Quantification of retinal blood leakage in fundus fluorescein angiography in a retinal angiogenesis model

Abstract

Blood leakage from the vessels in the eye is the hallmark of many vascular eye diseases. One of the preclinical mouse models of retinal blood leakage, the very-low-density-lipoprotein receptor deficient mouse (Vldlr^-/-), is used for drug screening and mechanistic studies. Vessel leakage is usually examined using Fundus fluorescein angiography (FFA). However, interpreting FFA images of the Vldlr^-/- model is challenging as no automated and objective techniques exist for this model. A pipeline has been developed for quantifying leakage intensity and area including three tasks: (i) blood leakage identification, (ii) blood vessel segmentation, and (iii) image registration. Morphological operations followed by log-Gabor quadrature filters were used to identify leakage regions. In addition, a novel optic disk detection algorithm based on graph analysis was developed for registering the images at different timepoints. Blood leakage intensity and area measured by the methodology were compared to ground truth quantifications produced by two annotators. The relative difference between the quantifications from the method and those obtained from ground truth images was around 10% ± 6% for leakage intensity and 17% ± 8% for leakage region. The Pearson correlation coefficient between the method results and the ground truth was around 0.98 for leakage intensity and 0.94 for leakage region. Therefore, we presented a computational method for quantifying retinal vascular leakage and vessels using FFA in a preclinical angiogenesis model, the Vldlr^-/- model.

Figure 1

(a,b) Flowcharts showing the steps of the proposed methodology for (a) detecting leakage regions and blood vessels and (b) registering the FFA image sequences. White rectangles indicate processing steps and shaded rectangles indicate the respective resulting images. Note that the large vessels mask image is used as input for image registration. (c,d) Illustration of the methodology used for detecting a stable point inside the optic disk. (c) Linear regression applied to the neighborhood of a pixel (red point) in the medial axis of a blood vessel. The pixels used in the calculation are those inside the top circle. The dashed line in top circle indicates the obtained line. (d) The line is drawn in the accumulator array. The dot indicates the original point for reference. ONH: optic nerve head; $L_0$ defines a scale of analysis for applying the linear regression.

Figure 2

(a)–(c) Representative results of the main steps of the blood leakage quantification methodology. (a) Original image, (b) image ${\tilde{I}}_l$ obtained after small vessel removal, (c) image  ${\tilde{I}}_v$ obtained after grayscale opening using orthogonal structuring elements. (d,e) Representative images obtained from our method. (d) Blood leakage image obtained after small and large vessel removal. (e) Segmented blood vessels.

Figure 3

(a)–(c)Reference point, shown in red, obtained for FFA images from Vldlr^−/− mice at different time points. The blue dot indicates the pixel position of the red dot in (a) for reference. The blue circle indicates the position of the ROI of the first time point on the subsequent time points and can be used as a visual guide for the movement of the sample along time. (d–f) Example of FFA leakage quantification allowed by the leakage segmentation methodology. (d) Representative FFA images from Vldlr^−/− mice at early and late time points post fluorescein injection, respectively. (e) Relative fluorescein intensity as a function of time for FFA images. (f) Leakage region divided by ROI area as a function of time calculated for the FFA images (n = 5). (g) Relative fluorescein intensity inside the identified leakage regions as a function of time for FFA images (n = 5).

Figure 4

(a) Example of manually marked leakage regions from two annotators. These annotations were used as ground truth for assessing the performance of the leakage identification methodology. (b) Comparison of leakage intensity (upper row of plots) and area (lower row of plots) between ground truth images produced by two annotators and the results of the method for three time points of three different eyes.