imaging of the human retina is a relatively new concept that has the
potential to determine the metabolic status of the retina. Oximetric studies have
been the main focus of previous research as the differential spectral
characteristics of the two functional haemoglobin derivatives may be exploited
to determine the oxygen saturation in the blood vessels.1,2,3
aims to demonstrate the ability to detect oximetric variations in the retinal
circulation amongst normal subjects and in patients with retinal arteriopathy
and glaucoma using hyperspectral imaging and spectral analysis
hyperspectral retinal imaging system consisting of a modified commercial fundus camera,
a liquid crystal tuneable filter and a low-noise CCD detector (figure 1) was
used to capture sequential hyperspectral images of the human retina.
A hyperspectral data cube with a spectral bandwidth of 500nm to 700nm
and a spectral resolution of 10nm at wavelength steps of 2nm were
obtained for each subject. Normal subjects (n = 11) were examined and
compared to subjects with retinal arterial occlusion (n = 3) and advanced
primary open angle glaucoma (POAG)(n = 1).
algorithms were used to dark calibrate and co-register the raw retinal
images. A further image processing algorithm produced a reflectance
optical density map of the retina for each wavelength (figure 2).
spectral unmixing is used in spectral imaging to determine the relative
abundance of materials (endmembers) in each pixel of a scene through the
analysis of its spectral characteristics. An average spectral profile of the
arteries in the optic disc were calculated from all normal eyes (figure 3). This spectral profile
was used to represent a pure endmember of arterial blood. In
addition, a region within the optic disc cup was included into the analysis and
represents a relatively non-oxygenated and spectrally inert endmember. Linear
spectral unmixing, performed in ENVI 4.1 (ITT Visual Information Solutions),
incorporating these two endmembers was used to produce a
qualitative abundance map of oxygenated blood in the retina.
arrangement of the Hyperspectral Retinal Camera.
spectral unmixing produced consistent oximetric maps of the retina in normal subjects
(figure 4) where oxygenated blood (red) has been identified within the
arteries and arterioles. In subjects with arterial occlusions and
advanced POAG, this technique was able to detect
changes in the oximetric status of the retinal circulation. Figure 5 illustrates
one of these subjects with arteritic retinal vasculopathy caused by giant cell
arteritis. The oximetry maps demonstrate an improvement in
oxygenation of the retinal vasculature and retina
following treatment with intravenous methylprednisolone. Figure 6 illustrates
the oximetric variation in the retinal circulation of a patient with advanced POAG which
corresponds with the severity of visual field loss.
Figure 2. Pre-processing
methods of the hyperspectral retinal images.
Figure 3. Linear spectral unmixing
Figure 4. Oximetric
retinal maps from a selection of normal subjects. Oxygenated blood (red) has
been identified in the retinal arteries and arterioles. Venous
blood, containing less oxygenated haemoglobin resulting in a different
spectral characteristic to oxygenated arterial blood, is shown
Figure 5. Oximetric maps of an 81
year old female with a right retinal vasculitic arteriopathy secondary to
giant cell arteritis.
(Left) Oximetric map of the
right retina at presentation.Linear
spectral unmixing revealsa reduction
in oxygenated blood in the retinal arteries (pale red/white) and an
increase in oxygenated blood in the superior retinal veins. This indicates a
reduced metabolic activity in the retina consistent a visual acuity of
perceiving hand movements.
(Middle) Colour photograph of the
right retina 3 days after presentation whilst on intravenous
(Right) Oximetric map of the
right retina 8 days after presentation. Linear spectral unmixing reveals an
improvement in the amount of oxygenated blood within the retinal
arteries associated with a change in the oxygenation of the blood in the
veins. This indicates an improved metabolic activity in the
retina which corresponds with an improvement in the visual acuity to
Hyperspectral imaging is
capable of detecting oximetric changes in the retina and monitoring its
response to treatment.However, the
sequential technique of capturing retinal images described here is
heavily dependent on other factors such as accurate co-registration of the images
This limitation will be addressed in poster 2582/B659 where we will describe the
development of a “snapshot” spectral retinal camera.4
Linear spectral unmixing
offers a powerful and visually useful method of producing
semi-quantitative oximetric maps of the retina, but to increase the effectiveness in
detecting changes caused by diabetic retinopathy and early glaucoma increased and
absolute accuracy is required. This requires the incorporation of a
physical model for light propagation in the retina into the calculation of
oxygenation – as is described in poster 2581/B658. 5
1Ophthalmology Department, Cheltenham General Hospital,
Cheltenham. United Kingdom.
2School of Engineering and Physical Sciences, Heriot-Watt
University, Edinburgh. United Kingdom.
Superior temporal artery
Superior temporal vein
Inferior temporal artery
Inferior temporal vein
1. Schweitzer D, Hammer M,
Kraft J, Thamm E, Konigsdorffer E, Strobel J. In vivo measurement of the
oxygen saturation of retinal vessels in healthy volunteers. IEEE Tram
Biomed Eng. 1999; 46:1454-1465.
2. Khoobehi B, Beach JM,
Kawano H. Hyperspectral imaging for measurement of oxygen saturation in the
optic nerve head. Invest Ophthalmol Vis Sci. 2004; 45(5):1464-72.
3. Lawlor J,
Fletcher-Holmes DW, Harvey AR and McNaught AI. In vivo hyperspectral imaging
of human retina and optic disc. ARVO Annual Meeting Fort Lauderdale, Florida 2002.
4. G.Muyo, A.Gorman, I.Al
Abboud, D.J. Mordant, A.I. McNaught, A.R. Harvey.En Face Snapshot Spectral Imaging of the Retina. ARVO
Annual Meeting Fort Lauderdale, Florida 2007.Tuesday, May 8 2007 8:30 AM - 10:15 AM. (2582/B659).
5. I.Alabboud, III,
A.McNaught, D.Mordant, A.R. Harvey. Quantitative Spectral Imaging of the
Retina. ARVO Annual Meeting Fort Lauderdale, Florida 2007. Tuesday, May 8 2007 8:30
AM - 10:15 AM. (2581/B658).
(Left) Colour photographs of
the optic discs showing advanced cupping (cup-disc ratios 0.9).
(Middle) Humphrey visual field
examination demonstrates asymmetrical visual field loss affecting the right
eye more than the left eye.
(Right) Oximetric maps of the
retina shows an increased oxygenation in the venous circulation of the right retina compared to the
left retina. This may suggest a reduced metabolic activity in the right eye.
Left eye: Snellen acuity 6/9, intraocular pressure 12.
Right eye: Snellen acuity 6/9, intraocular pressure 13.