Tomography is imaging by sectioning or "slicing".  In ophthalmology, an optical coherence tomographer is used to image either the anterior segment (cornea and anterior chamber) or the posterior segment (vitreous and retina).  At present, because of focusing limitations, most instruments cannot effectively image both anterior and posterior chambers.  This module discusses only posterior chamber imaging instruments.

Above left:  anterior chamber OCT                  Above right:  posterior chamber OCT

Both images captured with OPKO/OTI OCT instruments.

Time Domain and Spectral Domain OCT

Time domain technology was the first wave of OCT technology.  It is represented in clinical use by the Zeiss Stratus OCT 3.  The second wave of ophthalmic OCT technology is called "spectral domain" or "Fourier domain".  The time domain OCT market is dominated by the proprietary design of the Carl Zeiss Company.  Spectral domain technology has been developed by many companies, with many spectral domain OCT instruments coming on the market at about the same time.  Simply  put, spectral domain technology has a much faster scanning rate, and has better resolving power than time domain technology. 

In order to understand the advantages of spectral domain technology, it is useful to understand the shortcomings of the Zeiss Stratus OCT 3.  Despite the advantages of spectral domain technology, the Zeiss OCT 3 will continue to be a valuable clinical tool for years to come, particularly since thousands of them are currently in use.  For more detailed information regarding the Zeiss Stratus OCT 3, see Modules 37 and 38.

The Zeiss Time Domain OCT

Optical coherence tomography was first developed in 1991 by Swanson and Huang et al.  Time domain OCT has been commercially available as a proprietary instrument made by the Carl Zeiss company.  The Zeiss time domain OCT received FDA approval in 2002 and it has evolved through three model changes, culminating in the Zeiss Stratus OCT 3, with the 6000th unit having been installed in 2007.

The OCT uses an interferometer that measures the time it takes for light to be reflected back from structures in the retina, as compared to the time it takes for light to be reflected back from a reference mirror at specific distances.  The process is similar to that of ultrasonography, except that light is used instead of sound waves.

The Zeiss OCT 3 can make from 128 to 768 axial samples (A-scans) in a single "scan pass".  Each A-scan has 1024 data points and is 2mm long (deep).  The following illustration of a scan pass is an animation that is best viewed on a computer screen.  An OCT "scan pass" is sometimes referred to as a "B-scan", which is similar in concept to the ultrasound B-scan.  The A-scans are stacked horizontally or vertically to create a two dimensional image of the area being scanned.

OCT A-scans and B-scans vs. Ultrasonic A-scans and B-scans

When all of the A-scans are combined into one B-scan image, the image has a resolving power of about 10 microns vertically and 20 microns horizontally.  Ten microns is 1/100th of a millimeter.  Compare that to the resolution of a good ophthalmic ultrasound at 100 microns, or 1/10th of a millimeter.  The difference is illustrated below.  The image on the right has 1/10th of the pixels per inch that the image on the left does.  The image on the right would represent the resolution of the ultrasound as compared to the resolution of the OCT on the left.

The time domain OCT has been a revolutionary instrument for ophthalmology, giving us non-invasive views of the retinal structure with micron level resolution.

Above: a stage 2 macular hole imaged with the OCT 3

1. Motion artifacts
2. Failure to identify RNFL and RPE layers on low signal to noise ratio scans
3. Interpolated data
4. Lack of registration

Motion Artifacts

The highest resolution scan, covering about 6mm, takes just short of 2 seconds to complete.  This is because the reference arm (mirror) is moving.

A Retinal Map Analysis, which is generated from the six radial scans of the Fast Macular Thickness Map, is pictured below.  This analysis gives a quantification to the retinal "volume" within a 6mm diameter circle centered on the fovea.  The protocol is limited to six scans because of the time involved in the scanning process, which is nearly two seconds of scan time.  Remember that the longer the scan time, the more measurement error due to eye movement.   The degree of movement error can be monitored by noting the center thickness and the deviation from center thickness (arrow).  If there is no movement, then the thickness measurement at the center of the circle, and the center of each scan line, should be exactly the same.  A "good" measurement (minimal movement) is said to have a deviation within 10% of the center measurement.  Theoretically, the scan can be repeated until an acceptable measurement is achieved.

For quantitative evaluations, the Zeiss OCT 3 has a software protocol that traces the RNFL and RPE layers, so that the distance between the two can be measured to give a retinal thickness reading.  The software identifies the layers by the difference in the signal strength of adjoining layers.  The RNFL and RPE layers are more highly reflective (more orange and red colors) than the vitreous or the inner layers of the retina.  If the signal strength is low, because of a poor optical path within the eye, or because of poor scanning technique, then the software may fail to correctly identify the position of RNFL and/or the RPE, resulting in an erroneous measurement.

The figure below illustrates the problem.  Notice that the white line that identifies the RPE layer jumps up to the RNFL at the left side of the scan, causing an erroneous measurement.

Interpolated Data

The third major source of error has to do with interpolation when doing a Retinal Map Analysis.  Interpolation is a method of constructing new data points from a discrete set of known data. The volume measurement covers the area of a 6mm diameter circle, but thickness data is only available from 6 radial line scans that are evenly spaced around the circle.  The thickness measurements for the considerable area between the lines is interpolated, which means that it is not real data but an estimate based upon what might be there if the numerical trend were to continue.  The yellow area in the scan circle below represents the interpolated area.

The third major source of error has to do with "registration" and comparing serial measurements.  In other words, when the patient comes back at a future date you want the ability to compare scans, to see if the condition is getting better or worse.  The problem is that the scans from the Zeiss OCT 3 are not "registered" to an anatomical location on the retina.  When repeating a scan, you are not sure if you are scanning the same area of the retina or not.  The video image that is captured with the scan gives you a general idea of the location, but the video image is not taken at the same time that the scan is captured.  If there is an identifiable landmark in the scan, such as the foveal depression, then you are more certain of the location, but the foveal depression is not always present when there is retinal edema.

Consider the retinal thickness map imaged below.  We notice from the direction indicator in the center of the image that this was a vertical scan.  The video "fundus" image on the right was not taken at the same time the OCT was captured, but the image gives us an indication that the macular area of the right eye was being scanned.  The OCT scan shows marked intra-retinal edema.  We cannot be sure that the foveal area is included in the scan because there is no foveal depression present.

An RNFL thickness scan is show below.  The video image on the right shows the circular scan placed around the optic nerve head.  The arrows point to the software generated lines that identify the top and bottom of the nerve fiber layer.  The software measures the vertical distance between the two lines to to arrive at the RNFL thickness

1. Motion artifacts
2. Failure to identify the upper and lower borders of the RNFL on low signal to noise ratio scans
3. Lack of registration

Since only one scan is required, error due to motion during the scanning process is minimized, but is still a factor.  A more troublesome source of error is created because the eye continues to move after the operator centers the scan circle on the optic nerve head.  A video image of the circle scan placement is frozen after scan acquisition is initiated, but the image is not captured at the same time that the scan is captured.  Therefore, there is no certainty that the circle was centered around the optic nerve head at the time of the scan.

Measurement errors occur when the software fails to identify the upper and lower boundaries of the retinal nerve fiber layer.  These boundaries are identifiable because they are more highly reflective than the bordering structures.  A poor scan due to optical opacities or operator error will reduce the signal-to-noise ratio and prevent the software from making a proper identification and tracing.

When performing serial scans over time, it is important to scan in the exact same position around the optic nerve head.  The Zeiss Stratus OCT 3 offers no way to register the scanning circle with anatomical landmarks in order to keep the positioning the same.  The operator must be relied upon centering the circle visually as best he or she can.