The Next Step in OCT Technology
Spectral domain provides unprecedented views of the retina.
BY PETER K. KAISER, MD
It is an exciting time for retinal imaging. Optical coherence tomography (OCT), a technology that has become an invaluable part of practice, is undergoing a transformation. The newest instruments employ spectral domain technology, a more powerful way to create images than the current gold standard time domain technology.
Higher image resolution and acquisition speeds are what sets the new instruments apart from their predecessors. In the following text, I will explain how that higher resolution and speed is accomplished and illustrate their potential benefits.
ATTAINING HIGHER RESOLUTION
The overall resolution of an OCT instrument is determined by two independent directions of resolution: transverse and axial. Transverse resolution is based on the spacing of the A-scans performed and is limited by the optics of the eye. Therefore, it cannot be improved upon. In contrast, axial resolution depends on the wavelength and bandwidth of the instrument's light source, which can be improved. Both spectral domain and time domain OCT systems use superluminescent diode light sources, but those in the spectral domain systems have a slightly broader bandwidth, improving axial resolution over time domain. Research facilities use systems with even higher resolution because they have employed expensive titanium sapphire lasers as the light source.
Fig. 1. Heidelberg Noise Reduction software produces high quality images for easier identification of retinal structures and improved segmentation.
Increasing the speed of an OCT system is another way to increase its resolution. To understand how spectral domain and time domain systems differ in this regard, it helps to review how OCT works. First, the light source is split 50/50 by a beam splitter. Half of the light goes into the sample, ie, what is being imaged. The other half goes into a reference arm, typically a mirror. The two streams of light are then reflected back. The time it takes each stream of light to return is measured and compared to the reference light to create a reflectivity profile and resulting image of the sample.
In time domain systems, for this process to work, the reference mirror has to move back and forth with sequential scans. Because it depends on this mechanical moving part to perform its A-scans, time domain is a slower imaging modality. Approximately 400 A-scans per second is the maximum that can be achieved reliably. Because patient eye motion is also occurring, it is not feasible to use time domain OCT to precisely map retinal tissue.
In contrast, in spectral domain OCT, the reference arm does not move. Instead, when the light is reflected back, the entire signal (at all wavelengths) is recorded in parallel by a spectrometer. All of the wavelengths are then converted by Fourier transform into time delay signals to produce the image. Because all of the echoes are measured simultaneously as opposed to sequentially with the reference mirror, the process is 50 to 100 times faster than time domain OCT. When the speed of OCT is increased, motion artifacts are reduced and digital processing is not required to align adjacent A-scans, resulting in more accurate retina scans.
To further enhance resolution, multiple spectral domain scans of the same location can be combined to reduce noise. Noise is inevitable in image capture, but when multiple scans are taken, the instrument can identify differences between them as noise and remove it as shown in Figure 1 on the previous page. The result is a much sharper image. In general, spectral domain systems allow better visualization of the retinal structures (Figures 2–6).
|New OCT Instruments Have Varying Capabilities|
|Several companies now offer, or will soon offer, spectral domain OCT instruments. They all have B-scan and 3-D capabilities and similar axial resolution, but we can point to differences among them as well.|
Scanning speed, which has an impact on motion artifacts and scan accuracy, varies among the available instruments. At 40,000 A-scans per second, the Spectralis HRA+OCT from Heidelberg Engineering is the fastest. Also, some of the instruments are standalone OCT devices, while others are combined with other imaging modalities. For example, the 3D OCT-1000 from Topcon includes a color fundus camera, and the Spectral OCT/SLO from OTI can be used to perform microperimetry. In addition to spectral domain OCT, the Spectralis HRA+OCT can also be used for fluorescein angiography, indocyanine green angiography and autofluorescence.
Figs. 2 and 3. Spectral domain OCT allows better delineation and visualization of retinal structures.
IMPROVED RETINAL BOUNDARY DELINEATION
Since the data are obtained so slowly in time domain, motion artifacts often occur. To compensate for this motion, digital processing, i.e., interpolated data, must be used with time domain OCT. To image the macula, 6 scans of the macula are obtained, and the rest of the retinal thickness is interpolated from these 6 scans. These 6 radial line scans measure less than 5% of the mapped area. Therefore, errors can be propagated over a large area, and small abnormalities between scan lines may not be detected. Moreover, the software assumes the radial line scans are straight and intersecting at the center point. It relies on this assumption to reconstruct the data. However, as the eye moves, so does the scan location. As a result, the clinician has no way of knowing if he or she is ever imaging the same spot on the retina (Figure 6).
Fi. 4. Time domain OCT provides no useful information for making a treatment decision for this patient with diabetes. However, the spectral domain image taken minutes later shows an epiretinal membrane, which may be cause for surgery.
Fig. 5. Time domain shows little if any subretinal fluid in this patient with AMD, but the spectral domain image reveals disease activity that may warrant treatment.
Fig. 6. Example of motion artifacts seen on a time domain OCT system.
Fig. 7. Spectral domain OCT generates a more accurate image because its high speed ensures dense coverage of the retina.
In contrast, spectral domain systems do not need to interpolate data points because their high speed provides a much greater coverage of the macula (Figure 7). That means, for example, when a clinician is deciding whether a patient needs another age-related macular degeneration (AMD) treatment, he or she will not miss a pocket of fluid because a scan line did not pass through it.
In addition, the improved resolution of the spectral domain OCT can improve the machine's ability to accurately detect retinal boundaries. In general, time domain systems accurately detect the retinal boundaries in eyes with macular edema, but not as consistently in eyes with AMD. Sadda and colleagues showed that retinal thickness boundaries in these cases frequently are identified incorrectly.1 This is especially problematic for clinical trial reading centers and physicians treating glaucoma. With improved resolution, the spectral domain software is more successful at automatically detecting retinal boundaries resulting in more accurate retinal thickness maps.
Fig. 8. The TruTrack feature of the Spectralis HRA+OCT locks scans onto a location of the retina, improving the clinician's ability to track changes over time.
EYE TRACKING AND REGISTRATION IMPROVE PATIENT MONITORING
Time domain images are obtained slowly, so patient movement can occur, leading to motion artifacts. To reduce this problem, the Spectralis uses eye tracking technology. The device takes a continuous reference scan of the retina. When the eye moves during scanning, so does the location where the scan is performed. The scanner takes images only when the reference laser is tracking the retina. This eliminates motion artifact caused by patient movement. The Spectralis can then combine multiple scans taken from the exact same position to eliminate noise. The difference in noise level and improvement in resolution when this eye tracking feature is enabled is notable.
An ongoing challenge with the use of time domain OCT systems is that they do not register images visit to visit. The scans obtained between visits are not registered with time domain systems. Using the eye tracking feature and the reference image from the previous visit, the Spectralis can capture data at the next visit in the exact same position. This precise registration is a boon for reading centers, but also important for any clinician following a treatment over time to see how the OCT changes between visits (Figure 8).
CHANGE PRESENTS CHALLENGES
Along with the benefits of spectral domain OCT come a few challenges that are currently being addressed. The capabilities of these systems are derived from the vast amount of data they capture. The resultant files are very large, which is not a problem for viewing on-screen or with patients in the photography area. However, sending such large files to the clinical exam room or reading station is too much for most practice infrastructures to handle in a timely manner.
Also, the software for the new spectral domain systems is still relatively immature and is continuing to evolve. The lack of a validated normative database is an issue, especially in the management of glaucoma. Finally, clinicians will want to have access to their time domain data after they upgrade to a spectral domain system. How that can be accomplished is still unknown. For example, will every manufacturer's spectral domain unit allow viewing of the time domain data?
LEARNING MORE EACH DAY
As these issues are resolved, there is little doubt that spectral domain OCT is the technology for the future. As clinicians learn more about the significance of its capabilities, they will allow better diagnosis, monitoring and treatment for patients with retinal diseases.
|Peter K. Kaiser, MD, is director of the Digital OCT Reading Center at the Cleveland Clinic Cole Eye Institute. He can be reached at firstname.lastname@example.org or (216) 444-6702.|