Purchase this article with an account.
D. Izhaky, Y. Kaminitz, A. Grinvald, D. A. Nelson; A New Model for Correcting Retinal Oximetry Values Using the Retinal Function Imager. Invest. Ophthalmol. Vis. Sci. 2009;50(13):1402.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Determining retinal vessel oxygen saturation is desirable to assess retinal function. However, apparent optical density (OD) values of retinal vessels we and others obtain from multi-wavelength fundus camera images are substantially lower than single-pass Beer-Lambert expectations. An additive ("offset") diffuse media reflection, reported, but to our knowledge uncharacterized in such images, would account for this, if large. We measured typical offset values in human subjects, constructed a model for estimating their effects, and evaluated methods for correcting them.
Retinal image data sets were acquired at three wavelengths (569, 575, and 586 nm) using the Retinal Function Imager (RFI) with healthy volunteers (N=5). We found the vessel/background pixel value ratio under different conditions of illumination, and calculated the offset necessary to meet physical constraints like feature contrast invariance and constant OD ratio at isobestic wavelengths. Apparent OD was corrected for individual vessels (also avoiding specular reflections), and oxygen saturation levels calculated.
Offset values were consistent among methods of determination, and commonly approached 50% of background extravascular pixel values. In a single-pass Beer-Lambert model, offset-induced errors are most sensitive for thick vessels and high absorption coefficients, and normally aggravated when the fractional offset size changes with wavelength. Pre-correction, oxygen saturation levels were skewed to highly variable and unrealistically "supersaturated" values. Offset corrected mean oxygen saturation levels were 92% ± 4% for arteries and 56% ± 7% for veins.
Diffuse media reflection, often neglected, is a large light contributor in fundus camera images. Measuring it to correct the calculations brings oxygen saturation estimates to physiologically reasonable values. Among its advantages, this correction may be needed for saturation estimates in retinal areas where hemoglobin is not the dominant chromophore.
This PDF is available to Subscribers Only