Snapshot Hyperspectral Imaging: Retinal Oxygen Saturation Analysis From Hyperspectral Images of Rabbit Retina

Abstract

Purpose: : To introduce a 28 wavelength oximetry model for imaging the rabbit retina with computerized tomographic imaging spectroscopy (CTIS).

Methods: : We use a novel hyperspectral camera that is attached to a standard fundus camera and permits simultaneous recording of more than 50 wavelengths of spectral data from standard fundus images. Images were acquired from 5 albino and 5 pigmented rabbits. 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 algorithm to extract vessels from background. Automated calculation is performed to extract the transmittance and the optical density for each pixel in vessels. A least squares method is applied to calculate the oxygen saturation. The results are displayed as a pseudocolor map of relative oxygen saturation in the retinal vessels.

Results: : Our automatic vessel detection algorithm successfully isolated rabbit retinal arteries and veins. Our automated oximetry model shows relative differences in arterio-venous oxygen saturation. Image 1 and 2 (544nm and 592nm respectively) show excellent spatial resolution of arteries/veins and gross resolution of choroidal vessels. Image 3 is a pseudocolor oximetry map.

Conclusions: : Our results demonstrate the successful application of our software algorithms for vessel extraction using an unsupervised approach in the rabbit retina. Both retinal vessels as well as choroidal vessels can be demonstrated on raw images from the spectral data set. Preliminary retinal oximetry shows qualitative differences in retinal arterio-venous oxygen saturation. Future experiments will be aimed at quantitative calibration of oximetry values from retinal vessels as well as improving aquisition parameters for demonstration of choroidal vessel.

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