Determining retinal nerve fiber layer thickness vulnerability zones in mild glaucoma with rescaled optic nerve heads

Abstract

Purpose : Structure-function modeling of mild glaucoma is hindered by the variability of eye anatomy as well as data reliability. We adjusted the optical coherence tomography (OCT) retinal nerve fiber layer thickness (RNFLT) images to achieve congruent optic nerve head (ONH) representations. Then we computed correlations (corr) to identify the vulnerability zones (VZs), the areas where the RNFLT might have associations with the visual function (VF) deficiency. We suspect the revealed VZs are more representative since the images were compared at a notionally common scale.

Methods : We computed corr of the total deviations (TDs) of the VFs (n=419, SITA Standard 24-2 Humphrey with at most mild glaucoma, mean deviation ≥-3dB) and accompanying OCT RNFLT images (Fig. 1A). We then performed a re-scaling operation, ensuring that ONHs are equally sized and recomputed the corr (Fig. 1B).

Results : Corr maps showed that most of the high corr regions are localized around the major arteries. This observation suggests general VZs rather than location-specific areas as assumed previously. Rescaling based on the ONH size increased the maximal corr at 45 / 52 VF locations. At 9 locations, the corr improved over 50%, with the largest increase of 86.7% (Fig 1B, red *).

Conclusions : High-resolution structure-function corr help identify retinal VZs in mild glaucoma based on OCT imaging. At many locations, higher corr were found after ONH rescaling. These higher corr supported that the VZs are generally clustered around the arteries. The peripheral areas of the RNFL map are now showing stronger corr at superior and inferior locations. We did not find any specific locations that indicate direct association with VF dysfunction. Special attention to RNFLT thinning in the VZs in glaucoma suspects may improve early-stage glaucoma detection.

This is a 2020 Imaging in the Eye Conference abstract.

 

Fig 1: Structure-function corr of the RNFLT and VF were shown. A) Original. B) After ONH re-scaling, the corr maps were reshaped to retain areas with n>100 data for computation. The numbers at the center are the corr coefficients (Green: increased, Grey: no change, Blue: decreased). Orange boxes: corr increased > 50%. Red * is where the corr has the largest change, 86.7%.

 

Fig 2: The zoom-in areas where the corr has 86.7% change, before (upper) and after (lower) after ONH scaling. The superior and inferior areas at the edge of the peripheral are showing stronger corr.