Purpose
To estimate capillary perfusion pressure (CPP) and wall shear stress (WSS) in the perifoveal capillaries through advanced computational analysis of AOSLO images from diabetic and non-diabetic individuals.
Methods
AOSLO multiply scattered light imaging of perifoveal capillary networks was performed for eyes of diabetic and non-diabetic subjects. Standard deviation perfusion maps were generated from 50-frame videos (30 frames/sec, 2°×2° area) and montaged to form 5°×5° images. After Frangi filtering, image thresholding and manual editing by trained graders, a binary mask was generated to construct 3-D geometrical models with arteriolar inlets and venular outlets identified by registration to a 100° color fundus photograph. Mean arterial pressure, IOP and the geometry of the network were utilized in a computational model to estimate CPP (with respect to the venular outlet pressure) and WSS. On average, >600,000 data points were sampled across each vascular network.
Results
3 non-diabetic eyes (2 subjects) and 2 eyes of 1 subject with mild NPDR (DM duration 19 yrs) had mean±SD age of 33±4 yrs, and 2 were male. Mean estimated CPP was 18.5±11.6 mmHg and WSS was 6.0±5.3 Pa. In the 3 non-diabetic eyes, distribution of data points within the 1st, 2nd, 3rd & 4th quartiles of the average distribution were 21.7, 22.4, 22.2, 33.7% for CPP, and 22.7, 21.9, 27.0, 28.3% for WSS. In the 2 diabetic eyes the values were 24.6, 30.5, 34.0, 10.8%, and 24.4, 30.2, 24.7, 10.7%. On average, CPP and WSS did not differ between the inferior and superior macular quadrants.
Conclusions
Previous studies of CPP and WSS have not been able to assess human retinal capillaries. Our novel approach leverages the attributes of advanced computational modeling with AOSLO to permit estimation of CPP and WSS in these previously inaccessible vessels of the human eye in vivo. These preliminary data suggest that AOSLO-derived computational blood flow models have potential to evaluate human flow dynamics in the smallest retinal vessels of the retina.