Fig. 8. Cystoid macular edema appears as late staining of fovea with pooling of the dye into parafoveal cyst-like spaces in a petalloid pattern.

Ozdek and colleagues attempted to correlate FA patterns of diabetic macular edema with described OCT patterns (7). Mean foveal thickness as determined by OCT was the least in the no leakage group and progressively increased in order for focal, diffuse, and combined leakage groups. In addition, 63.3% of eyes that showed evidence of cystoid macular edema by OCT were not detected by FA. OCT also showed serous retinal detachment in 9.7% of eyes, none of which were detected by FA. Kang et al. showed that focal leakage correlated closely with homogeneous focal thickening on OCT (29). Focal leakage showed the least foveal thickness and the best visual acuity among FA types. The proportion of focal leakage type decreased as diabetic retinopathy progressed. Diffuse or cystoid leakage correlated closely with outer retinal layer or subretinal fluid accumulation.

OPTICAL COHERENCE TOMOGRAPHY

Low-Coherence Interferometry

Since its first clinical application in 1991 (30), OCT has dramatically increased our understanding of the morphological changes associated with many macular diseases, including diabetic macular edema (31). Using noncontact, noninvasive scanning, OCT produces high-resolution two-dimensional cross-sectional images of ocular tissues (32–34). Analogous to B-scan ultrasonography which uses sound echoes, OCT is based on reflections of light from the retinal tissue to produce a cross-sectional image. By using light instead of sound, OCT offers considerably higher axial resolution and faster acquisition times. When light is directed into the eye, it is reflected at the boundaries of tissues with different optical properties, as well as being scattered and absorbed by the ocular tissue. Low-coherence interferometry is used to measure the time-of-flight delay of light reflected from structures within the retina.

Fig. 9. Schematic diagram of a classic optical coherence tomography system. An interferometer splits the light source into a probe beam and a reference beam. The probe beam is reflected by retinal structures whose echoes result in a signal.

Current time-domain based OCT scanners use an infrared 200 W, 830 nm wavelength probe light from a continuous-wave superluminescent diode source that is coupled into a fiberoptic Michelson interferometer (Fig. 9). The interferometer splits the light source into a probe beam and a reference beam. The probe beam is directed into the eye and is reflected from retinal structures at different distances. The reflected probe beam is composed of multiple echoes that give information about the distance and thickness of the retinal structures. The reference beam is projected at a known distance. In order for the reference beam and the backscattered light of the probe beam to combine at a detector, the reference beam must be altered. The amount that the reference beam is altered compared to its baseline results in a signal. Software manipulations of the raw OCT image data produce a false-color map representing three-dimensional topographic retinal features and quantitative retinal thickness measurements. The images are stored on digital media to enable comparison of serial evaluations and for archiving purposes.

OCT Image Interpretation

The resulting OCT image closely approximates the histologic appearance of the retina (Fig. 10). The top of the image corresponds to the vitreous cavity. In a normal patient, this will be optically silent, or may show the posterior hyaloidal face in an eye with a posterior vitreous detachment. The posterior vitreous face appears as a thin horizontal or oblique greenish line above or inserting into the retina. The anterior surface of the retina demonstrates high reflectivity, and in the fovea of normal eyes, demonstrates the central foveal depression. The horizontally aligned nerve fiber layer demonstrates higher tissue signal strength and is thicker closer to the optic nerve. The internal structure of the retina consists of heterogeneous reflections, corresponding to the varying

Fig. 10. OCT of a normal human macula showing the characteristic foveal contour. The nerve fiber layer (NFL) is highly backscattering. The inner and outer plexiform layers (IPL, OPL) are more hyperreflective than the inner and outer nuclear layers (INL, ONL). There is a reflection from the boundary between the inner and outer segments of the photoreceptors (IS, OS). The RPE and choriocapillaris appear as the highly backscattering boundary beyond the posterior retina.

ultrastructural anatomy. The axially aligned cellular layers of the retina (inner nuclear, outer nuclear, and ganglion cell layers) demonstrate less backscattering and back-reflection of incident OCT light, and thus appear with a lower tissue signal (darker), compared to horizontally aligned structures (internal limiting membrane, Henle’s layer, and NFL) that appear brighter. The retinal pigment epithelium, Bruch’s membrane, and choriocapillaris complex collectively comprise the highly reflective external band. Just anterior to this band is another highly reflective line representing the junction between the photoreceptors’ inner and outer segments. Reproducible patterns of retinal morphology seen by OCT have been shown to correspond to the location of retinal layers seen on light microscopic overlays in both normal and pathologic retinas (17, 35–37, 40).

Image-processing software can quantify retinal thickness from the OCT tomograms as the distance between the anterior and posterior highly reflective boundaries of the retina (38). A software algorithm known as segmentation uses the processes of smoothing, edge detection, and error correction to facilitate this process. Retinal thickness can therefore be determined at any transverse location. Hee et al. (33) developed a standardized mapping OCT protocol, consisting of six radial tomograms, each 6 mm in length, in a spoke pattern centered on the fovea. Retinal thickness is then displayed in two different manners: first as a two-dimensional color-coded map of retinal thickness in the posterior pole; and secondly as a numeric average of nine parafoveal areas corresponding the ETDRS subfields. Additional acquisition algorithms include the fast macular mapping protocol, which allows six radial scans to be performed in a single session of 1.92 s; and the high-density scan protocol consisting of six separate 6-mm radial lines, acquired in 7.32 s.