Purpose: Structure-function models of the retinal ganglion cell layer (GCL) are limited by spatial inconsistencies and high measurement variability. Suboptimal account of physiological changes over time delays the identification of deviations, such as early changes with glaucomatous disease. Current advances aim to develop normative structural databases and establish correlation with spatially summated functional parameters. The current study leveraged off such paradigms mapping spatially discrete GCL structure-function correlations, thereby significantly improving capacity to differentiate age-and disease related loss.

Methods: Spectralis optical coherence tomography of the central 30 x25 field was performed on 254 patients with no evidence of optic neuropathy or posterior pole disease. Layer segmentation and GCL thickness over customized spatial areas was obtained through a custom designed algorithm. The software accuracy was determined by comparison of results to the Spectralis 8x8 posterior pole grid data. Structure-function models were subsequently developed from spatially adjusted GCL areas in conjunction with full threshold Humphrey visual field test results from 40 patients using previously established cluster methods.

Results: Customized segmentation and extraction of GCL thickness did not significantly differ from automated thickness analysis (1-way ANOVA, p> 0.05) when comparing equivalent posterior grid areas. GCL measurements adjusted to spatially coincide with visual field stimulus locations, however, significantly increased by 3.5 µm on average and 15µm for areas equivalent to macular visual field locations. Consequently, the structure-function relationship across 64 visual field test point is linear with R 2= 0.79 (Figure), increasing to above 0.98 for clustered correlations.

Conclusions: Spatially-temporally adjusted structural and functional measurements of the central GCL are highly concordant, providing a robust model to describe normal, age-related change. This paradigm is directly transferable to current clinical instrumentation and provides a refined normative data range, thus, facilitating earlier and more sensitive detection of glaucomatous damage. Furthermore, the current model allows the immediate prediction of localised visual function from structural imaging.