Longitudinal Evaluation of Retinal Capillaries in Experimental Glaucoma



Journal Title

Journal ISSN

Volume Title



Purpose: Glaucoma is a leading cause of irreversible blindness worldwide. The loss of radial peripapillary capillaries (RPCs) has been hypothesized to play a role in retinal ganglion cell axon degeneration in glaucoma. Previous histological work has shown that structural alterations and losses of RPCs are associated with glaucomatous pathology. However, the lack of longitudinal in vivo data characterizing RPC alterations in glaucomatous eyes has limited understanding of how changes in RPCs relate with axonal changes during disease progression. The experiments described in this work were designed to (1) develop robust methods of imaging and analyzing RPCs in vivo and (2) investigate how changes in RPC perfusion relate to changes in the retinal nerve fiber layer (RNFL) over time in a non-human primate (NHP) model of experimental glaucoma (EG). Methods: First, the longitudinal repeatability of quantifying manually segmented RPCs acquired using adaptive optics scanning laser ophthalmoscope (AOSLO) spit detector imaging was evaluated in healthy NHP and human eyes in two or more sessions separated by ≥ 2 weeks. Second, a convolutional neural network (CNN) was developed to automatically segment perfused RPCs in AOSLO images. Third, perfused RPCs were imaged and quantified (using the developed CNN) approximately every 2 weeks after inducing unilateral EG in 9 rhesus monkeys. The RNFL and optic nerve head structure were imaged using spectral domain optical coherence tomography and quantified at each time point. Results: No significant differences were found in any examined capillary metric between sessions across healthy monkey and human eyes (P>.05; Mann-Whitney). The developed CNN objectively segments perfused capillaries with high accuracy (0.938). Changes in RPC perfusion occurred prior to RNFL thinning in 6 of 9 EG eyes. RNFL thinning occurred prior to a change in RPC perfusion in 2 EG eyes, while one EG eye had simultaneous changes in RNFL thickness and RPC perfusion. Conclusions: AOSLO imaging can be used to provide repeatable measures of perfused RPCs. The longitudinal data suggest that vascular factors may make some eyes more susceptible to axonal loss in early stages of the disease. The methods developed here may be applied to other retinal and optic nerve pathologies.



Glaucoma, Adaptive Optics