A Study On Imaging Human Retinal Ganglion Cell Layer Structures Using Adaptive Optics Scanning Laser Ophthalmoscopy

Background and Objectives:Adaptive Optics(AO)permits highresolution imaging of the human retina in vivo at cellular level by compensating for the optical aberrations of the eye.AO can be added on multiple conventional ophthalmic imaging modalities and among which Adaptive Optics Scanning Laser Ophthalmoscopy(AOSLO)is an important device.AOSLO has multiple imaging modalities and multioffset imaging is capable of imaging human Retinal Ganglion Cells(RGCs)in vivo by placing larger aperture at multiple offset positions at the retinal conjugate plane.However,improvements of the system light pathway,multi-offset detection patterns,and image processing procedures still require further investigation.In this study,we used AOSLO imaging system to explore the optimal multi-offset detection and image processing procedures to achieve in vivo non-invasive imaging of the human retinal ganglion cell layer structures.Methods:24 healthy participants and 4 patients with posterior uveitis were enrolled from the clinical research platform of the University of Pittsburgh.AOSLO is located in Dr.Rossi’s lab,Department of Ophthalmology,University of Pittsburgh.A custom-made reflective pinhole mirror was installed in the light detection pathway that permits single wavelength light source for both confocal channel and multi-offset channel.Image processing and analysis were performed mainly with MATLAB1.Small aperture was used to characterize the light scattering at the retinal conjugate plane.The range of offset distance were determined based on analyzing the signal intensity at different offset distance.We then setup a series of offset distance within the range and performed sequential multioffset imaging using large aperture.Optimal offset distance was concluded by comparing the image quality.Metrics involve Michelson contrast of cellular structures and the radially averaged power spectrum density of the image.2.We compared the different types of images generated from sequential multi-offset imaging and determined the optimal image processing procedures.3.We imaged 19 healthy participants using the optimal sequential multi-offset pattern to evaluate the robustness and repeatability of the imaging technique.4.Based on the optimal parameters of sequential multi-offset imaging,we designed and installed an optical fiber bundle(FB)that allows for simultaneous acquisition of all offset positions and then compared it with the sequential approach.5.We imaged the ganglion cell layer of 6 healthy participants using FB multi-offset technique.Montages were made to quantify presumed microglial cells.Repeat imaging of the several retinal regions were carried out to form time-lapse videos and characterize microglia cell motility and morphological changes.6.We imaged 4 patients with infectious and non-infectious posterior uveitis in the region of retinal lesion.Immune cells(microglia and macrophages)distribution and dynamics were evaluated.Results:1.The optimal sequential multi-offset detection pattern is:radially arranged large apertures(around 8ADD)that sits 8-10ADD away from the central optical axis.This pattern was calculated in the unit of ADD and applies to systems with different ADD sizes.2.Three types of images were generated from sequential multi-offset acquisition including:single offset position images,differencing images,and MOD image.Both Michelson contrast of cellular structures and power spectrum density were highest in MOD image,following differencing images whereas single offset position image has relatively the lowest quality.Therefore,generating the MOD image should be considered the standard processing procedure.3.RGCs structures were revealed in 19 participants.We evaluated the cell density in images where RGCs mosaic is readily visible inspired by the calculation for cone mosaic.Some images failed to reveal clear RGCs mosaic,but we were able to discover several RGC structures within.Repeated cellular structures were seen in certain participants and we discovered presumed microglia cells.Repeat imaging of the same cellular structures were easier when the image acquisition was done on the same day.4.We replaced the single PMT with an optical fiber bundle that achieve simultaneous acquisition of all offset positions.The image quality was slightly higher than that from sequential multi-offset acquisition.The imaging efficiency and capability were largely improved.Single acquisition reduced from several minutes to tens of seconds that enabled more acquisition during one imaging session.We can also make montage image that covers larger region and image the optic nerve head region which was previous impossible.5.Presumed ganglion cell layer microglial cells were detected in healthy participants.These cells have an average soma diameter of 12.92μm,and an average soma size of 123.50 μm2.Microglial cells displayed two types of morphology,one being elongated that resembled ramified state and the other being more circular that resembled activated state.Microglial cells moved extremely slowly in healthy retinas with an average speed being 0.0017μm/sec with only a few moved higher,the maximum speed detected were 0.106μm/sec.Apart from displacement of cell location,they also showed morphological changes that resembled extending and retrieving of processes.6.Multiple circular structures were found in a patient’s retina with acute infectious posterior uveitis.Some of the structures have evenly reflective surface whereas others showed uneven reflectance that contain internal granular structures.These cells moved fast in acute phase of the disease with the maximum speed reaching 2.4μm/sec.Cell amount and speed in the same retinal location decreased as the patient recovered.In non-infectious posterior uveitis patients,we discovered increased amount of microglial cell-like structures;they were more active in patient in the active state of the disease but static in quiescent state of the disease.Cell activity in these patients were reflected not by the displacement of location but rather by the change of their morphology locally.Conclusion:1.Multi-offset AOSLO enabled robust and reliable imaging of the human retinal ganglion cell layer structures(RGCs and microglial cells)in vivo with ultra-high resolution.2.Custom-made reflective pinhole mirror permits using the same light source for confocal channel and multi-offset channel,thus reduced the chromatic aberration and increased the light power.3.The optimal detection pattern for multi-offset imaging is to set large apertures(diameter being～8ADD)arranged radially at 8-10ADD away from the central position.4.Optical fiber bundle based multi-offset detection simplified the optical design and improved the imaging efficiency and capability.It allows for montaging multiple images,visualizing regions such as the optic nerve head region,and monitoring dynamic within a short period of time.5.FB multi-offset AOSLO enabled visualization and characterization of the ganglion cell layer microglial cell morphology and activity.When applied on retinal inflammation patients it helped capture immune cell morphology and activities.