Advancing OCT technology yields high-resolution retinal images

Fast, precise spectral optical coherence tomography enables physicians to distinguish layers of the retina and monitor pathologies.

Recent developments in optical coherence tomography promise to enhance the treatment of retinal disease by rapidly generating high-resolution retinal images. Recent imaging innovations include three-dimensional spectral domain OCT, which offers speed and enhanced resolution without diminishing image quality, some experts say.

“The current state of the art is spectral domain OCT, which has been an extraordinary advance for clinical practice,” Richard Rosen, MD, said. “These systems provide the ability to appreciate microscopic pathology of each of the various retinal layers on a daily basis in the clinic. They challenge us to examine OCTs much more closely than we did a couple years ago because you can really distinguish all of the layers of the retina in a good system.”

There are seven commercially available spectral domain OCT devices: Cirrus HD-OCT (Carl Zeiss Meditec), Spectral OCT/SLO (Opko/OTI), Spectralis HRA+OCT (Heidelberg Engineering), RTVue-100 (Optovue), Copernicus (Optopol/Reichert), SD OCT (Bioptigen) and 3D-OCT 1000 (Topcon).

Ophthalmologists must become acclimated to these spectral domain OCT systems and the detailed images they create, Dr. Rosen said.

“[With] the spectral systems that are currently on the market, there’s a wealth of information acquired,” he said. “It’s imperative for clinicians to get comfortable with reading the images and start to use the information so they can distinguish which layers are involved in a particular disease and hopefully come to understand it a little bit better.”

Scanning speed is essential for getting high-quality three-dimensional images, Dr. Rosen said.

Current spectral systems typically run 25,000 to 40,000 A-scans per second. The next wave of spectral OCT devices will likely generate 250,000 to 500,000 scans per second, he said.

“That sounds pretty incredible, but you need that speed in order to acquire the data fast enough to eliminate movement, which will blur fine details,” Dr. Rosen said. “In order to get good 3-D images with quality comparable to the regular B-scan slices, you have to scan very rapidly. Currently, there are prototype swept source-based systems that provide such fast scanning. Swept source OCT will likely be the next generation of spectral OCT.”

Joel S. Schuman

Three-dimensional spectral domain OCT affords detailed imaging of the retina that was formerly unattainable, Joel S. Schuman, MD, FACS, said.

Dr. Schuman was senior author of a study on segmentation-free contour modeling C-mode used to enhance three-dimensional OCT. The authors compared their own ultrahigh-resolution OCT prototype and the Cirrus HD-OCT. The study, which is in press, was presented at this year’s annual meeting of the Association for Research in Vision and Ophthalmology.

“In retinal imaging, you can get a very high-density cross-sectional image of the retina,” Dr. Schuman said. “You can see the details of the different retinal layers within the retina and also structures within those layers that we weren’t able to see in the past because the OCTs didn’t have a high enough scan density.”

He said the C-mode three-dimensional OCT software, developed by Hiroshi Ishikawa, MD, and the University of Pittsburgh’s Glaucoma Imaging Group, is versatile.

“It’s not limited to one particular type of OCT,” Dr. Schuman said. “It’s not limited to our prototype or to any particular brand of OCT. It is software that could be applied as long as there is a three-dimensional OCT data set.”

Segmentation-free spectral OCT

Dr. Schuman said spectral domain OCT enhances observation of pathology within the macula.

“Because the scan density is fairly high, it’s unlikely that you’re going to miss pathology within the macula as you acquire your scans,” he said. “Now you can look at this set of cross-sections that has been acquired. You can also re-sample your data for a new cross-section.”

Coronal scans can be cut, providing a direct, head-on view of retinal layers.

“In the direction that we’re sampling (axially), the resolution is 3.5 µm to 10 µm. If we’re cutting coronally, we’re able to find details of the retina that we cannot see in cross-section,” he said.

However, curvature of the eye is often problematic in matching a coronal slice to a given retinal layer, Dr. Schuman said.

“Because segmentation algorithms often do not work perfectly, if you depend on segmentation, you might have a view of a tissue layer that’s also not adequate,” he said. “In other words, you may be simultaneously viewing multiple layers, including layers other than the one that you’re interested in.”

Using a segmentation-free malleable plane allows compensation for curvature of the plane at different points.

“You can mold your sampling plane to the actual curvature or actual shape of the layer that you’re interested in,” he said. “That allows you to essentially lift out that layer of the retina and look at it directly. It’s possible to see pathologies in great detail that we really were not able to see before.”

For example, segmentation-free OCT enables imaging the fine details of epiretinal membranes, macular edema, pigmented epithelial detachment and drusen, Dr. Schuman said.

Thresholding to view retinal layers

Thresholding is currently the most common approach to segmenting retinal layers in 3-D images, Dr. Rosen said.

“I think it’s exciting,” he said. “In the study, [Schuman et al] bring up the point that the current software technology that’s commercially available from Zeiss and most of the other manufacturers create C-scan coronal images using thresholding to separate retinal layers. With thresholding, you can simulate peeling off the different layers of a 3-D reconstruction of the retina.”

However, thresholding may be subject to artifacts, especially in cases involving thick lesions or lesions with high or low reflective properties, Dr. Rosen said.

Visualizing layers in isolation may help clinicians appreciate the extent of a pathological feature such as drusen, a common component of dry macular degeneration, he said.

“When we look with high-resolution OCT, we can see that in many cases, drusen are part of more extensive layers of deposited material,” Dr. Rosen said. “What you see as yellow spots are sometimes just the peaks of those layers. To get a really accurate idea and quantify the extent of pathology, it would be nice to be able to section these layers in a way that’s accurate and repeatable.”

One limitation to OCT is that it isolates a very thin cross-sectional slice of the retina, he said. For example, newer spectral domain OCT slices represent a 6- to 10-µm thick section from a 6,000- to 9,000-µm-square field.

“It’s difficult to be sure that if you’re looking at the same point two times in a row,” he said. “With some of these subtle pathologies, if you move just a little bit, they disappear.”

Localization when imaging fine details is one of the most significant challenges, Dr. Rosen said.

“Localization is key to accurate follow-up of most retinal diseases and evaluation of the success of therapy,” he said. “The newer technologies employ different approaches such as utilizing a scanning laser ophthalmoscope image in conjunction with each OCT slice to provide context of the slice and help ensure localization Stratus OCT (Zeiss) images are accompanied by very low quality infrared video images of the fundus, and you can rarely tell their exact location, let alone re-examine the same spot. … As OCTs reveal smaller, finer details, accurate localization will become an even bigger issue.”– by Matt Hasson

• Richard Rosen, MD, can be reached at New York Ear and Eye Infirmary, 310 E. 14th St., 6th Floor, South Building, New York, NY 10003; 212-979-4181; 212-979-4268; e-mail: rrosen@nyee.edu.
• Joel S. Schuman, MD, FACS, can be reached at University of Pittsburgh Medical Center, Eye Center Department of Ophthalmology, 203 Lothrop St., Suite 816, Pittsburgh, PA 15213; 412-647-2205; 412-647-5119; e-mail: schumanjs@upmc.edu.