-Pro Plus.

Figure 5

Binary image of region of interest. (a) The area was chosen as the region of interest. (b) By adjusting a suitable threshold, the image was converted to a binary image. (c) The result after using a morphologically open processing algorithm.

Figure 6

Blood vessels in the rabbit eye are clearer in the XPCI image than in the X-ray absorption image. (a) X-ray absorption image, in which the distance between the stage and the CCD is zero. (b) XPCI of the rabbit eye. The distance is 700 mm and the energy is 22 keV. Because of the limited beam size, our samples were divided the eye into four groups, CT 1, CT 2, CT 3 and CT 4.

Figure 7

Segments of the vascular three-dimensional model. (a) Circulus arteriosus iridis major and its branches. (b) A branch of the circulus arteriosus iridis major with the minimum diameter about 30 μm. (c) Blood vessels of the fundus. (d) Branches of the fundus vessels with a minimum vessel diameter of about 40 μm.

Figure 8

3D reconstruction of the vasoganglion in the rabbit eye. This figure is a vertical view from the top of the cornea to the fundus. The tissues surrounding the blood vessels have been removed. The 3D model could help observe the morphology and obtain the diameter of any vessel at any angle.

Figure 9

Mean angiographic scores in the XPCI image groups. Each group is composed of 27 ROIs and the score represents the density of the blood vessels in this region. The scores for CT 1, CT 2, CT 3 and CT 4 are 0.27 ± 0.029, 0.22 ± 0.026, 0.16 ± 0.032 and 0.25 ± 0.083 (mean ± SD), respectively.