Article Date: 9/1/2008

Computer-Assisted Glaucoma Progression Analysis

Computer-Assisted Glaucoma Progression Analysis

The second part of this series describes how the evolution of glaucoma progression technology is impacting optometrists.

contributing editor

Accurate measurement of disease progression in glaucoma is vital to patient welfare, yet difficult to achieve. Without it, you, the practitioner, are bereft of information on which to base clinical management decisions.

But, so little is known about how glaucoma advances — does it occur in a linear fashion, one cell at a time, or stepwise, great clumps of cells dying at separate locations? Should structural data supersede functional data, or vice versa? What differentiates glaucoma from normal, age-related structural decline? — that obtaining reliable information is a monumental task. Even a widely accepted definition of what constitutes progression has been impossible to reach.

In recent years, eye care has turned to computer software to help sort through the complex data diagnostic equipment has obtained. This article, second in a two-part series, explores some of those programs and the technology that supports them.

Humphrey visual field analyzer

In the old days, a technician would print out an overview of a series of visual fields, and the clinician would spread them side by side on a table and scan through the grayscale, total deviation and pattern deviations to identify progression. This took a lot of guesswork, and the imprecision meant that one doctor's diagnosis of progression was another's normal test variability finding.

To help reduce such uncertainly, Carl Zeiss Meditec (CZM) in 2003 introduced Glaucoma Progression Analysis (GPA) software, designed for use with the company's Humphrey Field Analyzer. The combined average of two visual fields forms the baseline. Then, the device performs an automated point-by-point analysis of subsequent exams, flagging each change with a small triangle — variously colored-in depending on the repeatability of the defect — or an "x" to indicate the deepest, most certain damage. During follow-up exams, the GPA compares results to both the baseline and the last two fields performed, thus helping to identify progression from baseline as well as more short-term changes. When the same three points worsen in two consecutive field tests, the printout reads "Possible Progression." Degradation of three or more points in three consecutive fields results in a "Likely Progression" reading.

Either event-based or trend-based methodology calculates all progression analysis.

To differentiate between clinically significant progression and normal variability, the software uses data collected during the Early Manifest Glaucoma Trial (EMGT), in which hundreds of glaucoma patients at multiple study sites underwent field testing four times in one month. Because the short time period of the trial precluded the occurrence of significant progression, an amalgam of the data gives the computer a fair range of the average glaucoma patient's test variability. So, unless changes exceed these parameters, the GPA_disregards them as noise.

The latest version of this software, now called Guidance Progression Analysis (GPA), offers a new piece of information, the Visual Field Index (VFI) — a number between 0 and 100, with 100 being a perfect visual field, and an accompanying trend analysis graphic.

The VFI provides trend-based analysis — a form of analysis not robustly present in the original software (except by way of mean deviation, over which VFI is a huge improvement, according to the company). The old version of GPA was chiefly an event-based analysis.

Either event-based or trend-based methodology calculates all progression analysis. Both have advantages and disadvantages. For example, event-based analysis offers better sensitivity than trend-based analysis to small, localized progression. On the other hand, trend-based analysis (sometimes called regression analysis) provides better data than the event-based analysis as to how fast the disease is progressing. Practitioners do not consider one superior to the other, and generally speaking, they want as much information as possible from both computations.

The GPA has been popular, and its success probably helped encourage the introduction of progression software for other glaucoma diagnostics. Practitioners say the biggest challenge with GPA is ensuring the baseline fields are of the highest possible quality, lest all subsequent data be skewed. If you have a complaint about GPA, it's most likely regarding the excessive subjectivity of visual fields themselves.

Gadi Wollstein, M.D., director of Ophthalmic Imaging Research Laboratories at the University of Pittsburgh, expresses the opinion of many in the field when he says, "Unfortunately, at this point I don't think we are at the time when we can completely abandon the visual field, although that would be to me the ultimate goal. It's a subjective way of assessing vision."

"While IOP, CCT, family history, vascular status and other factors play a role in painting the picture of the entire glaucomatous landscape, it is damage to the optic nerve that we are ultimately trying to ameliorate."

James Fanelli. O.D.

Optical coherence tomography (OCT)

OCT provides cross-sectional views of the retina, retinal nerve fiber layer (RNFL), and optic nerve by means of an echo time delay of back-scattered light.

The previous version of the progression software for CZM's Stratus OCT, known as Serial Analysis, provided a trend analysis in a visual graphic that overlaid the RNFL thickness at each of the 256 points along the scan. A different color represented each visit date. But, many practitioners noted it was difficult to tell one line from the other or identify any significant changes.

Earlier this year, CZM began offering a version of GPA Advanced Serial Analysis for the Stratus OCT. The software supplies quantitative values for overall average RNFL thickness as well as for each quadrant. It also plots average thickness through time and uses regression analysis to report average RNFL thickness as a slope, confidence interval and p value to track whether a statistically significant change has occurred.

Scanning laser polarimetry

Scanning laser polarimetry with variable corneal compensation (GDx VCC) assesses the RNFL by measuring the light retardation caused by the thickness and birefringence of the RNFL structure.

CZM offers a version of its GPA software for the GDx. This software provides a trend-based analysis that employs three separate algorithms to detect RNFL change, ranging from focal to diffuse: 1.) the image progression map, which is most sensitive to narrow, focal change; 2.) a TSNIT progression graph, which picks up broader focal change than the progression map; and 3.) summary parameter charts, which are most sensitive to diffuse change. At the top of the printout, a summary box indicates whether the software has flagged a statistically significant change, and if so, by which algorithm(s). The box's color code is yellow for possible progression; red for likely progression; and purple for possible increase.

This software offers you easy-to-interpret information about the structural health of the RNFL. The GDx measures polarimetric retinal nerve fiber layer (RNFL) thickness, an indicator of RNFL health status. Note that polarimetric thickness is not to be confused with the anatomical thickness measured by other devices. Although some clinicians have complained that birefringence changes aren't uniform around the optic nerve head, the clinical utility of the instrument remains valid. This is because the software makes comparisons with normal values and with individual baseline values, according to the company.1 The GDx contains a normative database of polarimetric RNFL thickness values found in known normal subjects, which it uses to separate abnormal from normal RNFLs.

For reasons that are not yet well understood, some patients generate atypical scans due to interference with the birefringence patterns. Enhanced Corneal Compensation (ECC) software, soon to be available commercially, should correct for these birefringence artifacts, according to the company.

The device does not provide information about the optic nerve head.

Heidelberg Retina Tomograph (HRT)

Usually known as HRT, con-focal scanning laser ophthalmoscopy creates three-dimensional images based on reflected light. It assesses the area of the optic nerve tissue, volume of the neuroretinal rim tissue and perioptic RNFL thickness.

Its software, the Topographical Change Analysis (TCA), differentiates true biological changes from random variability by using the Moorfields Regression Analysis (MRA) technique. The MRA shows the rim area, compared with the disk area, in six sectors, and the device graphically presents the status of an individual optic nerve head in relation to normal eyes. It identifies abnormal sectors of the optic nerve as "within normal limits," "borderline," or "outside normal limits."

Doctors say the HRT's reputation stands on its ability to depict optic disk morphology. "I prefer the morphometric change analysis, as it clearly shows which specific areas of the optic nerve have undergone change," says James Fanelli, O.D., of Wilmington, N.C., "and you can then evaluate those specific areas, for example, to see whether they coincide with like changes in the visual field, or whether they represent progression in the areas typically affected by glaucoma — namely the inferior and superior temporal rims."

Ultimately, glaucoma is a disease of the optic nerve head, Fanelli adds. "While IOP, central corneal thickness, family history, vascular status and other factors play a role in painting the picture of the entire glaucomatous landscape, it is damage to the optic nerve that we are ultimately trying to ameliorate," he says.


Because measurements of basic ocular structures, such as the optic nerve head and the RNFL vary so greatly from person to person, it's difficult to determine what constitutes normal. However, less inter-patient fluctuation appears in the macula than in the optic disk. This implies a smaller deviation from normal in the macula enables progression detection.

Such is the thinking behind the Retinal Thickness Analyzer (RTA-5), made by Talia/Marco, which is ideal for imaging the macula but less so for the optic nerve head (though optic nerve head information is provided as well). The device shoots a narrow slit of helium neon laser into the eye and identifies the RNFL and retinal pigment epithelium, calculating the difference between the two to determine RNFL thickness. The data is essentially a combination of scanning laser ophthalmoscopy and a high resolution digital fundus camera. A color-coded deviation probability map compares patient results to a large, normative database and indicates statistically significant thickness variances. The full macula, perimacular or peripapillary regions can present data.

The main advantage of the RTA-5 is that it can image very large areas of the retina at one time, and many clinicians consider this the best strategy for tracking glaucoma in its early stages. In addition, the RTA can generate a serial follow-up report, which contains change-probability maps for both macula thickness deviation and C/D ratio deviation.

"I feel that I will see the most subtle changes in the macular thickness," says Janet M. Mint, O.D., of Jacksonville, Fla. "A 40 to 50 micron change may require a change in the treatment plan."

Looking ahead

Manufacturers will continue to improve the progression analysis capabilities of their diagnostic instruments. For example, a software upgrade to the Octopus series of perimeters (Haag-Streit USA) is pending a U.S. release. Called EyeSuite, this software is the database for the Octopus series of perimeters.

The new generation spectral domain OCTs (also known as Fourier Domain OCTs), offer high resolution, obtain images about 70 times faster than previously and — perhaps most significantly — image ganglion cell layers in the macula region to a degree that almost rivals histology. No fewer than seven companies are in the process of launching SD-OCTs.

"The technology is superb," says Richard J. Madonna, O.D., F.A.A.O., an associate professor at the State University of New York (SUNY) College of Optometry, who has experience with four of the new spectral domain instruments. "When you are looking at axial resolution of four or five microns, you are certainly able to qualitatively look at the retina in a way that you have never been able to before."

Also, the spectral domain OCTs offer tools to measure progression analysis. For example, the RTVue (Optovue) has tools to diagnose and measure glaucoma progression, including a normative database, ganglion cell complex analysis, pattern deviation analysis, optic nerve head/RNFL analysis, progression/trend analysis and asymmetry analysis, according to company literature.

As is the case with other categories of diagnostic equipment, each manufacturer's sprectral domain OCT uses a unique set of features to address glaucoma progression. For example, the Spectralis SD-OCT (Heidelberg Engineering) offers an eye tracking feature known as TruTrack. With this feature, the Spectralis takes its scan focusing on a specific location of the retina. At subsequent visits, the Spectralis locks onto the same position.

In the near future, an advanced software package will be available for CZM's new Cirrus HD-OCT — one that will provide change from baseline analysis and flags for possible or likely nerve fiber layer loss, according to CZM's senior marketing manager Marianne Whitby.

Art, science and detection

Despite the proliferation of technology to help detect glaucoma progression, the task remains as much art as science. Many of the conclusions drawn from these technologies will likely prove in variance with each other. For instance, a paper presented at last year's American Academy of Ophthalmology annual meeting found poor agreement among data obtained from visual fields, OCT and HRT. Given the lack of a gold standard for defining or measuring glaucoma progression, this should come as no shock, according to the paper's lead author, Dr. Wollstein. "When looking at agreement, in a way it is kind of disappointing, but it is not really surprising." OM

1. Huang, XR, Bagga, H, Greenfield, GS, Knighton, RW. Variation of peripapillary retinal nerve fiber layer birefringence in normal human subjects. Invest Ophthalmol Vis Sci. 2004 Sep;45(9):3073-80.

Mr. Celia is a freelance healthcare writer based in the Philadelphia area.

Optometric Management, Issue: September 2008