Spatial Patterns of Ganglion Cell Complex Thickness Are Associated with Brain Parameters

Abstract

Purpose : Brain abnormalities have been associated with glaucoma, which damages the ganglion cell complex (GCC). In this study, we will directly associate the GCC thickness with brain parameters in subjects without glaucoma.

Methods : From the large population-based LIFE-Adult study (Leipzig Research Centre for Civilization Diseases), we extracted the machine-segmented GCC thickness from macular optical coherence tomography (SD-OCT) scans. We applied an artificial intelligence method termed non-negative matrix factorization to determine GCC thickness patterns. The brain parameters used include left and right hippocampal volumes, white matter volume and white matter hyperintensity volume, which were measured using automatic segmentation tools (FreeSurfer and LesionTOADS) on magnetic resonance imaging scans. The global average GCC thickness and each of the GCC thickness patterns were associated with the brain parameters by linear regression with adjusted effects for age, gender, scan focus and intracranial volume. P values were corrected for multiple comparisons for the GCC thickness patterns by the false discovery rate method. For linear regression analyses, subjects with glaucoma diagnosis were excluded and one random eye per subject was selected if data for both eyes were available.

Results : We determined 12 GCC thickness patterns (Figure 1) using 17,877 OCT scans from 17,877 eyes of 9,029 subjects (Age: 57.5 ± 12.4 years, 52.1% female). The brighter regions in the GCC thickness patterns indicate the more informative zones with greater variations across subjects than the blue regions. 1,747 eyes from 1,747 subjects (Age: 58.2 ± 15.8 years, 45.6% female) were used to link the GCC parameters with the brain parameters. The global average GCC thickness was not significantly associated with any of the brain parameters (Table 1). A lower Pattern 3 coefficient was correlated with lower left (p < 0.001) and right (p = 0.006) hippocampal volumes. A lower Pattern 5 coefficient was correlated with lower right hippocampal volume and higher white matter hyperintensity volume (p = 0.04 for both). A lower Pattern 5 coefficient was also correlated with lower white matter volume with borderline significance (p = 0.07; p = 0.005 before p value corrections).

Conclusions : Our results suggest that parafovea ring-shape thinning and inferior-nasal arcuate-shape thinning in GCC are specifically linked with unfavorable brain parameters.

This is a 2021 Imaging in the Eye Conference abstract.