Microglia are resident central nervous system macrophages and the first responders to neural injury. Until recently, microglia have been studied only in animal models with exogenous or transgenic labeling. While these studies provided a wealth of information on the delicate balance between neuroprotection and neurotoxicity within which these cells operate, extrapolation to human immune function has remained an open question. Here we examine key characteristics of retinal macrophage cells in live human eyes, both healthy and diseased, with the unique capabilities of our adaptive optics–optical coherence tomography approach and owing to their propitious location above the inner limiting membrane (ILM), allowing direct visualization of cells. Our findings indicate that human ILM macrophage cells may be distributed distinctly, age differently, and have different dynamic characteristics than microglia in other animals. For example, we observed a macular pattern that was sparse centrally and peaked peripherally in healthy human eyes. Moreover, human ILM macrophage density decreased with age (∼2% of cells per year). Our results in glaucomatous eyes also indicate that ILM macrophage cells appear to play an early and regionally specific role of nerve fiber layer phagocytosis in areas of active disease. While we investigate ILM macrophage cells distinct from the larger sample of overall retinal microglia, the ability to visualize macrophage cells without fluorescent labeling in the live human eye represents an important advance for both ophthalmology and neuroscience, which may lead to novel disease biomarkers and new avenues of exploration in disease progression.

Purpose To characterize retinal ganglion cell morphological changes in patients with primary open-angle glaucoma associated with hemifield defect (HD) using adaptive optics–optical coherence tomography (AO-OCT). Methods Six patients with early to moderate primary open-angle glaucoma with an average age of 58 years associated with HD and six age-matched healthy controls with an average age of 61 years were included. All participants underwent in vivo retinal ganglion cell (RGC) imaging at six primary locations across the macula with AO-OCT. Ganglion cell layer (GCL) somas were manually counted, and morphological parameters of GCL soma density, size, and symmetry were calculated. RGC cellular characteristics were correlated with functional visual field measurements. Results GCL soma density was 12,799 ± 7747 cells/mm 2 , 9370 ± 5572 cells/mm 2 , and 2134 ± 1494 cells/mm 2 at 3°, 6°, and 12°, respectively, in glaucoma patients compared with 25,058 ± 4649 cells/mm 2 , 15,551 ± 2301 cells/mm 2 , and 3891 ± 1105 cells/mm 2 ( P < 0.05 for all locations) at the corresponding retinal locations in healthy participants. Mean soma diameter was significantly larger in glaucoma patients (14.20 ± 2.30 µm) compared with the health controls (12.32 ± 1.94 µm, P < 0.05 for all locations); symmetry was 0.36 ± 0.32 and 0.86 ± 0.13 in glaucoma and control cohorts, respectively. Conclusions Glaucoma patients had lower GCL soma density and symmetry, greater soma size, and increased variation of GCL soma reflectance compared with age-matched control subjects. The morphological changes corresponded with HD, and the cellular level structural loss correlated with visual function loss in glaucoma. AO-based morphological parameters could be potential sensitive biomarkers for glaucoma.

Adaptive optics—optical coherence tomography (AO-OCT) is a non-invasive technique for imaging retinal vascular and structural features at cellular-level resolution. Whereas retinal blood vessel density is an important biomarker for ocular diseases, particularly glaucoma, automated blood vessel segmentation tools in AO-OCT have not yet been explored. One reason for this is that AO-OCT allows for variable input axial dimensions, which are not well accommodated by 2D-2D or 3D-3D segmentation tools. We propose a novel bidirectional long short-term memory (LSTM)-based network for 3D-2D segmentation of blood vessels within AO-OCT volumes. This technique incorporates inter-slice connectivity and allows for variable input slice numbers. We compare this proposed model to a standard 2D UNet segmentation network considering only volume projections. Furthermore, we expanded the proposed LSTM-based network with an additional UNet to evaluate how it refines network performance. We trained, validated, and tested these architectures in 177 AO-OCT volumes collected from 18 control and glaucoma subjects. The LSTM-UNet has statistically significant improvement (p < 0.05) in AUC (0.88) and recall (0.80) compared to UNet alone (0.83 and 0.70, respectively). LSTM-based approaches had longer evaluation times than the UNet alone. This study shows that a bidirectional convolutional LSTM module improves standard automated vessel segmentation in AO-OCT volumes, although with higher time cost.