Snaphot Retinal Oximetry: Improved Vessel Detection and Multispectral Oximetry in Normal Subjects

Abstract

Purpose: : To report a hyperspectral approach to retinal vessel oximetry in humans.

Methods: : We use a novel hyperspectral camera that can be attached to the port of a standard fundus camera. This approach utlizes the computerized tomographic imaging spectroscopy ( CTIS) technology, which permits the simultaneous recording of more than 50 wavelengths of spectral data from standard fundus images obtained in less than 3 milliseconds. This approach does not require multiple image acquisitions at different wavelengths and hence there is no need for image registration. We developed an unsupervised software to extract the vessels from the background. Automated calculation is performed to extract the apparent transmittance and the optical density for each pixel on the vessel. A least square method is applied, by using 28 wavelengths, to calculate the oxygen saturation. The results are then converted into a pseudo-color chromatic map of relative oxygen saturation in the retinal vessels. In addition we devised an error index which represents an objective index to measure the performance of the mathematical measurements and the wavelengths used in all vessels. This error index is obtained for each individual patient.

Results: : Thirteen normal subjects were studied using this approach. The software was able to accurately identify the vessels in all subjects. The oxygen saturation maps showed a relative difference between arteries and veins of about 30% in all subjects, with accurate classification of first and second order bifurcations of arteries and veins in all subjects. The error index showed an average error of 4.7% in all images.

Conclusions: : Our results demonstrate improved software algorithms for vascular detection using an unsupervised approach. The retinal vascular oximetry measured by this approach in this group of normal subjects shows correlation with known physiologic differences. The user-friendly software module shows great promise for clinical use. Further studies are underway in patients with retinal vascular disease.

Keywords: imaging/image analysis: clinical • retina • blood supply