Pattern Recognition Analysis of Age-Related Retinal Ganglion Cell Signatures in the Human Eye

Abstract

PurposeTo characterize macular ganglion cell layer (GCL) changes with age and provide a framework to assess changes in ocular disease. This study used data clustering to analyze macular GCL patterns from optical coherence tomography (OCT) in a large cohort of subjects without ocular disease.MethodsSingle eyes of 201 patients evaluated at the Centre for Eye Health (Sydney, Australia) were retrospectively enrolled (age range, 20–85); 8 × 8 grid locations obtained from Spectralis OCT macular scans were analyzed with unsupervised classification into statistically separable classes sharing common GCL thickness and change with age. The resulting classes and gridwise data were fitted with linear and segmented linear regression curves. Additionally, normalized data were analyzed to determine regression as a percentage. Accuracy of each model was examined through comparison of predicted 50-year-old equivalent macular GCL thickness for the entire cohort to a true 50-year-old reference cohort.ResultsPattern recognition clustered GCL thickness across the macula into five to eight spatially concentric classes. F-test demonstrated segmented linear regression to be the most appropriate model for macular GCL change. The pattern recognition–derived and normalized model revealed less difference between the predicted macular GCL thickness and the reference cohort (average ± SD 0.19 ± 0.92 and −0.30 ± 0.61 μm) than a gridwise model (average ± SD 0.62 ± 1.43 μm).ConclusionsPattern recognition successfully identified statistically separable macular areas that undergo a segmented linear reduction with age. This regression model better predicted macular GCL thickness. The various unique spatial patterns revealed by pattern recognition combined with core GCL thickness data provide a framework to analyze GCL loss in ocular disease.

Consistency of Structure-Function Correlation Between Spatially Scaled Visual Field Stimuli and In Vivo OCT Ganglion Cell Counts

“…1C), each frame was manually assessed and corrected for segmentation error if pronounced layer misclassification, resulting in more than 50% difference in thickness to adjacent areas, was detected by an experienced operator (NY) following previously established protocol. 21 A total of 21.6% of the grids required manual correction and 4.6% individual grids were excluded from analysis as adequate segmentation was hindered due to the presence of major blood vessels; shadowing; poor quality B-scan (<15 dB signal scores); the optic nerve head encroached the grid; or more than 10% of the grid could not be analyzed because it was located outside of the scan area. Subanalysis confirmed that manual adjustments did not bias final measurements (Supplementary Fig.…”

“…21,[44][45][46] The age difference between individual subjects and the cohorts may therefore increase the variability within the dataset and confound the comparison of structure-function relationship between the normal and glaucomatous cohort. Hence, to minimize the variability imposed by intra-and intercohort age differences, the GCL thickness and DLS of all normal and glaucomatous subjects were age-corrected to the average age of the glaucomatous cohort using previously established regression analysis models as conducted in previous studies, 11,21,[24][25][26]31,32,45,47 including the commonly used SITAstandard algorithm of the HFA. 48 Specifically, for the GCL thickness, measurements from each normal and glaucomatous subjects were clustered into eight statistically separable classes and adjusted to match the average age of the glaucomatous cohort using a previously published conversion for all but the foveal area.…”

“…48 Specifically, for the GCL thickness, measurements from each normal and glaucomatous subjects were clustered into eight statistically separable classes and adjusted to match the average age of the glaucomatous cohort using a previously published conversion for all but the foveal area. 21 For the four central grids, corresponding to the foveal area, thickness measurements were recalculated for the cohort used in the previous study (n ¼ 201) 21 as detailed in the supplementary material ( Supplementary Figs. S2A-C) providing the appropriate conversion formula subsequently applied to the corresponding data in the current study ( Supplementary Fig.…”

“…[9][10][11][12][13][14] More recently, the development of optical coherence tomography (OCT) has allowed large-scale investigations to be conducted by utilizing thickness profile of various retinal structures such as nerve fiber layer and ganglion cell/inner plexiform layer as an in vivo surrogate measure of GC density, [16][17][18][19] including models for estimating the GC counts in vivo from ganglion cell layer (GCL) thickness. 20,21 The white-on-white Goldmann size III stimulus (GIII) is widely accepted as the standard stimulus size for clinical and scientific measurement of differential light sensitivity (DLS). While the choice of this stimulus size is based on historical rather than psychophysical origins, 4,15,22 it has been utilized in numerous structure-function investigations.…”

“…In particular, the use of GCc as a common structural metric permits structurefunction data from multiple stimulus sizes to be combined and compared directly. Therefore, we have investigated the correlation between previously developed in vivo GCc estimates 21 with DLS obtained from different-sized, white-onwhite visual field stimuli within the central retina in both a normal and glaucomatous cohort.…”