Purpose
We hypothesized that site-specific features within the branching generations of arterial and venous trees imaged by 30° Spectralis® fluorescein angiography (FA) can be mapped and quantified by VESsel GENeration Analysis (VESGEN) software, and that this methodology would be useful for longitudinal analysis of diabetic retinopathy (DR) progression.
Methods
The retina of a patient diagnosed with mild nonproliferative DR was photographed at12.5 μm/pixel with 30° Heidelberg Spectralis® imaging following injection of fluorescein (Fig 1). Binary (black/white) vascular patterns of the branching arterial and venous trees were extracted from the photograph (768x768 pixels) as described previously for 50° FA (IOVS 2010, 51:498). The resulting arterial and venous images served as sole inputs to the VESGEN software. The vascular patterns were automatically analyzed to first map vessel branching generations (Gx, Fig 2) and then to quantify the resulting vascular maps by vascular parameters that include the fractal dimension (Df), vessel number (Nv), and densities of vessel length (Lv) and area (Av). Branching generations mapped by VESGEN were further assigned into two groups of large and small vessels.
Results
Overall vascular density by Df was 1.57 for the arterial tree and 1.60 for the venous tree. Nv1-3 was 21 for large arteries and 15 for large veins (Figure 1). For small arteries, Nv≥4 was 161, compared to 234 for small veins. Trends for Df and Nv were confirmed by Av and Lv. Several key parameters revealed that the density of small veins was greater than that of small arteries.
Conclusions
Our study demonstrates the feasibility of VESGEN analysis for mapping and quantifying the remodeling of arterial and venous trees during progressive DR from clinical images obtained by 30° Spectralis® fluorescein angiography. This methodology will enable evidence-based conclusions for ongoing longitudinal studies on how and where site-specific remodeling occurs and progresses within retinal vasculature.
Keywords: 688 retina •
499 diabetic retinopathy •
550 imaging/image analysis: clinical