Optical Coherence Tomography

**Figure 2:** Schematic image of the eye
[width=7cm]Eye.eps

The goal of this project is to detect glaucoma at early stages of the disease. In a glaucomatous eye the nerve fibres die, which results in a decrease of the retinal nerve fibre layer (RNFL) thickness.

All nerve fibres leave the eye through the blind spot. The majority of them emerge from the yellow spot, see Figure 2. Thus, the thickness distribution of the RNFL on a circle circumnavigating the blind spot is not uniform. The thickness is measured with Optical Coherence Tomography, OCT. The measurement data does not immediately give the thickness. It shows the reflectivity at different depths of the eye. The RNFL has large reflectivity, while the deeper photoreceptor layer has low. Below the photoreceptor layer the reflectivity is again higher. If the photoreceptor layer thickness is assumed to be constant, the RNFL thickness is proportional to the distance between the upper edge of the upper bright ribbon to the upper edge of the lower bright ribbon, see Figure 3.

To measure the RNFL thickness, the edges thus have to be detected. This is achieved in a straightforward manner using convolutions with differentiated Gaussian functions, see Figure 4.

**Figure 3:** The measurement data
[width=4cm]/usr/matematik/kalle/tex/mig_verk02/figs/OCTbildfshow.ps

**Figure 4:** Edges detected
[width=4cm]/usr/matematik/kalle/tex/mig_verk02/figs/OCTkantfshow.ps

The RNFL thickness for some healthy eyes can be seen in Figure 5. While the general shape seems to be constant, the thickness decreases with age, so the mean value of each curve is subtracted to compensate for age effects, see Figure 6.

**Figure 5:** Thickness of fibre layer
[width=3cm]/usr/matematik/kalle/tex/mig_verk02/figs/Kurvor.ps

**Figure 6:** Age compensated thickness
[width=3cm]/usr/matematik/kalle/tex/mig_verk02/figs/Kurvorn.ps

The work aims at improving present clinical routines for detecting glaucoma using these measurements. Usually, the RNFL thickness of a glaucomatous eye decreases first in the regions where it is thick. In Figure 7 and 8 a comparison between a healthy and glaucoma eye with the mean value of the healthy eyes above is depicted.

**Figure 7:** Healthy eye
[width=3cm]/usr/matematik/kalle/tex/mig_verk02/figs/Friskomv.ps

**Figure 8:** Glaucomatous eye
[width=3cm]/usr/matematik/kalle/tex/mig_verk02/figs/Sjukomv.ps

To use OCT measurements for glaucoma diagnostics, it would be of medical importance to develop models for the thickness of the RNFL, and in particular its dependence on the density of the ganglion cells, from which the nerve fibres emerge. A couple of models for both healthy and glaucomatous RNFL have been developed, and further investigations and refinements of these are still going on.

Another way of drawing information from OCT measurements of the RNFL thickness is to use machine learning techniques for determining whether a thickness profile comes from a healthy or a glaucomatous eye. Some preliminary investigations on small datasets show promising results.