Subjects

All patients with POAG who visited the Glaucoma Service at Kyoto University Hospital between November 2012 and July 2015, as well as healthy volunteers, were considered as potential participants in this study. All patients and volunteers underwent a comprehensive ophthalmologic examination, which included autorefractometry/keratometry, uncorrected and best-corrected visual acuity (BCVA) evaluation using a 5-m Landolt chart, IOP measurement using a Goldmann applanation tonometer, axial length measurement using an IOLMaster (Carl Zeiss Meditec, Dublin, California, USA), visual field testing using the Humphrey Field Analyzer (HFA; Carl Zeiss Meditec), gonioscopy, dilated funduscopy, stereo fundus photography (3Dx simultaneous stereo disc camera; Nidek, Gamagori, Japan), red-free SLO fundus imaging (Heidelberg Retina Angiography [HRA] 2; Heidelberg Engineering, Heidelberg, Germany), SDOCT (Spectralis HRA+OCT; Heidelberg Engineering), SSOCT (DRI-OCT-1; Topcon, Tokyo, Japan), and our prototype AO-SLO.

Glaucoma was diagnosed in cases of (1) diffuse or localized rim thinning evident on stereo disc photography and/or (2) RNFL defects—determined using red-free fundus imaging—that corresponded to glaucomatous visual field defects. This diagnosis was carried out during our glaucoma service meeting before the recruitment to this study. The optic disc appearance was independently evaluated by 3 glaucoma specialists, who were masked to all other information about the eyes. If any examiner disagreed with the others regarding the classification of an eye, the group reviewed and discussed the fundus color and stereo photographs until they had reached a consensus.

We defined glaucoma-related visual field defects according to the Anderson and Patella criteria (using standard automated perimetry in accordance with the 24-2 Swedish Interactive Threshold Algorithm [SITA] standard); the criteria were as follows: (1) abnormal range on the glaucoma hemifield test; (2) pattern deviation of <5% in 3 adjacent points, 1 of which was <1%; or (3) pattern standard deviation of <5% of the normal reference value, which is confirmed by 2 consecutive tests, and the tests were considered to be reliable when there was a fixation loss of <20%, false-positive results of <20%, and false-negative results of <20%.

Patients with POAG were only included if (1) the visual field sensitivity in the parafoveal region (within 5 degrees of the fovea) had deteriorated by less than 20 dB, and (2) the deterioration had occurred at least 3 years before their recruitment in the study. These inclusion criteria were determined in order to detect any glaucomatous damage to the cones sensitively. The exclusion criteria were as follows: (1) contraindication for mydriasis; (2) spherical equivalent refractive error of >5.0 diopters or <-10.0 diopters, or cylindrical refractive error of <-3.0 diopters; (3) axial length of ≥26.0 mm; (4) unreliable HFA results (ie, fixation loss of ≥20%; false positives or false negatives of ≥20%); (5) history of intraocular surgery other than cataract surgery; (6) evidence of vitreoretinal disease, or of brain diseases, diabetes mellitus, or other systemic diseases that might affect the eye; (7) poor image quality. If both eyes in a glaucoma patient were eligible, the eye with the more severe visual field loss was selected. If both eyes of a healthy subject were eligible, 1 eye was selected randomly.

Adaptive-optics Scanning Laser Ophthalmoscopy

We developed an original prototype AO-SLO system in collaboration with Canon Inc (Tokyo, Japan). The system comprises an AO subsystem, a high-resolution confocal SLO imaging subsystem, a wide-field imaging subsystem, and a pupil observation subsystem. This system allows the user to visualize individual cone photoreceptors in vivo. We obtained a series of AO-SLO images from several locations ( Figure 1 ); at each location, we moved the focus from the RNFL to the retinal pigment epithelium (RPE) to obtain the image series. We obtained AO-SLO images that focused on the cone mosaic, capturing up to 3 different fields of view at each location (L: 1700 × 1700 μm, M: 820 × 820 μm, and S: 340 × 340 μm). First, we obtained images at 3 different fields of view (L, M, and S) at the foveal center; then we captured images at 2 fields of view (M and S) centered 0.5 mm and 1.0 mm from the foveal center (both superior and inferior). Then we obtained images at 3 fields of view (L, M, and S) that were centered 1.5 mm superior or inferior, and 1.5 mm nasal or temporal, to the foveal center (diagonally 2.1 mm). At each location, 32 images were averaged to reduce noise. After image acquisition, we created montages of the AO-SLO images by matching the vessels and cone mosaic patterns; this allowed automatic calculation of the distances between the foveal center and target areas. We only used the highest-magnification images (S) for quantification, and we quantified these images in smaller areas that contained approximately 100 cones, each of which was at a targeted distance from the foveal center. We also obtained AO-SLO images that focused on different depths between the RNFL and cone mosaics.

Images of the left eye of a healthy 59-year-old woman. (Top left) Fundus photograph showing no abnormalities—there are no signs of glaucoma, such as disc-rim thinning or nerve fiber layer defects. (Top right) Montage from adaptive-optics scanning laser ophthalmoscopy (AO-SLO) images; asterisk indicates the foveal center. Three different field-of-view images—centered at the foveal center—were obtained: L (yellow box; 1700 × 1700 μm), M (blue box; 820 × 820 μm), and S (green box; 340 × 340 μm), followed by 2 field-of-view images (M and S) centered 0.5 mm and 1.0 mm from the foveal center (superior and inferior). Finally, 3 field-of-view images (L, M, and S) centered 2.1 mm superior–nasal, superior–temporal, inferior–nasal, and inferior–temporal to the foveal center were obtained. (Second row) Magnified spectral-domain optical coherence tomography (SDOCT) image—taken in the direction of the vertical arrow in top left—showing normal retinal structures. White arrowheads indicate where the retinal thickness was measured. (Bottom row) AO-SLO images showing a normal cone mosaic in the area indicated by the white boxes in top right (0.5 mm and 1.0 mm superior and inferior, 2.1 mm superior–nasal, superior–temporal, inferior–nasal, and inferior–temporal to the foveal center). Scale bar: 50 μm.

To assess the cones, we applied the automated cone labeling process established by Li and Roorda. When the algorithm failed to properly label the cones, an observer (who was masked to clinical information) manually labeled the areas where cones were visible but not detected automatically. To assess the interobserver intraclass correlation coefficients (ICCs), another masked observer independently carried out the manual correction of the automated cone labeling in randomly selected eyes, and measured cone density, the hexagonal Voronoi domains, and the ratio of average nearest-neighbor distance (NND) to expected NND. To assess the intersession reproducibility, ICCs were assessed between images repeatedly obtained at the same locations. To obtain accurate scan lengths, we corrected for the magnification effect using the adjusted axial length method reported by Bennett and associates. Spatial organization of the cone mosaics was assessed in terms of NNDs and Voronoi domains. The NNDs were determined by measuring the minimum distances to the center of the neighboring cells within the cone mosaic. Expected NNDs were calculated using a virtual, perfectly hexagonally packed mosaic with a density equal to that at each measured location. The ratio of the average NND to the expected NND was calculated at each location. The Voronoi domain of each cell was constructed by identifying all points that were closer to that cell than to any other cell in the mosaic. The ratio of the hexagonal Voronoi domain was calculated at each location.

In the current study, we only assessed locations that corresponded to HFA test points where (1) the sensitivity had deteriorated by less than 20 dB and (2) this deterioration had occurred at least 3 years prior to recruitment in the study. We defined these as “affected locations,” and the quantified data in each table are limited to these locations. If there were 2 or more affected locations on the different sides (superior and inferior), cone density, hexagonal Voronoi domain ratio, and average NND/expected NND were measured at each location, and the mean density or ratio of each distance was calculated. Eyes with glaucoma of >15 dB vertically asymmetric severity were analyzed separately.

Optical Coherence Tomography

The Spectralis HRA+OCT system was used to scan the macula. Our macular scan protocol comprised 19 vertical line scans and scanned a macular area that was centered at the fovea, 30 degrees in height, and 15 degrees in width. To reduce speckle noise, each image was obtained by averaging 50 B-scans.

In order to confirm that the glaucomatous damage had occurred where the AO-SLO images were obtained, we measured the thicknesses of the RNFL, as well as those of the INL, the ONL plus the outer plexiform layer (OPL), and the photoreceptor layer. The ONL + OPL thickness was defined as the distance from the inner border of the OPL to the external limiting membrane (ELM). Photoreceptor thickness was defined as the distance from the ELM to the inner border of the RPE. The continuity of the ellipsoid zone (EZ) line was also assessed. We measured the combined thicknesses of the ONL and OPL, instead of the ONL thickness alone, because it is difficult to accurately measure the ONL thickness using standard SDOCT images. We performed the measurements using the built-in digital caliper tool of the SDOCT system, with adjustments for the displacement between the cones and RGCs, in accordance with the formula introduced by Sjostrand and associates : namely, y = 1.29 × (x + 0.046) ^0.67 , where “y” is the RGC eccentricity and “x” is the cone eccentricity. The measurements were conducted independently in randomly selected eyes by 2 observers, and the interobserver ICCs were assessed. Additionally, we performed SSOCT (DRI-OCT-1) examinations. En face images of each layer were automatically reconstructed from a 3-dimensional volumetric scan centered at the fovea using software developed by the Topcon Corporation.

Statistical Analysis

We used the unpaired t test to compare glaucomatous eyes with healthy eyes in terms of mean age, refractive error, axial length, mean deviation (MD) value, and IOP. Cone density, hexagonal Voronoi domain ratio, and ratio of average NND to expected NND were compared between glaucomatous eyes and normal eyes using the unpaired t test. In glaucomatous eyes with vertically asymmetric severity, the paired t test was used to compare the same parameters between more and less affected locations. The unpaired t test was used to compare parameters between eyes with dark areas on AO-SLO and eyes without them. A P value < .05 was considered statistically significant. Where applicable, data are presented as mean ± standard deviation (SD).