Comparison of Retinal Imaging Techniques in Individuals with Pulmonary Artery Hypertension Using Vessel Generation Analysis

Abstract

(1) Background: Retinal vascular imaging plays an essential role in diagnosing and managing chronic diseases such as diabetic retinopathy, sickle cell retinopathy, and systemic hypertension. Previously, we have shown that individuals with pulmonary arterial hypertension (PAH), a rare disorder, exhibit unique retinal vascular changes as seen using fluorescein angiography (FA) and that these changes correlate with PAH severity. This study aimed to determine if color fundus (CF) imaging could garner identical retinal information as previously seen using FA images in individuals with PAH. (2) Methods: VESGEN, computer software which provides detailed vascular patterns, was used to compare manual segmentations of FA to CF imaging in PAH subjects (n = 9) followed by deep learning (DL) processing of CF imaging to increase the speed of analysis and facilitate a noninvasive clinical translation. (3) Results: When manual segmentation of FA and CF images were compared using VESGEN analysis, both showed identical tortuosity and vessel area density measures. This remained true even when separating images based on arterial trees only. However, this was not observed with microvessels. DL segmentation when compared to manual segmentation of CF images showed similarities in vascular structure as defined by fractal dimension. Similarities were lost for tortuosity and vessel area density when comparing manual CF imaging to DL imaging. (4) Conclusions: Noninvasive imaging such as CF can be used with VESGEN to provide an accurate and safe assessment of retinal vascular changes in individuals with PAH. In addition to providing insight into possible future clinical translational use.

Keywords: VESGEN; fluorescein angiography; image processing; retinal imaging; vascular segmentation.

Figure 1

Representative output image of VESGEN analysis of the retinal vessels from PAH subjects. VESGEN imaging maps with a 55-degree imaging resolution of the right retina from a (A): 51-year-old female PAH subject; (B) 36-year-old female PAH subject; (C) 72-year-old female PAH subject. First Column: Color fundus (CF) output image. Second column: Overlap of arteries (red) and veins (blue) generated manually. Third and fourth columns: Branching generation of arteries and veins, respectively, generated by VESGEN output. Legend (center) identifies branching generations 1–9.

Figure 2

Fluorescein angiography and color fundus comparisons of vascular tortuosity. Manual segmentations of fluorescein angiography (FA) binary images (A,B) color fundus (CF) prepared manually from the retina of a 45-year-old female with PAH. Right and left eyes were split on Bland–Altman plot where FA VESGEN output was compared to CF VESGEN output for total tortuosity, showing a scatter of points suggesting similar results between imaging methods (C), and total artery tortuosity showing scatter of points suggesting similar results between image methods (D). The mean of the measures (x axis) and the difference of the measures (y axis). The reference line of no difference at 0 (solid black), a reference line at the mean of the difference in solid red (observed actual mean), a reference line at +/−2SD of the mean of the difference in dashed red, and a reference line at +/−3SD of the mean of the differences in dashed green (C,D).

Figure 3

Fluorescein angiography and color fundus comparisons in vessel area density. (A) Left and right eyes were split on Bland–Altman plot where fluorescein angiography (FA) VESGEN output was compared to color fundus (CF) VESGEN output for total vessel area density. Scattered points for both left and right eye data suggests no difference between imaging methods. (B) Total artery vessel area density with scattered points suggests no difference between imaging methods with points gather near mean. Plots show similarities between FA and CF imaging among total Av and total artery Av suggesting that CF can be used as a safer method for vascular change monitoring. The mean of the measures (x axis) by the difference of the measures (y axis). The reference line of no difference at 0 is the solid black, a reference line at the mean of the difference in solid red (observed actual mean), a reference line at +/−2SD of the mean of the difference in dashed red, and a reference line at +/−3SD of the mean of the differences in dashed green.

Figure 4

Fluorescein angiography and color fundus comparisons in microvessel number. Left and right eyes were split on Bland–Altman plot where fluorescein angiography (FA) VESGEN output was compared color fundus (CF) VESGEN output for total number of microvessels (A) and total number of microvessel of the arteries (B). Plots show differences between FA and CF imaging in microvessels of total vessels due to the linear scattering of the points. The mean of the measures (x axis) and the difference of the measures (y axis). The reference line of no difference at 0 (solid black), a reference line at the mean of the difference in solid red (observed actual mean), a reference line at +/−2SD of the mean of the difference in dashed red, and a reference line at +/−3SD of the mean of the differences in dashed green.

Figure 5

Binary images of manual and deep learning. CF binary images prepared (A) manually and by (B) deep processing using the retina of a 45-year-old female with PAH. CF binary images of the artery prepared (C) manually and by (D) deep processing using the retina of a 45-year-old female with PAH.

Figure 6

Manuel and deep learning comparisons. Right and left eyes were split on Bland–Altman plot where color fundus (CF) VESGEN output was compared deep learning VESGEN output for total tortuosity (A), arterial tortuosity (B), total fractal dimension (C), arterial fractal dimension (D), total vessel area density (E), arterial vessel area density (F), total number of vessels (G) and arterial number of vessels (H). Plots show difference between FA and CF imaging among total vessels (Tv) (A), total tortuosity (C), and total number of vessels due to the liner scattering of the points. Plots show similarities between FA and CF imaging among total Df. This suggesting CF can be used as a safer method for vascular change monitoring. The mean of the measures (x axis) by the difference of the measures (y axis). The reference line of no difference at 0 is the solid black, a reference line at the mean of the difference in solid red (observed actual mean), a reference line at +/−2SD of the mean of the difference in dashed red, and a reference line at +/−3SD of the mean of the differences in dashed green.