Automatic montaging of adaptive optics SLO retinal images based on graph theory.

We present a fully automatic montage pipeline for adaptive optics SLO retinal images. It contains a flexible module to estimate the translation between pairwise images. The user can change modules to accommodate the alignment of the dataset using the most appropriate alignment technique, provided that it estimates the translation between image pairs and provides a quantitative confidence metric for the match between 0 and 1. We use these pairwise comparisons and associated metrics to construct a graph where nodes represent frames and edges represent the overlap relations. We use a small diameter spanning tree to determine the best pairwise alignment for each image based on the entire set of image relations. The final stage of the pipeline is a blending module that uses dynamic programming to improve the smoothness of the transition between frames. Data sets ranging from 26 to 119 images were obtained from individuals aged 24 to 81 years with a mix of visually normal control eyes and eyes with glaucoma or diabetes. The resulting automatically generated montages were qualitatively and quantitatively compared to results from semi-automated alignment. Data sets were specifically chosen to include both high quality and medium quality data. The results obtained from the automatic method are comparable or better than results obtained by an experienced operator performing semi-automated montaging. For the plug-in pairwise alignment module, we tested a technique that utilizes SIFT + RANSAC, Normalized cross-correlation (NCC) and a combination of the two. This pipeline produces consistent results not only on outer retinal layers, but also on inner retinal layers such as a nerve fiber layer or images of the vascular complexes, even when images are not of excellent quality.

Retinal Arteriolar Wall Remodeling in Diabetes Captured With AOSLO.

Purpose: Adaptive optics scanning laser ophthalmoscopy (AOSLO) enables the visualization and measurement of the retinal microvasculature structure in humans. We investigated the hypothesis that diabetes mellitus (DM) induces remodeling to the wall structure in small retinal arterioles. These alterations may allow better understanding of vascular remodeling in DM. Methods: We imaged retinal arterioles in one eye of 48 participants (26 with DM and 22 healthy controls) with an AOSLO. Structural metrics of 274 arteriole segments (203 with DM and 71 healthy controls) ≤ 50 µm in outer diameter (OD) were quantified and we compared differences in wall thickness (WT), wall-to-lumen ratio (WLR), inner diameter (ID), OD, and arteriolar index ratio (AIR) between controls and participants with DM. We also compared the individual AIR (iAIR) in groups of individuals. Results: The WLR, WT, and AIRs were significantly different in the arteriole segments of DM participants (P < 0.001). The iAIR was significantly deviated in the DM group (P < 0.001) and further division of the participants with DM into groups revealed that there was an effect of the presence of diabetic retinopathy (DR) on the iAIR (P < 0.001). Conclusions: DM induces remodeling of wall structure in small retinal arterioles and in groups of individuals. The use of AIR allows us to assess remodeling independently of vessel size in the retina and to compute an index for each individual subject. Translational Relevance: High-resolution retinal imaging allows noninvasive assessment of small retinal vessel remodeling in DM that can improve our understanding of DM and DR in living humans.

Imaging fine structures of the human trabecular meshwork in vivo using a custom design goniolens and OCT gonioscopy.

The trabecular meshwork (TM), located within the iridocorneal angle, is a target for many glaucoma treatments aimed at controlling intraocular pressure. However, structural variations between individuals are poorly understood. We propose a newly designed gonioscopic lens optimized for high-resolution imaging to image fine structures of the human TM in vivo. The body of the new lens is index-matched to the human cornea and includes a choice of two gonioscopic mirrors (59° and 63°) and matching air-spaced doublets placed on the anterior surface of the goniolens. The new design allows a diffraction-limited image plane at the iridocorneal angle structures. The goniolens design was built and then placed on the subjects eyes coupled to the cornea with goniogel and a 3D adjustable mount. Images were obtained using a commercially available OCT device (Heidelberg Spectralis). The optical resolution was measured in a model eye as 40.32 and 45.25 cy/mm respectively for each mirror angle. In humans, dense OCT scans with minimum spacing oriented tangential to the iris and ICA were performed on 7 healthy subjects (23-73 yrs). The TM was successfully imaged in all subjects. The custom goniolens improved the contrast of the uveoscleral meshwork structures and corneoscleral meshwork revealing limbus parallel striations, not visible with previous goniolens designs. Transverse OCT images were constructed along the segmentation line, providing an enface image of the TM structures including corneoscleral beams, previously only imaged in vivo using custom adaptive optics systems.

Characterization of the human iridocorneal angle in vivo using a custom design goniolens with OCT gonioscopy.

Intraocular pressure (IOP) is the only modifiable risk factor for glaucoma progression, and many treatments target the trabecular meshwork (TM). Imaging this region in vivo is challenging due to optical limitations of imaging through the cornea at high angles. We propose a gonioscopic OCT approach using a custom goniolens and a commercially available OCT device to improve imaging of the TM, Schlemm's canal (SC) and adjacent structures within the iridocorneal angle (ICA). The goniolens is modified with a plano-convex focusing lens and placed on the eye optically mated with goniogel and aided by a 3D adjustable mount. Gonioscopic OCT volume scans are acquired to image SC. Transverse enface images allowed measurements of SC over a 45° section of the ICA for the first time and revealed locations of SC narrowing. The band of extracanalicular limbal lamina and corneoscleral bands were imaged in most subjects and these bands were confirmed using exterior OCT imaging. The polarization dependence of the visibility of these structures is studied by polarization rotation the OCT beam with a half-wave plate, allowing increased contrast of SC. Gonioscopic OCT has successfully been used to image the human ICA in 3D in vivo. This approach provides more detailed characterization of the TM and SC, enhancing their contrast against their birefringent backgrounds.

Cone Photoreceptors in Diabetic Patients.

Purpose: Cones in diabetic patients are at risk due to metabolic and vascular changes. By imaging retinal vessel modeling at high magnification, we reduced its impact on cone distribution measurements. The retinal vessel images and retinal thickness measurements provided information about cone microenvironment. Methods: We compared cone data in 10 diabetic subjects (28-78 yr) to our published norms from 36 younger and 10 older controls. All subjects were consented and tested in a manner approved by the Indiana University Institutional Review Board, which adhered to the Declaration of Helsinki. Custom adaptive optics scanning laser ophthalmoscopy (AOSLO) was used to image cones and retinal microcirculation. We counted cones in a montage of foveal and temporal retina, using four non-contiguous samples within 0.9-7 deg that were selected for best visibility of cones and least pathology. The data were fit with a two parameter exponential model: ln(cone density) = a * microns eccentricity + b. These results were compared to retinal thickness measurements from SDOCT. Results: Diabetic cone maps were more variable than in controls and included patches, or unusually bright and dark cones, centrally and more peripherally. Model parameters and total cones within the central 14 deg of the macula differed across diabetic patients. Total cones fell into two groups: similar to normal for 5 vs. less than normal for 2 of 2 younger diabetic subjects and 3 older subjects, low but not outside the confidence limits. Diabetic subjects had all retinal vascular remodeling to varying degrees: microaneurysms; capillary thickening, thinning, or bends; and vessel elongation including capillary loops, tangles, and collaterals. Yet SD-OCT showed that no diabetic subject had a Total Retinal Thickness in any quadrant that fell outside the confidence limits for controls. Conclusions: AOSLO images pinpointed widespread retinal vascular remodeling in all diabetic eyes, but the SDOCT showed no increased retinal thickness. Cone reflectivity changes were found in all diabetic patients, but significantly low cone density in only some. These results are consistent with early changes to neural, glial, or vascular components of the retinal without significant retinal thickening due to exudation.

Template free eye motion correction for scanning systems.

Scanning imaging systems are susceptible to image warping in the presence of target motion occurring within the time required to acquire an individual image frame. In this Letter, we introduce the use of a dual raster scanning approach to correct for motion distortion without the need for prior knowledge of the undistorted image. In the dual scanning approach, the target is imaged simultaneously with two imaging beams from the same imaging system. The two imaging beams share a common pupil but have a spatial shift between the beams on the imaging plane. The spatial shift can be used to measure high speed events, because it measures an identical region at two different times within the time required for acquisition of a single frame. In addition, it provides accurate spatial information, since two different regions on the target are imaged simultaneously, providing an undistorted estimate of the spatial relation between regions. These spatial and temporal relations accurately measure target motion. Data from adaptive optics scanning laser ophthalmoscope (AOSLO) imaging of the human retina are used to demonstrate this technique. We apply the technique to correct the shearing of retinal images produced by eye motion. Three control subjects were measured while imaging different retinal layers and retinal locations to qualify the effectiveness of the algorithm. Since the time shift between channels is readily adjustable, this method can be tuned to match different imaging situations. The major requirement is the need to separate the two images; in our case, we used different near infrared spectral regions and dichroic filters.

Alterations to the Foveal Cone Mosaic of Diabetic Patients.

Purpose: We measured localized changes occurring in the foveal cone photoreceptors and related defects in the cone mosaic to alterations in the nearby retinal vasculature. Methods: The central 4° of the retina of 54 diabetic (53.7 ± 12.5 years) and 85 control (35.8 ± 15.2 years) participants were imaged with the Indiana adaptive optics scanning laser ophthalmoscope. Foveal cones and overlying retinal capillaries were imaged and infrared scanning laser ophthalmoscopy (IR SLO) images and optical coherence tomography (OCT) B-scans were obtained. Follow-up imaging sessions were performed with intervals from 4 to 50 months for 22 of the 54 diabetic participants. Results: The foveal cone mosaics of 49 of 54 diabetic participants were of sufficient quality to assess the absence or presence of small localized defects in the cone mosaic. In 13 of these 49 diabetic participants we found localized defects, visualized as sharp-edged areas of cones with diminished reflectivity. These small, localized areas ranged in size from 10 × 10 µm to 75 × 30 µm. Of these 13 participants with cone defects, 11 were imaged over periods from 4 to 50 months and the defects remained relatively stable. These dark regions were not shadows of overlying retinal vessels, but all participants with these localized defects had alterations in the juxtafoveal capillary network. Conclusions: The foveal cone mosaic can show localized areas of dark cones that persist over time, that apparently correspond to either missing or nonreflecting cones, and may be related to local retinal ischemia.