Validation of Quantitative Retinal Vessel Oximetry Using an Eye-Phantom and Monte-Carlo Ray-Tracing

Abstract

Purpose: : Spectral imaging can be used together with physical models and computational algorithms to determine oxygen saturation in the retinal vasculature. There is however difficulty in validating in-vivo measurements.This study aims to validate the results of retinal vessel oximetry using quantitative analysis of spectral-images of an eye-phantom together with Monte-Carlo modelling of light propagation.

Methods: : Spectral images of an eye phantom were acquired using conventional fundus cameras modified to enable two spectral imaging techniques to be implemented; high resolution time-sequential imaging and a lower resolution snapshot imaging. The eye-phantom incorporates the most important components determining retinal image signatures: namely physically realistic artificial blood vessels located close to a simulated retinal pigment epithelium and choroid. Control of retinal parameters, allows their effect to be assessed and enables in-depth understanding of light paths in the retina to be established for the first time. Images recorded with the eye-phantom were modelled using Monte-Carlo ray-tracing and compared with a range of measurements recorded with the eye phantom and also with in-vivo measurements. An algorithm based on a physical model of light propagation was used to determine oxygen saturation from the spectral images.

Results: : Quantitative oximetry of blood within artificial vessels in an eye phantom has achieved with an accuracy of 8% in venules and 5% in arteriols. Profiles across blood vessels predicted by Monte Carlo modelling is in good agreement with those recorded in the eye phantom and in vivo. The wavelength dependence of scattering within the blood column determined from the data are in good agreement, but significant deviations between conventional models of light propagation in the eye are have been measured indicating the need for enhanced rigour in the design of oximetry algorithms.

Conclusions: : The ability to change the structure and composition of the fundus allows validation of oxygen saturation retrieval algorithms for a wide range of retinal characteristics. Monte-Carlo ray-tracing enables enhanced understanding of light propagation in the retina with consequent improvements in the reliability of oximetry algorithms.

Keywords: retina • oxygen • imaging/image analysis: clinical