Hyperspectral Imaging of the Retina

Balancing between image resolution and spectral accuracy

Dirk De Brouwere and Chris Dainty

Motivation

The clear optical media in the eye give clinicians a unique opportunity to obtain high quality images of the (micro)-vasculature of the retina and other pathologic structures using non-invasive imaging techniques.

For many years, researchers have been investigating the spectral reflectivity of the retinal tissue in fundus images. The aim of this study is to gain further knowledge about the composition of the tissues in the retina. The most popular area of interest in spectral imaging is retinal oximetry, but other fields such as evaluation of macular pigment epithelium (MPE) or cytochrome have been investigated using this technique.

Until recently, spectral retinal imaging has yielded to a trade off between accurate spectral information in a large area of the retina and a high-resolution image of the retina as seen under a limited number of spectral bands. The goal of this project is to find a balance between both approaches using currently available technologies.

Instrumentation

We have attached a line spectrograph (Specim V2/3”) on the image plane of a standard fundus camera (Zeiss FF450+). The spectrograph separates the spectrum in one direction on a sensitive camera, whereas the image of the selected line is conserved in the perpendicular direction. After correction for some optical effects, we obtain a “hyperspectral slice” of our retinal image. To enhance the morphologic information of the spectrogram, right before image acquisition, a retinal (broadband) image is acquired on a camera placed in another exit pupil of the fundus camera. Despite minor eye movements, image registration of the retinal image and the spectral image gives us a high resolution spectrum on each pixel of the slit.

Modified Zeiss fundus camera for registration of spatio-spectral imaging

In other words, this configuration is useful as it balances a high spectral resolution (3nm) with accurate spatial image information. This allows us to study spectral absorption of various tissues and microstructures in the retina.
Another benefit of the technique is that, given the large spectral information, our setup can compare the efficiency of several multi-band oximetry technologies that have been proposed by several authors.

Principles of the line spectrograph.

In a parallel study, we have developed a custom made fundus camera for the purpose of spectral imaging in animal models. It employs a bright white LED light source to illuminate approx 20degrees on the retina through a maxwellian viewing system. Imaging is done through a (typically) 2mm pupil. A reflex system separates the spectral image from the retinal image. A sketch of the setup is illustrated in figure…

Custom designed retinal spectrograph focussed on in vivo research in wetlabs

Data analysis

 Image registration of a line spectrum on the retinal image                Spectral outcomes of custom fundus camera (albino rabbit)

Several spectral algorithms that have been proposed to unmix are implemented in a user friendly labview environment (National Instruments). Most algortihms are available in the literature (see references), although some of them are slightly modified in the software. Please click on the links below for a brief description of the algorithms

Relative Multiband Oximetry

Hyperspectral Unmixing of retinal reflectance

References

Francois C. Delori and Kent P. Pflibsen:"Spectral reflectance of the human ocular fundus",Appl. Opt.,28(1989).


James Beach, Jinfeng Ning, and Bahram Khoobehi:"Oxygen saturation in optic nerve head structures by hyperspectral image analysis.",Curr Eye Res,32(2007).


Francois C. Delori:"Noninvasive technique for oximetry of blood in retinal vessels",Appl. Opt.,27(1988).


Martin Hammer, Walthard Vilser, Thomas Riemer, and Dietrich Schweitzer:"Retinal vessel oximetry-calibration, compensation for vessel diameter and fundus pigmentation, and reproducibility.",J Biomed Opt,13(2008).


Sveinn Hakon Hardarson, Alon Harris, Robert Arnar Karlsson, Gisli Hreinn Halldorsson, Larry Kagemann, Ehud Rechtman, Gunnar Mar Zoega, Thor Eysteinsson, Jon Atli Benediktsson, Adalbjorn Thorsteinsson, Peter Koch Jensen, James Beach, and Einar Stefansson:"Automatic retinal oximetry.",Invest Ophthalmol Vis Sci,47(2006).


Bahram Khoobehi, James M Beach, and Hiroyuki Kawano:"Hyperspectral imaging for measurement of oxygen saturation in the optic nerve head.",Invest Ophthalmol Vis Sci,45(2004).


Vasile Diaconu:"Multichannel spectroreflectometry: a noninvasive method for assessment of on-line hemoglobin derivatives",,48(2009).

Funding

This work is funded by IRCSET Embark Initiative and Science Foundation Ireland

Interested?

Send us an email!