This PDF file contains the front matter associated with SPIE Proceedings Volume 12360, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

Near-infrared Optical Coherence Tomography (OCT) can image bands of alternating reflectivity that delineate the major layers of the retina in living subjects. With the latest technical developments, visible light OCT has taken this capability even further, revealing fine sub-bands and strata within conventional NIR OCT bands. Here we performed visible light OCT study of the retina in a cross-section of mice with varying ages. C57BL/6J, BALB/cJ, and a selection of mutant mice and controls (aged 1.15-20 months) were imaged with 1.0 micrometer axial resolution visible light OCT while under isoflurane anesthesia. Retinal layers and sub-layers were analyzed through a combination of manual and automated segmentation procedures. In agreement with previous studies, we found a decrease in outer nuclear layer (ONL) thickness with age. In addition, we found age-related changes in two closely-associated visible light OCT features. First, a hyporeflective band at the outer edge of the ONL was found to thin with age. Second, a moderately reflective band inner to the ONL was found to thin with age. Histological correlations suggest that outer ONL reflectivity oscillations, or striations, arise from the nuclei being arranged into rows, and that the moderately reflective band inner to the ONL arises from rod spherules. The concomitant age-related thinning of these visible light OCT features, along with the ONL, strengthens these hypothesized associations. These observations also suggest that changes in the organization of the ONL accompany age-related thinning, and that photoreceptor loss can be studied at the level of the soma and at the synapse by visible light OCT in vivo.

As one modality extension of optical coherence tomography (OCT), OCT angiography (OCTA) provides unrivaled capability for depth-resolved visualization of retinal vasculature at the microcapillary level resolution. For OCTA image construction, repeated OCT scans from one location are required to identify blood vessels with active blood flow. The requirement for multi-scan-volumetric OCT can reduce OCTA imaging speed, which will induce eye movements and limit the image field-of-view. In principle, the blood flow should also affect the reflectance brightness profile along the vessel direction in a single-scan-volumetric OCT. In this article, we report a retinal vascular connectivity network (RVC-Net) for deep learning OCTA construction from single-scan-volumetric OCT. We compare the effects of RVC with three adjacent B-scans and a single B-scan input models into RVC-Net. The structural-similarity index measure (SSIM) loss function was selected to optimize deep learning contrast enhancement of microstructures, i.e., microcapillaries, in OCT. This was confirmed by comparing RVC-Net performances with SSIM and mean-squared-error (MSE) loss functions. The involvement of RVC and SSIM loss function enabled microcapillary resolution OCTA construction from singlescan- volumetric OCT.

Retinal pigment epithelial (RPE) cells play an integral role in maintaining visual function and retinal health. Adaptive optics-optical coherence tomography (AO-OCT) has enabled the in vivo visualization of the hexagonal structure of RPE at cellular scale resolution. However, it is difficult to clearly visualize the RPE mosaic in single AO-OCT volumes due to inherent speckle noise, which can be overcome by averaging a large number of AO-OCT volumes acquired at sufficiently spaced time intervals to allow speckle decorrelation for improved cell contrast. Here, we present a deep learning based siamese discriminator (DS) generative adversarial network (GAN) to recover the hexagonal RPE mosaic from only a single AO-OCT volume. The DS provides additional feedback to the generator that resulted in improved visualization of RPE morphology compared to traditional GAN. Experimental results from five healthy subjects suggest that the RPE images generated using DS-GAN were comparable to ground truth images obtained by averaging multiple AO-OCT volumes. Quantitative comparison of cell-to-cell spacing, density, and image quality assessment metrics further confirmed the accuracy of recovered RPE mosaics relative to ground truth. These results establish a potential strategy in which deep learning can be leveraged to eliminate the need for volume averaging and speckle decorrelation for more efficient RPE imaging.

The high resolution of adaptive optics optical coherence tomography (AO-OCT) allows 3-dimensional imaging of individual cone photoreceptors in vivo. Histology has revealed that short-wavelength-sensitive (S) cones have distinct structural features compared with medium-wavelength-sensitive (M) and long-wavelength-sensitive (L) cones. Quantifying these structural features in images of living human retinas may provide a simpler and quicker method for identifying S cones than by imaging cone function (e.g., optoretinography). Here, we present a quantitative method for using AO-OCT measurements of cone structure in a support vector machine (SVM) classifier to identify individual S cones. For every cone cell, we measured six key structural parameters: inner segment length (ISL), outer segment length (OSL), inner segment / outer segment conjunction (IS/OS) diameter, cone outer-segment tip (COST) diameter, IS/OS reflectance, and COST reflectance. ISL and OSL were determined from depth differences between reflections of the external limiting membrane (ELM) and IS/OS, and IS/OS and COST, respectively. Each reflection’s depth was measured with sub-pixel accuracy using Gaussian interpolation; its diameter was measured using the gradient information from the en face projection at that depth. Among 6,398 analyzed cones in six subjects, we found S cones had significantly longer ISLs, shorter OSLs, and wider IS/OS diameters than did cones of other types. We used these structural differences in our SVM model to classify cone spectral types and compared results with cone optoretinography. In five of the six subjects, S cones were identified with F1 scores ranging from 0.78 to 0.93.

Numerous retinal pathologies affect cone photoreceptor photopigment density, making it a potentially attractive functional biomarker for detecting and tracking disease progression. Conventional methods to measure photopigment density include psychophysical color matching, microspectrophotometry, and retinal densitometry, but these are either subjective, measure the aggregate response/change of thousands of cones, or are performed ex vivo. Recently, we have developed a method to measure spectral sensitivities of individual human cone photoreceptors objectively, non-invasively, and in vivo with adaptive optics optical coherence tomography (AO-OCT). In preliminary results we have observed variability in the spectral sensitivities of individual cones of the same type (S, M or L) that we hypothesize attributes to inter-cone variations in photopigment density. If correct, this may be of significant clinical interest. Here, we test this hypothesis by (1) deriving an expression for the relative photopigment densities of individual cone photoreceptors based on a theoretical model of the cone absorption process and (2) using this expression to estimate photopigment density from our AO-OCT measurements of spectral sensitivity. Our mean spectral sensitivity measurements align well to Stockman & Sharpe’s well-recognized cone fundamentals with a total least-squared error of 0.12 and confidence intervals (CI) <0.36, <0.025 and <0.017 for S, M, and L cones, respectively. The substantive variability in individual cone spectral sensitivities once related to photopigment density exhibits a distribution of standard deviation=0.177 for a group of 703 cones. This indicates a two-fold difference in light sensitivity between the least sensitive cone (least amount of photopigment) and the most sensitive cone (largest amount of photopigment) for 95% of the cones measured. Furthermore, we found relative photopigment density decreased with increasing retinal eccentricity from nasal to temporal retina at 3.8° eccentricity with a slope of -0.24/° (p < .001). Both density distribution and eccentricity dependence are consistent with the literature.

In vivo longitudinal monitoring of the inner retinal cellular morphology is of great importance in both clinical and experimental ophthalmology due to its importance in many blinding diseases, including glaucoma, the second leading cause of blindness in the adult population worldwide. At the cellular level, Ganglion Cells (GCs) damage and, ultimately, death often plays a central role in inner retinal disease progression. Recently, advances in OCT-based observation of highly translucent cell somas in the Ganglion Cell Layer (GCL) in both the living human and experimental animal eyes opened the door to noninvasive, label-free monitoring of RGC in vivo. However, a longitudinal validation of OCT's ability to follow individual RGC in experimental animals was still needed. To address this, we will report on our quantitative longitudinal studies in mice lines with fluorescently labeled ganglion cells. Here we used our custom-built mouse retinal Scanning Laser Ophthalmoscopy / Optical Coherence Tomography (SLO/OCT) system to acquire serial OCT volumes (with corresponding SLO intensity and fluorescence data) to provide input for Temporal Speckle Averaging (TSA) OCT volume processing method. To allow in vivo validation of TSA-OCT-based RGC quantification and monitoring, two mouse lines with fluorescently labeled RGC based on RGCs transcription factor (Brn3b-mCherry and Isl2-GFP) have been used and imaged simultaneously with fluorescence SLO (fSLO).

A high-speed adaptive optics (AO) partially-confocal ophthalmoscope using a digital micromirror device (DMD) and high-speed 2D CMOS camera is presented. The system allows for easy control of the trade-off between image acquisition rate and contrast by applying different illumination patterns on the DMD. The camera is synchronized with the DMD projecting multi-spot patterns on the human retina, which is pre-corrected by AO, for parallel scanning. Frame acquisition rates up to 250 fps was achieved this applying multi-spot scheme, with the contrast improving 2-3 fold compared to standard flood illumination. Partially confocal images of the human retina showed cone and rod photoreceptors over a range of retinal eccentricities.

Phase-contrast adaptive optics scanning light ophthalmoscopy (AOSLO) has provided a new way to image transparent retinal cells without exogenous contrast. Building on this approach, we propose a novel strategy to achieve 3D quantitative phase imaging (QPI) which has potential to reveal the cellular geometry, sub-cellular contents and refractive index of translucent cells in the living retina. The approach is based on a working model that harnesses the forward propagation of an illumination beam as it passes through translucent retinal cells and is backscattered by a deeper reflective layer (DRL) such as the photoreceptor-RPE complex. The distance between the illumination focal plane and DRL provides the opportunity to measure the angular scatter/refraction of light as it passes through the cells that reside near the focal plane. Our approach positions an array detector (a digital micromiror device, DMD coupled with a photomultiplier tube) at a plane conjugate with the DRL to capture the angle-resolved, 4D information of the illuminated cells. By measuring the deviation of angular light distribution, phase of the retinal cells can be reconstructed quantitatively. The geometry of this light distribution also encodes depth information of the cells in a way similar to rendering 4D light field imaging. Here, we demonstrate the ability to perform depth ranging, perspective imaging and digital refocusing of various retinal cells and structures in the living retina including immune cells, ganglion cell somata, photoreceptor somata, and microglia using a single 4D data set that captures the angle-resolved steering of light in an AOSLO. Keywords: Adaptive optics, scanning light ophthalmoscopy, retinal imaging, quantitative phase imaging, 3D imaging

Adaptive optics (AO) measures and corrects ocular wavefront aberrations, enabling cellular-resolution retinal imaging and stimulation, and enhanced visual performance. AO is a dynamic control system that must track and correct temporal changes in ocular aberrations in real time. This necessitates a wavefront sensor whose integration time and readout time are sufficiently short to minimize the latency of the feedback system and hence maximize AO performance. Most current ophthalmic AO systems use long wavefront sensor integration times on the order of 10−60 ms, resulting in long latencies, low AO loop rates (typically no greater than 10 Hz with a discontinuous-exposure scheme), and small AO closed-loop bandwidths (less than 1.5 Hz). Here, by using an integration time (0.126 ms) that is 100−500× shorter and a readout speed of the wavefront sensor that is 3−100× higher, we reduce the AO latency and increase the AO bandwidth by ~30× to 37.5 Hz. Although our wavefront sensor detects 100−500× fewer photons, our noise analysis shows that this limited number of photons is still sufficient for diffraction-limited wavefront measurements and hence our wavefront sensing is photon-efficient. We demonstrate that the resulting ultrafast AO running at 233 Hz significantly improves aberration correction and retinal image quality over conventional AO in a clinically-relevant scenario.

In this work we characterize excised human crystalline lenses optically, by measuring objective scattering parameters of cataracts and their Optical Memory Effect. This is completed with a wavefront optimization determining the optimized-image’s isoplanatic patch.

Degree of polarization uniformity (DOPU) provides promising biomarkers of abnormalities in the retinal pigment epithelium (RPE) and is obtained by polarization-sensitive optical coherence tomography (OCT), which is not commercially available. A U-Net shape model was used to synthesize retinal DOPU from OCT with OCT angiography (OCTA) images. Sets of OCT, OCTA, and DOPU images from 175 subjects, 107 subjects, and 30 subjects were used for training, validation, and evaluation, respectively. RPE abnormalities were compared between True DOPU and synthesized DOPU. Healthy structure, RPE elevation, and RPE thickening were synthesized with high recall and precision. However, further improvements are required for RPE defect and hyperreflective retina foci synthesizing.

Adaptive optics optical coherence tomography (AO-OCT) has allowed for the reliable 3-D imaging of individual retinal cells. The current AO-OCT systems are limited to tabletop implementation due to their size and complexity. This work describes the design and implementation of the first dual modality handheld AO-OCT (HAOOCT) and scanning laser ophthalmoscope (SLO) probe to extend AO-OCT imaging to previously excluded patients. Simultaneous SLO imaging allows for tracking of imaging features for HAOOCT localization. Pilot experiments on stabilized and recumbent adults using HAOOCT, weighing only 665 grams, revealed the 3-D photoreceptor structure for the first time using a handheld AO-OCT/SLO device.

Visual simulation is an emerging technology used in ophthalmology where a subject sees through manipulated optical conditions. The existing visual simulation tools are quite advanced and trailblazing. However, they cannot follow the miniaturization and mobility trends of the technology. Current visual simulators are bulky in size and require a tabletop arrangement for operation. Here we propose a novel handheld and portable adaptive optics visual simulator that can induce a variety of optical corrections and measure ocular aberrations in real-time while realizing the form-factor similar to the typical virtual reality headsets. The proposed device has two main parts: a wavefront shaping module for manipulation of visual stimuli and a wavefront sensing module for evaluation of ocular aberrations, which are integrated into a standalone handheld unit. The device incorporates a micro-display and a liquid crystal-on silicon (LCOS) phase modulator for wavefront shaping combined with a Hartmann-Shack (HS) wavefront sensor. Our prototype device incorporates miniature optical components and a path folding mechanism combined with in-house 3D printed mounts and covers to reduce the device footprints. The prototype device is tested by inducing the known values of defocus and astigmatism through a set of trial lenses, measuring the induced aberrations, and evaluating the simulated corrections. Our results show high measurement accuracy (R²>0.999) when tested on spherical and cylindrical trial lenses ranging from -10 to 10 diopters and -5 to 5 diopters, respectively. The visual correction performance shows better than 20/20 visual acuity resolution for the defocus correction of -5 to 5 diopters.

Traditional benchtop OCT systems require upright patient fixation and often impede ophthalmic imaging in bedridden, uncooperative, and pediatric patients. Point-of-care OCT systems have demonstrated ophthalmic imaging in supine patients. However, manually aligning and correcting for refractive power variations between patient eyes to ensure optimal image quality can be clinically cumbersome with point-of-care imaging systems. Here, we demonstrate our improved handheld spectrally encoded coherence tomography and reflectometry (HH-SECTR) imaging probe with mechanical focus-adjust and improved optical throughput in a clinically robust form-factor. SECTR uses spatiotemporally co-registered multimodal spectrally encoded reflectometry (SER) and OCT acquisition for volumetric motion-correction and retinal mosaicking. Our previous HH-SECTR prototype had three major limitations: 1) poor alignment stability caused by reduced mechanical stiffness in a completely rapid-prototyped resin body; 2) lossy SER optical throughput; and 3) manual focus adjust that was cumbersome during point-of-care imaging. Here, we demonstrate optical and optomechanical design improvements to address these limitations, including a modular aluminum probe chassis and increased optical power throughput for sustained system alignment and imaging performance. We have also incorporated a mechanical focusing subsystem to correct refractive errors, which can be integrated with our acquisition software for hands-free focus tracking. We demonstrate in vivo human retinal imaging, and mechanical focusing capabilities using a stationary model eye and stepping through ± 10 diopters focal shift. We predict the addition of focusing capabilities and design improvements in form-factor and optical throughput to our HH-SECTR probe will benefit clinical translation and point-of-care multimodal OCT imaging.

We demonstrated a contact handheld ultra-widefield (UWF) swept-source OCT (SS-OCT) imaging system with a 400 kHz VCSEL light source that achieved an unprecedented 140° field of view (FOV), which was capable to extend the imaging area from the posterior pole to peripheral retina in a single shot. The contact imaging approach provided faster and more efficient retinal imaging and improve image quality. To the best of our knowledge, this prototype achieved the widest FOV among all the retinal OCT research prototypes and commercial systems in desktop and portable format.

In this work, we developed a wearable, head-mounted device that automatically calculates the precise Relative Afferent Pupillary Defect (RAPD) value of a patient. The device consists of two RGB LEDs, two infrared cameras, and one microcontroller. In the RAPD test, the parameters like LED on-off durations, brightness level, and color of the light can be controlled by the user. Upon data acquisition, a computational unit processes the data, calculates the RAPD score and visualizes the test results with a user-friendly interface. Multiprocessing methods used on GUI to optimize the processing pipeline. We have shown that our head-worn instrument is easy to use, fast, and suitable for early-diagnostics and screening purposes for various neurological conditions such as RAPD, glaucoma, asymmetric glaucoma, and anisocoria.

In this work we demonstrate the potential of Vertical-Aligned Spatial Light Modulators in the simulation and correction of High Order Aberrations by testing visual acuity and contrast sensitivity function in a set of healthy subjects.

The laser systems currently used in ophthalmology either have some pulse length dependent side effects or are very expensive due to their complexity. Therefore, a newly developed approach using picosecond laser sources is investigated. These lasers combine the advantages of the low price of currently used short-pulse laser sources with the cold material ablation possibilities of high-end femtosecond sources. The surgeries intended are laser iridotomy, capsulotomy/post-cataract treatment and selective laser-trabeculoplasty (SLT). They are demonstrated on post mortem porcine eyes. The result is a more precise, less frayed tissue ablation with picosecond pulses in comparison to nanosecond pulses. The pulse energy could be reduced to (50 20) µJ per pulse instead of 1mJ to 10mJ per pulse, which is currently applied. The study of shock waves and cavitation bubbles revealed a huge difference in pressure between picosecond pulses (0:25MPa at 50 µJ) and nanosecond pulses (37MPa at 5 mJ). Therefore, the risk of collateral damage leading to potential additional clinical patterns and adverse effects could be significantly reduced.

In this paper, we describe how using a liquid lens can improve the quality of iris images acquired by a hyperspectral system. This improvement in the image quality is especially noticeable for systems that scan the iris over a wide range of wavelengths, e.g. visible and near-infrared spectrum. We have tested this approach on the previously developed system able to acquire iris images in the spectral range 480 - 900 nm. The key novelty presented in this paper is represented by the possibility of adaptively adjusting the focus of the imaging system, allowing for chromatic aberration compensation and ensuring a constant image sharpness among all wavelengths. A fast-tunable liquid lens has been placed in front of the chromatically corrected camera objective to adaptively change the overall focus of the imaging system. The findings imply that the device can rapidly perform hyperspectral measurements of the iris over a broad wavelength range ensuring optimal focus for all images.

Fluorescein video angiographies (FVAs) are a diagnostic tool for eye diseases, such as diabetic retinopathy (DR). Currently, kinetic tracer model methods based on indicator-dilutions theory use FVAs to extract biomarkers (e.g., volumetric blood flow and retinal vascular permeability) via pixel mapping using two-step non-linear least square fitting. Prior to biomarker extraction, the FVAs must attain optimal quality. The objective of this research is to create a program to remove all frames experiencing signal drops (causes include blinking, squinting, and head movement). 15 FVAs (6 healthy control subjects, 6 diabetes mellitus no DR (DMnoDR) subjects, and 3 mild non-proliferative DR (NPDR) subjects) were analyzed for low quality frames. The average signal of each frame was analyzed as top, middle, and bottom thirds. The frame with maximum average signal up to the final frame of a created “Gold Standard” was compared with the raw AVI’s frame with maximum average signal and subsequent frames. All frames before maximum average signal and any remaining frames were compared with the previous good-quality raw frame to determine if the frame of interest was of good quality. All remaining frames were subsequently re-evaluated and flagged if they had a local minimum prominence of 10% of the maximum average signal. The flagged frames’, as well as former and subsequent frames’, quality were subjectively determined. The AVI quality was subsequently tested via pre-DTKM processing and biomarker extraction via DTKM methods. Results displayed that the semi-automated frame removal process provides sufficient quality AVIs.

We report a large field of view FOV low-cost smart fundus camera with non-mydriatic and spectral band illumination features for the purpose of diagnosing central nervous system diseases such as Alzheimer’s disease. By combining a customized optical lens group, ring light emitting diode light source, Li-battery, and Raspberry Pi we can port the spectrally selected light source through a 4 mm diameter pupil. Taking advantage of the open hardware platform and Linux operating system, we integrate two narrowband filters explicitly selected based on the unique spectrums of biomarkers (580 nm and 660 nm) with a Raspberry Pi camera module to obtain high quality images for the purpose of enhancing the visibility of retinal vasculature and nerve fiber layer.

Light-evoked functional retinal imaging is of great interest in clinical and experimental ophthalmology because the alterations in retinal function hold the promise of being potentially more sensitive for disease diagnosis than purely morphology-based assays. Recent progress in Optoretinography (ORG), a technique employing OCT to extract light-evoked functional response of retinal tissue, promises to provide the needed sensitivity to probe early alterations in retinal function. The most promising implementations of the ORG probing of photoreceptor function in clinical settings are using parallel OCT detection schemes such as Line Field and Full Field designs. Herein, we present the instrumentation scheme of a Full-Field-Swept-Source OCT (FF-SS-OCT) system incorporating a high-precision light stimulation channel for facilitating phase-based functional retinal imaging in mice. The assessment of the OCT signal phase errors and their correction using short-time Fourier transform (STFT) is detailed. The performance of the system is investigated using a model eye.

Chemical injuries to the cornea account for 11 to 22% of all ocular injuries. Acidic injuries are commonly due to sulfuric, hydrochloric, hydrofluoric, and battery acids, while basic injuries are commonly due to sodium hydroxide, chlorine bleach, and ammonia products. We have previously studied potential-driven electrochemical clearing (P-ECC) for alkaline injuries. In this study, we investigated the use of P-ECC on both acidic and alkaline injuries to determine its effect on restoring corneal transparency. Optical coherence tomography (OCT) was performed before and after P-ECC to determine adequate corneal clearing. Severity of chemical injury was measured through second harmonic generation (SHG) imaging. HCl or NaOH was applied to the corneas of New Zealand white rabbit globes. P-ECC was performed on opacified cornea while OCT imaging was simultaneously performed to evaluate depth resolved clarity. SHG imaging evaluated the structure of collagen before HCl or NaOH application and after P-ECC. Irrigation with water served as positive control. Native rabbit corneas were used as a negative control group. P-ECC induced clearing in the rabbit cornea, shown through OCT. Clearing occurred in regions where the working electrode made contact with the cornea. SHG imaging showed restoration of collagen fibril signal in P-ECC treated corneas compared to control. P-ECC is a potentially effective therapy for clearing acidic and alkaline corneal injuries. However, more ex-vivo experiments are required to determine the specific parameter for optimal clearing. In-vivo experiments are necessary to determine its potential for clinical use.

It is estimated that by the year 2050, approximately 50.0 % of the world’s population will be myopic. It has been suggested that the thickening sub-foveal choroidal thickness (SFChT) is a precursor for reduced eye growth and slowed myopia progression. Hence, it is highly important to identify structural changes, during myopia management. Literature suggests that the SFChT show short-term changes and has been proposed as an ocular marker for many ocular conditions including pathological myopia. A major limitation to its use is that none of the commercially available instruments give a direct measure of SFChT. This paper describes a new semi-automated method for quantification of the SFChT using ocular biometry. This image processing method is used on healthy pediatric myopes to quantify the SFChT. Both axial length (AXL) and SFChT were quantified from a 2-D A-scan graph generated from an ocular biometer (ARGOS, Aichi, Japan). An experienced clinician manually selected three peak points corresponding to anterior corneal, retinal, and choroidal peaks which were used as the input to the algorithm. Using the pixel properties, the overall AXL was calculated by subtracting the distance between the anterior corneal and retinal peak. Similarly, the SFChT was calculated as the distance difference between retinal and choroidal peaks. These calculated values were compared with a standard clinical method. The intraclass correlation coefficient ICC showed a good (κ=0.79, CI: 0.63 – 0.87) agreement between the methods. Similarly, the Bland-Altman plot showed a good agreement (The mean difference between the two methods was -21.28 μm) with a wide limit of agreement (LOA: 79.08 & -121.64 μm). Compared to SS-OCT method and new-semi automated method, there is significant overestimation of SFChT (p<0.001). However, there is moderate agreement between these two methods.

We developed a prototype device for dynamic gaze and accommodation measurements based on 4 Purkinje reflections (PR) suitable for use in AR and ophthalmology applications. PR1&2 and PR3&4 are used for accurate gaze and accommodation measurements, respectively. Our eye-model was developed in Zemax and matches the experiments well. Our model predicts the accommodation from 25cm to infinity (<4 diopters) with better than 0,25D accuracy. We performed repeatability tests and obtained accurate gaze and accommodation estimations using 15 subjects. We are generating a large synthetic data set using physically accurate models and machine learning algorithms.

Myopia is the most common ocular disorder worldwide and the leading cause of visual impairment in children. Its progression affects the normal retinal development, increasing the risk of suffering severe ocular pathologies that lead to irreversible blindness. Although the mechanism for myopia progression is not completely elucidated yet, it is suggested that the quality of the peripheral retinal images may play a role. This produced an enormous interest in using spectacle lenses with different focusing properties for foveal and peripheral vision as a treatment to slow myopia progression. In this work, we developed a novel instrument to optically characterise spectacle lenses designed for myopia control under realistic viewing conditions. First, the testing beam impinges the spectacle under test with different eccentricities. For each eccentricity, a steering mirror guides the beam towards a liquid crystal spatial light modulator (SLM). The SLM allows for reproducing several pupil sizes, shaping the pupil according to the eccentricity, reproducing ocular aberrations for different levels of myopia, and acquiring the through-focus point spread functions (PSFs). An electrically tunable liquid lens, conjugated to the SLM, compensates the base power of the tested lens and focuses the beam into a camera’s sensor. Three spectacle lenses were tested, and different optical metrics were calculated from the PSF and the modulation transfer function.

A convolutional neural network (CNN) with multimodal fusion options was developed for artery-vein (AV) segmentation in OCT angiography (OCTA). We quantitatively evaluated multimodal architectures with early and late OCT-OCTA fusions, compared to the unimodal architectures with OCT-only and OCTA-only inputs. OCT-only architecture is limited for segmentation of large AV branches. The OCTA-only architecture, early OCT-OCTA fusion architecture, and late OCT-OCTA fusion architecture provide competitive performances for AV segmentation with further details. Compared to OCTA-only architecture, the late fusion architecture is slightly better, while the early fusion architecture is slightly worse.

The Electroretinogram measures the overall electrical activity of the retina in response to light stimulation, while the dynamics of Pupillary Light Reflex reveal information about how the visual pathway innervates the iris muscles in response to such stimuli. By simultaneously evaluating PLR and ERG responses, a deeper understanding of both image-forming and non-image-forming mechanisms of the human visual system can be gained. Additionally, ERG and PLR responses to bursts of light are contributed by all primary classes of photoreceptors, including rods, (L-M-S) cones, and intrinsically photosensitive ganglion cells (ipRGCs). This study presents and tests a low-cost, Maxwellian view-based optical setup that can be used to acquire synchronous PLR and ERG recordings with silent stimulation techniques on cones photoreceptors in human subjects.

Fundus imaging is indispensable for clinical management of eye diseases such as diabetic retinopathy (DR), age-related macular degeneration (AMD), and glaucoma. Emerging portable and smartphone fundus cameras hold promise to advance telemedicine ophthalmology in home and family care environments. However, currently available portable fundus cameras have limited field of view (FOV). Miniaturized indirect ophthalmoscopy illumination has been recently demonstrated as a feasible solution to expand the FOV in portable fundus cameras. However, for the indirect illumination, the light reflected from the ophthalmic lens causes image artifacts. Moreover, the image contrast of a local fundus region can be limited as the light efficiency varies significantly among different regions, such as the optic nerve head and central fovea. We report here a portable, non-mydriatic, reflectance artifact free, high dynamic range (HDR) fundus camera with a 67° visual-angle (101° eye-angle) snapshot FOV. Orthogonal polarization control was used to eliminate illumination reflectance artifact due to the ophthalmic lens back-reflection. With independent power control, three sequential fundus images were captured within the pupillary reflex time. These three images, with different local features enhanced, are registered, and fused together to get the HDR image. The output HDR images shows better biomarker contrast compared to input low dynamic range (LDR) images.

Nystagmus is a periodic, involuntary movement of eyes examined for diagnosis of various vestibular diseases such as benign paroxysmal positional vertigo, the most frequent vestibular disorder. In recent years, videonystagmography has been widely used in the examination of nystagmus due to its non-invasive feature. However, identifying and classifying nystagmus still requires professional knowledge and training. To this end, a pupil tracking algorithm was proposed in this paper using convolutional neural networks. U-Net was selected for pupil segmentation, and we constructed the ground truth of a new dataset for the training procedure. An additional tracking algorithm was designed to prevent false outputs of the U-Net model. Results show that the proposed pupil tracking algorithm scored higher performance than conventional methods.

Cataract is a common ophthalmic disease in which a cloudy area is formed in the lens of the eye and requires surgical removal and replacement of eye lens. Careful selection of the intraocular lens (IOL) is critical for the post-surgery satisfaction of the patient. Although there are various types of IOLs in the market with different properties, it is challenging for the patient to imagine how they will perceive the world after the surgery. We propose a novel holographic vision simulator which utilizes non-cataractous regions on eye lens to allow the cataract patients to experience post-operative visual acuity before surgery. Computer generated holography display technology enables to shape and steer the light beam through the relatively clear areas of the patient’s lens. Another challenge for cataract surgeries is to match the right patient with the right IOL. To evaluate various IOLs, we developed an artificial human eye composed of a scleral lens, a glass retina, an iris, and a replaceable IOL holder. Next, we tested different IOLs (monofocal and multifocal) by capturing real-world scenes to demonstrate visual artifacts. Then, the artificial eye was implemented in the benchtop holographic simulator to evaluate various IOLs using different light sources and holographic contents.

In aqueous environments crosslinked hydrophobic acrylic intraocular lenses (IOL) can form fluid-filled microvacuoles (MV). The water inclusions lead to reflection and scattering of light at the interface of the MV which causes the IOLs to become cloudy. This phenomenon is called glistening. A new approach to reduce glistening is to decrease the free volume and therefore the space for water inclusions in acrylic hydrophobic materials. By using certain types and amounts of crosslinkers, the size and density of MVs can be controlled to eliminate the glistening phenomenon without using hydrophilic comonomers. The procedure has only a slight influence on other physical parameters and can even improve them with the right choice of structure and amount of the crosslinker.

Injection of fluorescent dye is a safety concern in fluorescein angiography (FA). This has led to the cautious use of this clinical diagnostic modality in certain populations (e.g., children, allergic populations). In recent years, the development of non-invasive functional imaging of fundus blood flow by computational means has become a hot spot in ophthalmic research, such as OCT angiography. Deep learning-based color fundus to FA prediction is another emerging approach, which takes advantage of the nonlinear and high-dimensional mapping capabilities of deep neural networks to establish the relationship of these two imaging modalities explicitly. Most of such studies use a small publicly-available dataset and rely on algorithm design to improve the prediction accuracy. However, the limited performance has attracted little attention and raised doubts about its viability. Here, we show that the prediction accuracy can be significantly improved by simply expanding the training dataset by a factor of ~10 without introducing new algorithms. While this result is expected based on the nature of the data-driven model, it suggests that the development of such deep learning-based prediction requires a more diverse approach rather than focusing only on algorithmic improvements.

Publisher's Note: This paper, originally published on 14 March 2023, was replaced with a corrected/revised version on 5 May 2023. If you downloaded the original PDF but are unable to access the revision, please contact SPIE Digital Library Customer Service for assistance.

Publisher's Note: This paper, originally published on 14 March 2023, was replaced with a corrected/revised version on 5 January 2024. If you downloaded the original PDF but are unable to access the revision, please contact SPIE Digital Library Customer Service for assistance.