Perimetry and retinal imaging

Abstract

The Imaging and Perimetry Society’s 20th International Symposium was held between the 22 and 25 of January 2012 at the University of Melbourne (VIC, Australia). It is a well-established bi-annual international meeting, which focuses on methods for quantifying localized damage to the retina in diseases such as glaucoma, macular degeneration and diabetic retinopathy. These methods are critical to the diagnosis and management of these diseases. Attendees have backgrounds ranging from ophthalmology and psychology, to optical physics and statistics.

Keywords::

• glaucoma
• OCT
• perimetry
• polarimetry
• structure–function

The Imaging and Perimetry Society Symposium is a well-established biannual international meeting. It focuses on methods for quantifying focal damage to the retina. Accordingly, attendees have backgrounds ranging from ophthalmology and psychology to optical physics and statistics.

Retinal imaging has undergone explosive innovation over the past 10 years, the main developments being in the area of frequency domain (spectral) optical coherence tomography (OCT), and several new perimeters have also been introduced. Key topics are methods for comparing or combining data from these measurement techniques, and also addressing perceived problems with well-accepted techniques. One fruitful area arising from the overlap of imaging and perimetry is the creation of structure–function maps that relate the loss of nerve fibers around the optic disc with sensitivity loss at particular parts of the visual field in glaucoma. These maps allow us to decide whether observed field and disc margin defects are concordant, providing powerful diagnostic evidence.

The interesting

Chota Matsumoto and colleagues from Kinki University (Osaka, Japan) decided to try something fun; the CLOCK screener. They placed A3-sized advertisements in 62 million Japanese newspapers in the autumn of 2009. The advert displayed a few small targets at different radii from a fixation point. The concept was that as people rotated the advert about the fixation point they might see some of the targets disappear (when the targets fell in a scotoma, an obvious hole). The physiological blind spot was used to calibrate viewing distance and to train people to notice a disappearing target. A total of 3.5 million people tried it, 350,000 consulted a doctor, and 31,000 were diagnosed with glaucoma. The test also included an Amsler grid, thus 30,000 age-related macular degeneration patients and patients with related illnesses were also detected.

Ron Schuchard and colleagues of Stanford University (CA, USA) overturned ideas often portrayed in depictions of glaucomatous visual field defect by showing that 55% of patients had no percept of their scotoma, while 43% reported that sometimes smaller objects appeared and disappeared as they looked about. Perhaps training people on what they should attend to, rather than producing pamphlets with misleading depictions of scotomas, would help reduce undiagnosed glaucoma of 50%.

The perimeters

Functional testing (e.g., perimetry) is needed to identify pathology in several retinal diseases. Although imaging technologies are giving us ever better pictures of the retina, much of what we measure is only a correlate of disease. A major problem for perimetry is test–retest variability (TRV) [1,2]. Consensus opinion says that higher TRV means that even if we test six times in 2 years we will only have an 80% chance of detecting change in a patient whose mean defect is declining at 2 dB per year [3].

The Matrix^® and FDT^® perimeters from Zeiss use stimuli that display the spatial frequency-doubling illusion (FDi). The FDT came on the market in 1997, and the Matrix in 2004, and differ mainly in the density with which they can test the visual field. These and related perimeters were the subject of several presentations. It is proposed that a class of optic nerve cells, which relatively sparsely cover the retina, report FDi to the brain. Thus, when a few die, detectable gaps in the sparse phalanx of these cells occur. Other more redundant cell types would need larger numbers to die before a scotoma would be seen. FDi occurs when big blurry grating stimuli (i.e., having low spatial frequencies) are contrast modulated at more than 15 Hz. The concept is that under these conditions, a well-known nonlinear retinal mechanism, peculiar to the sparse cells, effectively squares the contrast, producing a stable percept with twice as many stripes as the original grating.

A new form of perimetry was introduced by Mitchell Dul of the State University of New York (NY, USA) and workers at Indiana University (IN, USA). Their study compared between-subject and between-TRV for size II stimuli and for Gabor wavelet-like stimuli, the latter scaling with visual field eccentricity. Both types of variability were lower for the new stimuli. The stimuli were similar to those used in FDT/Matrix perimeters but flicker more slowly, which may give better tolerance of cataract.

A more recent perimeter is flicker-defined form (FDF) perimetry, also known as the Heidelberg Edge Perimeter. Here, a 5° diameter patch of dots (referred to as dark dots) is presented against a field of similar-sized dots of the opposite contrast (bright). The contrasts are then reversed at >15 Hz and a stable ring is perceived, the ring margin corresponding to the edge where the contrast of the dots reverses. The dots contain low and high spatial frequencies, but at high flicker rates the contrast of the high spatial frequencies will be reduced, leaving the conditions for FDi mechanism to operate. That would produce a stable disc percept (with blurry edges), but spatial band-pass filtering by the visual cortex would cause a ring percept. Certainly, FDF and FDi stimuli have more in common than traditional, unflickered standard automated perimetry (SAP) stimuli.

A few presentations suggested that the dynamic range of FDF perimetry was less able than SAP to track changes in severe glaucomatous damage. The FDF perimeter also does an alternative test which is similar to SAP. Some papers indicated better detection of early damage by FDF. Earlier reports had its TRV being similar to Matrix and better than SAP. John Flanagan (University of Toronto, ON, Canada) showed that structure–function maps were better for FDF than for SAP, at least in early glaucoma.

There were two presentations on diabetic retinopathy and age-related macular degeneration from my group (Australian National University, ACT, Australia) using multifocal pupillographic objective perimetry. In response to visual stimuli the pupil makes small-scale modulations in diameter that are related to the sum of the activity on the two retinas. With independent stimuli for the two eyes, each can be tested concurrently and novel information on afferent and efferent defects, and the delay of responses, can be obtained.

Hot individuals

The classic problem with many test instruments is the use of population-based normative data. An alternative that can add value is sometimes to compare the left and right eyes of a patient, as even if your eyes are well outside the population norms, your two eyes should be similar. So why not create normative data from the subject’s own data?

Paul Artes and Balawantray Chauhan (Dalhousie University, NS, Canada) took this approach with data from a series of fields that showed progressive damage over time. To summarize, they created numerous data sets by scrambling the temporal order of the fields to create distributions that can indicate the probability that a particular real sequence of fields was obtained by chance. For the method to work, five or more fields are needed.

The concept of individual normative data was extended to structure–function relationships between visual field defects and the optic disc rim in glaucoma. Thus far, the correlations with population data are relatively poor, due in part to individual variation in the paths taken by ganglion cell axons to the rim. Jonathan Denniss and colleagues (University of Melbourne, VIC, Australia) explored developmental models of ganglion cell axons populating the optic disc, which permitted individual structure–function relationships to be created based on the relative positions of the fovea and the optic disc.

Imaging

Shaban Demerel and colleagues of Devers Eye Institute (Portland) compared Matrix and HFA II perimeters versus retinal nerve fiber layer thickness (RNFLt) as measured by Spectralis^® (Heidelberg Engineering, CA, USA) and GDx. Neither RNFLt measure was correlated with field damage above a certain thickness. SAP then showed a nonlinear relationship with initially poor discrimination, while Matrix adopted a linearly steep correlation with RNFLt once the threshold for thinning was broached. This may suggest that perimetry detects some damage before it shows up on imaging, and that Matrix indicates steeper change earlier, as suggested by other studies [4].

The last word on how to compare methods quantitatively was had by Ryo Asaoka and colleagues from Moorfields Eye Hospital (London, UK). They used so-called ‘random forests’ of decision trees to compare the diagnostic power of four retinal imaging methods to detect glaucoma. Patients had elevated intraocular pressure and abnormal Humphrey Field Analyzer mean deviation or pattern standard deviation or glaucoma hemifield test. They then ran the method on data from HRT, Stratus OCT and the GDx polarimeter to see how well they could diagnose glaucoma. In a cycle repeated 10,000 times, 60% of the data was selected at random and used to define the best diagnostic decision tree, which was then tested on the other 40%. Average diagnostic power was then computed. The outcome was GDx > OCT > HRT, with OCT as the single most important variable on average.

Wrap-up

The Imaging and Perimetry Society’s meeting is a small but high quality international conference that mixes serious multidisciplinary science with events such as the traditional national singing at the meeting dinner. It is particularly useful for younger eye care professionals and scientists. Accompanying persons are also catered for. The 2014 meeting will be held in Manhattan (Mitchell Dul, SUNY College of Optometry, NY, USA), and the 2016 meeting will be in Udine Italy (Marco Zeppieri, S Maria della Misericordia Hospital, Udine, Italy).

Financial & competing interests disclosure

T Maddess receives royalty income from Carl Zeiss Meditec (CA, USA) for the sale of the FDT^® and Matrix^® perimeters. Several patents and patent applications for multifocal pupillographic objective perimetry methods on which he is an inventor are licensed by the Australian National University to Seeing Machines Ltd (Canberra, ACT, Australia). Seeing Machines Ltd. also supplies the author with equipment and funding for research into multifocal pupillographic objective perimetry. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.