Algorithms and instrumentation for quantified retinal oximetry in a snapshot

Abstract

Abstract Purpose To develop an instrument and techniques for useful absolute and relative clinical oximetry of the retina. Methods A novel snapshot multispectral imaging system has been optimised for retinal oximetry. Eight monochromatic images of the retina are recorded with a field of view of 24 degrees. Algorithms calculate the absorption of light at each point along delineated blood vessels and for each of the eight wavebands. Data inversion using an analytical model for light propagation enables oximetry at along the blood vessels. Spectral inversion has been refined using ray tracing and Monte Carlo modelling of light propagation using a realistic phantom eye. Results A comparison of Monte Carlo modelling of light propagation in the phantom retina with recorded images for blood of various oxygenations indicates the influence of several unknowns, including scatter from optical surfaces within the ophthalmoscope and eye, the geometry of the blood vessels and eye, the optical constants of the ocular media and the complexity of light propagation in the retinal structure. Relative oximetry within the retina is possible with repeatability of about 1% in the phantom and 5% in a real eye but the influence of various systematic effects can introduce systematic differences between actual and calculated oxygenation that can significantly exceed these values. Conclusion Although oximetry using only a very small number of spectral bands is possible, this is prone to systematic errors that can effect both absolute and relative oximetry. The ability to record eight spectral images in a single snapshot offers promise to provide an enhanced clinically useful and validated oximetry technique.