A mathematical description of nerve fiber bundle trajectories and their variability in the human retina

Spatially Selective Retinal Ganglion Cell Activation Using Low Invasive Extraocular Temporal Interference Stimulation

Conventional retinal implants involve complex surgical procedures and require invasive implantation. Temporal Interference Stimulation (TIS) has achieved noninvasive and focused stimulation of deep brain regions by delivering high-frequency currents with small frequency differences on multiple electrodes. In this study, we conducted in silico investigations to evaluate extraocular TIS's potential as a novel visual restoration approach. Different from the previously published retinal TIS model, the new model of extraocular TIS incorporated a biophysically detailed retinal ganglion cell (RGC) population, enabling a more accurate simulation of retinal outputs under electrical stimulation. Using this improved model, we made the following major discoveries: (1) the maximum value of TIS envelope electric potential ([Formula: see text] showed a strong correlation with TIS-induced RGC activation; (2) the preferred stimulating/return electrode (SE/RE) locations to achieve focalized TIS were predicted; (3) the performance of extraocular TIS was better than same-frequency sinusoidal stimulation (SSS) in terms of lower RGC threshold and more focused RGC activation; (4) the optimal stimulation parameters to achieve lower threshold and focused activation were identified; and (5) spatial selectivity of TIS could be improved by integrating current steering strategy and reducing electrode size. This study provides insights into the feasibility and effectiveness of a low-invasive stimulation approach in enhancing vision restoration.

Standard automated perimetry for glaucoma and diseases of the retina and visual pathways: Current and future perspectives

Static automated perimetry (SAP) remains a mainstay of functional assessment of the visual field in diseases of the visual pathway, such as glaucoma and age-related macular degeneration. The fundamental psychophysical task of responding to stimuli of different levels of contrast has remained minimally changed since its inception in the 1980s, and this is potentially the root of several unresolved issues involving the technique. Enduring issues include the optimisation of SAP parameters for maximising defect detection, the influence of subjective behaviour on the response, structure-function discordance, and ageing- and disease-related changes of the visual pathway. Addressing these issues has been a focus of our research program and is the subject of this manuscript. We will review some of the basic psychophysical principles and methods that have contributed to the development of SAP and their contributions to its output measurements. Parameters that are interrogated include stimulus size and background luminance and their modification to improve defect defection in glaucoma and age-related macular degeneration. We propose frameworks for optimising testing parameters and leveraging the results for changing clinical care. In our pursuit of optimising the structure-function relationship in the eye, several areas of research have been developed and explored, including: the reconciliation of subjective responses in perimetry; by minimising sources of biases, such as Method of Limits we have been able to equate static and kinetic perimetry outputs in relation to underlying structural loci. This also formed the basis for our clustering framework, which groups together statistically similar structural and functional test locations to maximise structure-function concordance. Throughout the manuscript, we review the scientific underpinnings of clinical measurements, framing application into real-world patients to improve clinical practice.

A computational model to design wide field-of-view optic nerve neuroprostheses

Retinal stimulation (RS) allows restoring vision in blind patients, but it covers only a narrow region of the visual field. Optic nerve stimulation (ONS) has the potential to produce visual perceptions spanning the whole visual field, but it produces very irregular phosphenes. We introduced a geometrical model converting retinal and optic nerve firing rates into visual perceptions and vice versa and a method to estimate the best perceptions elicitable through an electrode configuration. We then compared in silico ONS and RS through simulated prosthetic vision of static and dynamic visual scenes. Both simulations and SPV experiments showed that it might be possible to reconstruct natural visual scenes with ONS and RS, and that ONS wide field-of-view allows the perception of more detail in dynamic scenarios than RS. Our findings suggest that ONS could represent an interesting approach for vision restoration and that our model can be used to optimize it.

Supervised Automated Kinetic Perimetry (SAKP) Using Simulated Visual Field Data - Presentation of a New Examination Technique

PURPOSE

The aim of this study was to develop, optimise, train, and evaluate an algorithm for performing Supervised Automated Kinetic Perimetry (SAKP) using digitalised perimetric simulation data.

METHODS

The original SAKP algorithm was based on findings from a multicentre study to establish reference values by semi-automated kinetic perimetry (SKP) combined with an automated examination method with moving stimuli ("Program K", developed in Japan). The algorithm evaluated the outer angles of isopter segments and responded to deviations from expected values by placing examination vectors to measure the outer boundaries of the visual field (VF). Specialised interpolation methods were also used to create individual 3D hills of vision and local "probing vectors" to optimise the eccentricity of the vector origins. This algorithm was trained iteratively on seven representative digitalised 3D VF results from five typical classes and optimised in each step: (1) Normal VF, (2) Central scotoma, (3) Concentric VF constriction, (4) Retinal nerve fibre layer defects in the visual field (VFDs), (5) VFDs with respect to the vertical meridian. The optimised SAKP algorithm was then applied to a new set of twenty 3D VF results of varying origin and severity. The primary targets were measured in agreement between actual calculated VF expressed as accuracy (A), that is, the ratio between the area containing correct predictions and total area of predictions measured between 0 = worst and 1 = best, and examination duration (T). The results are given as median (and interquartile range). We also verified the test's robustness by varying individual error rates (ERs) and error magnitudes (EMs).

RESULTS

The median and interquartile range (IQR, in brackets) for the total of representative VFs were 0.93 (0.02) for A and 7.0 min (5.2 min) for T, respectively. A gave the best result for altitudinal VFDs and VFDs with hemianopic character and macular sparing (0.98 each) and worst in superior wedge-shaped VFDs (0.78); T was shortest in blind spot displacement (3.9 min) and longest in hemianopic VFDs with hemianopic character and macular sparing with preserved temporal crescent (12.1 min). Error rate and magnitude (up to 30% each) only showed a comparatively low influence on A and T.

CONCLUSION

The SAKP algorithm presented here achieves a comparatively high degree of accuracy and robustness for actual, simulated visual field data within acceptable examination times. This algorithm is currently being prepared for application in real patient examinations under clinical conditions.

Dysfunction and Morphological Involvement of Inner Macular Layers in Glaucoma

Objectives: This study aimed to study the inner retina functional and morphological impairment of retinal ganglion cells (RGCs) from specific macular rings and sectors to identify whether selective macular regions were more vulnerable than others within the 20 central degrees in patients with open-angle glaucoma (OAG). Methods: In total, 21 OAG patients [mean age 50.19 ± 7.86 years, Humphrey Field Analyzer (HFA) 24-2 mean deviation (MD) between -5.02 and -22.38 dB, HFA 10-2 MD between -3.07 and -17.38 dB], providing 21 eyes, were enrolled in this retrospective case-control study. And 20 age-similar healthy subjects, providing 20 eyes, served as controls. The multifocal photopic negative response (mfPhNR) response amplitude density (RAD) from concentric rings and macular sectors and ganglion cell layer thickness (GCL-T) assessed by Spectral Domain-Optical Coherence Tomography (SD-OCT) was measured. Mean RAD and GCL-T values were compared between OAG and control ones by ANOVA. In OAG eyes, the relationship between mfPhNR and SD-OCT data was examined by linear regression analysis, and Pearson's correlation coefficients were computed. Results: In considering all rings and sectors, compared to the controls, the OAG group showed a significant (p < 0.01) reduction in mean mfPhNR RAD and in GCL-T values with the greatest reduction in the central area. In OAG eyes, a significant (p < 0.01) correlation between all mfPhNR RAD and GCL-T values, with significant (p < 0.01) correlation coefficients, were found. Conclusions: In OAG eyes, RGC dysfunction was detectable by abnormal mfPhNR responses in localized macular areas, mainly in the central one. Localized macular RGC dysfunction was linearly correlated with the GCL morphological changes.

Wide-angle simulated artificial vision enhances spatial navigation and object interaction in a naturalistic environment

Objective. Vision restoration approaches, such as prosthetics and optogenetics, provide visual perception to blind individuals in clinical settings. Yet their effectiveness in daily life remains a challenge. Stereotyped quantitative tests used in clinical trials often fail to translate into practical, everyday applications. On the one hand, assessing real-life benefits during clinical trials is complicated by environmental complexity, reproducibility issues, and safety concerns. On the other hand, predicting behavioral benefits of restorative therapies in naturalistic environments may be a crucial step before starting clinical trials to minimize patient discomfort and unmet expectations.Approach. To address this, we leverage advancements in virtual reality technology to conduct a fully immersive and ecologically valid task within a physical artificial street environment. As a case study, we assess the impact of the visual field size in simulated artificial vision for common outdoor tasks.Main results. We show that a wide visual angle (45°) enhances participants' ability to navigate and solve tasks more effectively, safely, and efficiently. Moreover, it promotes their learning and generalization capability. Concurrently, it changes the visual exploration behavior and facilitates a more accurate mental representation of the environment. Further increasing the visual angle beyond this value does not yield significant additional improvements in most metrics.Significance. We present a methodology combining augmented reality with a naturalistic environment, enabling participants to perceive the world as patients with retinal implants would and to interact physically with it. Combining augmented reality in naturalistic environments is a valuable framework for low vision and vision restoration research.

Proposing a Methodology for Axon-Centric Analysis of IOP-Induced Mechanical Insult

Purpose

IOP-induced mechanical insult on retinal ganglion cell axons within the optic nerve head (ONH) is believed to be a key factor in axonal damage and glaucoma. However, most studies focus on tissue-level mechanical deformations, overlooking that axons are long and thin, and that their susceptibility to damage likely depends on the insult's type (e.g. stretch/compression) and orientation (longitudinal/transverse). We propose an axon-centric approach to quantify IOP-induced mechanical insult from an axon perspective.

Methods

We used optical coherence tomography (OCT) scans from a healthy monkey eye along with histological images of cryosections to reconstruct the axon-occupied volume including detailed lamina cribrosa (LC) pores. Tissue-level strains were determined experimentally using digital volume correlation from OCT scans at baseline and elevated IOPs, then transformed into axonal strains using axon paths estimated by a fluid mechanics simulation.

Results

Axons in the LC and post-LC regions predominantly experienced longitudinal compression and transverse stretch, whereas those in the pre-LC and ONH rim mainly suffered longitudinal stretch and transverse compression. No clear patterns were observed for tissue-level strains.

Conclusions

Our approach allowed discerning axonal longitudinal and transverse mechanical insults, which are likely associated with different mechanisms of axonal damage. The technique also enabled quantifying insult along individual axon paths, providing a novel link relating the retinal nerve fiber layer and the optic nerve through the LC via individual axons. This is a promising approach to establish a clearer connection between IOP-induced insult and glaucoma. Further studies should evaluate a larger cohort.

Deep-Learning-Based Group Pointwise Spatial Mapping of Structure to Function in Glaucoma

Purpose

To establish generalizable pointwise spatial relationship between structure and function through occlusion analysis of a deep-learning (DL) model for predicting the visual field (VF) sensitivities from 3-dimensional (3D) OCT scan.

Design

Retrospective cross-sectional study.

Participants

A total of 2151 eyes from 1129 patients.

Methods

A DL model was trained to predict 52 VF sensitivities of 24-2 standard automated perimetry from 3D spectral-domain OCT images of the optic nerve head (ONH) with 12 915 OCT-VF pairs. Using occlusion analysis, the contribution of each individual cube covering a 240 × 240 × 31.25 μm region of the ONH to the model's prediction was systematically evaluated for each OCT-VF pair in a separate test set that consisted of 996 OCT-VF pairs. After simple translation (shifting in x- and y-axes to match the ONH center), group t-statistic maps were derived to visualize statistically significant ONH regions for each VF test point within a group. This analysis allowed for understanding the importance of each super voxel (240 × 240 × 31.25 μm covering the entire 4.32 × 4.32 × 1.125 mm ONH cube) in predicting VF test points for specific patient groups.

Main Outcome Measures

The region at the ONH corresponding to each VF test point and the effect of the former on the latter.

Results

The test set was divided to 2 groups, the healthy-to-early-glaucoma group (792 OCT-VF pairs, VF mean deviation [MD]: -1.32 ± 1.90 decibels [dB]) and the moderate-to-advanced-glaucoma group (204 OCT-VF pairs, VF MD: -17.93 ± 7.68 dB). Two-dimensional group t-statistic maps (x, y projection) were generated for both groups, assigning related ONH regions to visual field test points. The identified influential structural locations for VF sensitivity prediction at each test point aligned well with existing knowledge and understanding of structure-function spatial relationships.

Conclusions

This study successfully visualized the global trend of point-by-point spatial relationships between OCT-based structure and VF-based function without the need for prior knowledge or segmentation of OCTs. The revealed spatial correlations were consistent with previously published mappings. This presents possibilities of learning from trained machine learning models without applying any prior knowledge, potentially robust, and free from bias.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Maculopapillary Bundle Degeneration in Optic Neuropathies

PURPOSE OF REVIEW

Degeneration of the maculopapillary bundle (MPB) is a prominent feature in a spectrum of optic neuropathies. MPB-selective degeneration is seen in specific conditions, such as nutritional and toxic optic neuropathies, Leber hereditary optic neuropathy (LHON), and dominant optic atrophy (DOA). Despite their distinct etiologies and clinical presentations, which encompass variations in age of incidence and monocular or binocular onset, these disorders share a core molecular mechanism: compromised mitochondrial homeostasis. This disruption is characterized by dysfunctions in mitochondrial metabolism, biogenesis, and protein synthesis. This article provides a comprehensive understanding of the MPB's role in optic neuropathies, emphasizing the importance of mitochondrial mechanisms in the pathogenesis of these conditions.

RECENT FINDINGS

Optical coherence tomography studies have characterized the retinal nerve fiber layer changes accompanying mitochondrial-affiliated optic neuropathies. Selective thinning of the temporal optic nerve head is preceded by thickening in early stages of these disorders which correlates with reductions in macular ganglion cell layer thinning and vascular atrophy. A recently proposed mechanism underpinning the selective atrophy of the MPB involves the positive feedback of reactive oxygen species generation as a common consequence of mitochondrial dysfunction. Additionally, new research has revealed that the MPB can undergo degeneration in the early stages of glaucoma, challenging the historically held belief that this area was not involved in this common optic neuropathy. A variety of anatomical risk factors influence the propensity of glaucomatous MPB degeneration, and cases present distinct patterns of ganglion cell degeneration that are distinct from those observed in mitochondria-associated diseases. This review synthesizes clinical and molecular research on primary MPB disorders, highlighting the commonalities and differences in their pathogenesis.

KEY POINTS (BOX)

1. Temporal degeneration of optic nerve fibers accompanied by cecocentral scotoma is a hallmark of maculopapillary bundle (MPB) degeneration. 2. Mechanisms of MPB degeneration commonly implicate mitochondrial dysfunction. 3. Recent research challenges the traditional belief that the MPB is uninvolved in glaucoma by showing degeneration in the early stages of this common optic neuropathy, yet with features distinct from other MPB-selective neuropathies. 4. Reactive oxygen species generation is a mechanism linking mitochondrial mechanisms of MPB-selective optic neuropathies, but in-vivo and in-vitro studies are needed to validate this hypothesis.

The structure-function relationship between multifocal pupil perimetry and retinal nerve fibre layer in glaucoma

BACKGROUND

Multifocal pupillographic objective perimetry (mfPOP) is a novel method for assessing functional change in diseases like glaucoma. Previous research has suggested that, in contrast to the pretectally-mediated melanopsin response of intrinsically photosensitive retinal ganglion cells, mfPOP responses to transient onset stimuli involve the extrastriate cortex, and thus the main visual pathway. We therefore investigate the correlation between peripapillary retinal nerve fibre layer (pRNFL) thickness and glaucomatous visual field changes detected using mfPOP. Parallel analyses are undertaken using white on white standard automated perimetry (SAP) for comparison.

METHODS

Twenty-five glaucoma patients and 24 normal subjects were tested using SAP, 3 mfPOP variants, and optical coherence tomography (OCT). Arcuate clusters of the SAP and mfPOP deviations were weighted according to their contribution to published arcuate divisions of the retinal nerve fibre layer. Structure-function correlation coefficients (r) were computed between pRNFL clock-hour sector thickness measurements, and the local visual field sensitivities from both SAP and mfPOP.

RESULTS

The strongest correlation was observed in the superior-superotemporal disc sector in patients with worst eye SAP MD < -12 dB: r = 0.93 for the mfPOP LumBal test (p < 0.001). Correlations across all disc-sectors were strongest in these same patients in both SAP and mfPOP: SAP r = 0.54, mfPOP LumBal r = 0.55 (p < 0.001). In patients with SAP MD ≥ -6 dB in both eyes, SAP correlations across all sectors were higher than mfPOP; mfPOP correlations however, were higher than SAP in more advanced disease, and in normal subjects.

CONCLUSIONS

For both methods the largest correlations with pRNFL thickness corresponded to the inferior nasal field of more severely damaged eyes. Head-to-head comparison of mfPOP and SAP showed similar structure-function relationships. This agrees with our recent reports that mfPOP primarily stimulates the cortical drive to the pupils.

Axonal stimulation affects the linear summation of single-point perception in three Argus II users

Objective.Retinal implants use electrical stimulation to elicit perceived flashes of light ('phosphenes'). Single-electrode phosphene shape has been shown to vary systematically with stimulus parameters and the retinal location of the stimulating electrode, due to incidental activation of passing nerve fiber bundles. However, this knowledge has yet to be extended to paired-electrode stimulation.Approach.We retrospectively analyzed 3548 phosphene drawings made by three blind participants implanted with an Argus II Retinal Prosthesis. Phosphene shape (characterized by area, perimeter, major and minor axis length) and number of perceived phosphenes were averaged across trials and correlated with the corresponding single-electrode parameters. In addition, the number of phosphenes was correlated with stimulus amplitude and neuroanatomical parameters: electrode-retina and electrode-fovea distance as well as the electrode-electrode distance to ('between-axon') and along axon bundles ('along-axon'). Statistical analyses were conducted using linear regression and partial correlation analysis.Main results.Simple regression revealed that each paired-electrode shape descriptor could be predicted by the sum of the two corresponding single-electrode shape descriptors (p < .001). Multiple regression revealed that paired-electrode phosphene shape was primarily predicted by stimulus amplitude and electrode-fovea distance (p < .05). Interestingly, the number of elicited phosphenes tended to increase with between-axon distance (p < .05), but not with along-axon distance, in two out of three participants.Significance.The shape of phosphenes elicited by paired-electrode stimulation was well predicted by the shape of their corresponding single-electrode phosphenes, suggesting that two-point perception can be expressed as the linear summation of single-point perception. The impact of the between-axon distance on the perceived number of phosphenes provides further evidence in support of the axon map model for epiretinal stimulation. These findings contribute to the growing literature on phosphene perception and have important implications for the design of future retinal prostheses.

Multifunctional adaptive optics optical coherence tomography allows cellular scale reflectometry, polarimetry, and angiography in the living human eye

Clinicians are unable to detect glaucoma until substantial loss or dysfunction of retinal ganglion cells occurs. To this end, novel measures are needed. We have developed an optical imaging solution based on adaptive optics optical coherence tomography (AO-OCT) to discern key clinical features of glaucoma and other neurodegenerative diseases at the cellular scale in the living eye. Here, we test the feasibility of measuring AO-OCT-based reflectance, retardance, optic axis orientation, and angiogram at specifically targeted locations in the living human retina and optic nerve head. Multifunctional imaging, combined with focus stacking and global image registration algorithms, allows us to visualize cellular details of retinal nerve fiber bundles, ganglion cell layer somas, glial septa, superior vascular complex capillaries, and connective tissues. These are key histologic features of neurodegenerative diseases, including glaucoma, that are now measurable in vivo with excellent repeatability and reproducibility. Incorporating this noninvasive cellular-scale imaging with objective measurements will significantly enhance existing clinical assessments, which is pivotal in facilitating the early detection of eye disease and understanding the mechanisms of neurodegeneration.

Learning to see via epiretinal implant stimulationin silicowith model-based deep reinforcement learning

OBJECTIVE

Diseases such as age-related macular degeneration and retinitis pigmentosa cause the degradation of the photoreceptor layer. One approach to restore vision is to electrically stimulate the surviving retinal ganglion cells with a microelectrode array such as epiretinal implants. Epiretinal implants are known to generate visible anisotropic shapes elongated along the axon fascicles of neighboring retinal ganglion cells. Recent work has demonstrated that to obtain isotropic pixel-like shapes, it is possible to map axon fascicles and avoid stimulating them by inactivating electrodes or lowering stimulation current levels. Avoiding axon fascicule stimulation aims to remove brushstroke-like shapes in favor of a more reduced set of pixel-like shapes.

APPROACH

In this study, we propose the use of isotropic and anisotropic shapes to render intelligible images on the retina of a virtual patient in a reinforcement learning environment named rlretina. The environment formalizes the task as using brushstrokes in a stroke-based rendering task.

MAIN RESULTS

We train a deep reinforcement learning agent that learns to assemble isotropic and anisotropic shapes to form an image. We investigate which error-based or perception-based metrics are adequate to reward the agent. The agent is trained in a model-based data generation fashion using the psychophysically validated axon map model to render images as perceived by different virtual patients. We show that the agent can generate more intelligible images compared to the naive method in different virtual patients.

SIGNIFICANCE

This work shares a new way to address epiretinal stimulation that constitutes a first step towards improving visual acuity in artificially-restored vision using anisotropic phosphenes.

Axonal stimulation affects the linear summation of single-point perception in three Argus II users

Purpose

Retinal implants use electrical stimulation to elicit perceived flashes of light ("phosphenes"). Single-electrode phosphene shape has been shown to vary systematically with stimulus parameters and the retinal location of the stimulating electrode, due to incidental activation of passing nerve fiber bundles. However, this knowledge has yet to be extended to paired-electrode stimulation.

Methods

We retrospectively analyzed 3548 phosphene drawings made by three blind participants implanted with an Argus II Retinal Prosthesis. Phosphene shape (characterized by area, perimeter, major and minor axis length) and number of perceived phosphenes were averaged across trials and correlated with the corresponding single-electrode parameters. In addition, the number of phosphenes was correlated with stimulus amplitude and neuroanatomical parameters: electrode-retina and electrode-fovea distance as well as the electrode-electrode distance to ("between-axon") and along axon bundles ("along-axon"). Statistical analyses were conducted using linear regression and partial correlation analysis.

Results

Simple regression revealed that each paired-electrode shape descriptor could be predicted by the sum of the two corresponding single-electrode shape descriptors (p < .001). Multiple regression revealed that paired-electrode phosphene shape was primarily predicted by stimulus amplitude and electrode-fovea distance (p < .05). Interestingly, the number of elicited phosphenes tended to increase with between-axon distance (p < .05), but not with along-axon distance, in two out of three participants.

Conclusions

The shape of phosphenes elicited by paired-electrode stimulation was well predicted by the shape of their corresponding single-electrode phosphenes, suggesting that two-point perception can be expressed as the linear summation of single-point perception. The notable impact of the between-axon distance on the perceived number of phosphenes provides further evidence in support of the axon map model for epiretinal stimulation. These findings contribute to the growing literature on phosphene perception and have important implications for the design of future retinal prostheses.

Patient-specific computational models of retinal prostheses

Retinal prostheses stimulate inner retinal neurons to create visual perception for blind patients. Implanted arrays have many small electrodes. Not all electrodes induce perception at the same stimulus amplitude, requiring clinicians to manually establish a visual perception threshold for each one. Phosphenes created by single-electrode stimuli can also vary in shape, size, and brightness. Computational models provide a tool to predict inter-electrode variability and automate device programming. In this study, we created statistical and patient-specific field-cable models to investigate inter-electrode variability across seven epiretinal prosthesis users. Our statistical analysis revealed that retinal thickness beneath the electrode correlated with perceptual threshold, with a significant fixed effect across participants. Electrode-retina distance and electrode impedance also correlated with perceptual threshold for some participants, but these effects varied by individual. We developed a novel method to construct patient-specific field-cable models from optical coherence tomography images. Predictions with these models significantly correlated with perceptual threshold for 80% of participants. Additionally, we demonstrated that patient-specific field-cable models could predict retinal activity and phosphene size. These computational models could be beneficial for determining optimal stimulation settings in silico, circumventing the trial-and-error testing of a large parameter space in clinic.

A Model of Progression to Help Identify Macular Damage Due to Glaucoma

The central macula contains a thick donut shaped region of the ganglion cell layer (GCL) that surrounds the fovea. This region, which is about 12 degrees (3.5 mm) in diameter, is essential for everyday functions such as driving, reading, and face recognition. Here, we describe a model of progression of glaucomatous damage to this GCL donut. This model is based upon assumptions supported by the literature, and it predicts the patterns of glaucomatous damage to the GCL donut, as seen with optical coherence tomography (OCT). After describing the assumptions and predictions of this model, we test the model against data from our laboratory, as well as from the literature. Finally, three uses of the model are illustrated. One, it provides an aid to help clinicians focus on the essential central macula and to alert them to look for other, non-glaucomatous causes, when the GCL damage does not fit the pattern predicted by the model. Second, the patterns of progression predicted by the model suggest alternative end points for clinical trials. Finally, the model provides a heuristic for future research concerning the anatomic basis of glaucomatous damage.

Human-in-the-Loop Optimization for Deep Stimulus Encoding in Visual Prostheses

Neuroprostheses show potential in restoring lost sensory function and enhancing human capabilities, but the sensations produced by current devices often seem unnatural or distorted. Exact placement of implants and differences in individual perception lead to significant variations in stimulus response, making personalized stimulus optimization a key challenge. Bayesian optimization could be used to optimize patient-specific stimulation parameters with limited noisy observations, but is not feasible for high-dimensional stimuli. Alternatively, deep learning models can optimize stimulus encoding strategies, but typically assume perfect knowledge of patient-specific variations. Here we propose a novel, practically feasible approach that overcomes both of these fundamental limitations. First, a deep encoder network is trained to produce optimal stimuli for any individual patient by inverting a forward model mapping electrical stimuli to visual percepts. Second, a preferential Bayesian optimization strategy utilizes this encoder to optimize patient-specific parameters for a new patient, using a minimal number of pairwise comparisons between candidate stimuli. We demonstrate the viability of this approach on a novel, state-of-the-art visual prosthesis model. We show that our approach quickly learns a personalized stimulus encoder, leads to dramatic improvements in the quality of restored vision, and is robust to noisy patient feedback and misspecifications in the underlying forward model. Overall, our results suggest that combining the strengths of deep learning and Bayesian optimization could significantly improve the perceptual experience of patients fitted with visual prostheses and may prove a viable solution for a range of neuroprosthetic technologies.

Patient-specific computational models of retinal prostheses

Retinal prostheses stimulate inner retinal neurons to create visual perception for blind patients. Implanted arrays have many small electrodes, which act as pixels. Not all electrodes induce perception at the same stimulus amplitude, requiring clinicians to manually establish a visual perception threshold for each one. Phosphenes created by single-electrode stimuli can also vary in shape, size, and brightness. Computational models provide a tool to predict inter-electrode variability and automate device programming. In this study, we created statistical and patient-specific field-cable models to investigate inter-electrode variability across seven epiretinal prosthesis users. Our statistical analysis revealed that retinal thickness beneath the electrode correlated with perceptual threshold, with a significant fixed effect across participants. Electrode-retina distance and electrode impedance also correlated with perceptual threshold for some participants, but these effects varied by individual. We developed a novel method to construct patient-specific field-cable models from optical coherence tomography images. Predictions with these models significantly correlated with perceptual threshold for 80% of participants. Additionally, we demonstrated that patient-specific field-cable models could predict retinal activity and phosphene size. These computational models could be beneficial for determining optimal stimulation settings in silico, circumventing the trial-and-error testing of a large parameter space in clinic.

Detectability of Visual Field Defects in Glaucoma Using Moving Versus Static Stimuli for Perimetry

Purpose

We have previously shown that using moving, instead of static, stimuli extends the effective dynamic range of automated perimetry in glaucoma. In this study, we further investigate the effect of using moving stimuli on the detectability of functional loss.

Methods

We used two experimental perimetry paradigms to test 155 subjects with a diagnosis of glaucoma or glaucoma suspect, and 34 healthy control subjects. One test used stimuli moving parallel to the average nerve fiber bundle orientation at each location; the other used static stimuli. Algorithms were otherwise identical. Sensitivities to moving stimuli were transformed to the equivalent values for static stimuli based on a Bland-Altman plot. The proportions of locations outside age-corrected normative limits were compared, and test-retest variability was compared against defect depth for each stimulus type.

Results

More tested locations were below the fifth percentile of the normative range for that location using static stimuli. However, among locations abnormal according to standard clinical perimetry on the same day, 19.2% were abnormal using static stimuli, versus 20.5% using moving stimuli (P = 0.372). Test-retest variability was 44% lower for moving stimuli across the range of defect depths.

Conclusions

When compared with static automated perimetry and expressed on a common scale, moving stimuli extend the effective dynamic range and decrease variability, without decreasing the detectability of known functional defects.

Translational Relevance

Moving stimuli provide a method to improve known problems of current clinical perimetry.

Correlation between retinal structure and brain multimodal magnetic resonance imaging in patients with Alzheimer's disease

Background

The retina imaging and brain magnetic resonance imaging (MRI) can both reflect early changes in Alzheimer's disease (AD) and may serve as potential biomarker for early diagnosis, but their correlation and the internal mechanism of retinal structural changes remain unclear. This study aimed to explore the possible correlation between retinal structure and visual pathway, brain structure, intrinsic activity changes in AD patients, as well as to build a classification model to identify AD patients.

Methods

In the study, 49 AD patients and 48 healthy controls (HCs) were enrolled. Retinal images were obtained by optical coherence tomography (OCT). Multimodal MRI sequences of all subjects were collected. Spearman correlation analysis and multiple linear regression models were used to assess the correlation between OCT parameters and multimodal MRI findings. The diagnostic value of combination of retinal imaging and brain multimodal MRI was assessed by performing a receiver operating characteristic (ROC) curve.

Results

Compared with HCs, retinal thickness and multimodal MRI findings of AD patients were significantly altered (p < 0.05). Significant correlations were presented between the fractional anisotropy (FA) value of optic tract and mean retinal thickness, macular volume, macular ganglion cell layer (GCL) thickness, inner plexiform layer (IPL) thickness in AD patients (p < 0.01). The fractional amplitude of low frequency fluctuations (fALFF) value of primary visual cortex (V1) was correlated with temporal quadrant peripapillary retinal nerve fiber layer (pRNFL) thickness (p < 0.05). The model combining thickness of GCL and temporal quadrant pRNFL, volume of hippocampus and lateral geniculate nucleus, and age showed the best performance to identify AD patients [area under the curve (AUC) = 0.936, sensitivity = 89.1%, specificity = 87.0%].

Conclusion

Our study demonstrated that retinal structure change was related to the loss of integrity of white matter fiber tracts in the visual pathway and the decreased LGN volume and functional metabolism of V1 in AD patients. Trans-synaptic axonal retrograde lesions may be the underlying mechanism. Combining retinal imaging and multimodal MRI may provide new insight into the mechanism of retinal structural changes in AD and may serve as new target for early auxiliary diagnosis of AD.

On the protective role of the blood vessels in glaucomatous damage: A transversal study

PURPOSE

To corroborate whether vessels on the surface of the optic nerve head can provide protection against the loss of underlying axons in subjects with manifest glaucoma.

METHODS

In this pilot study, thirty-six glaucomatous eyes with a perimetric defect in the Bjerrum area were included. The retinal nerve fiber layer (RNFL) thickness was measured in each of the sectors of the clock-hour map obtained by Cirrus HD-OCT considering the presence or absence of blood vessels. These sectors were related with their corresponding areas of the retina examined in the visual field using a mathematical model of the retina introduced by Jansonius, in order to determine the values of threshold sensitivity in those areas in the presence or absence of vessels.

RESULTS

We corroborated the protective role of the blood vessel for peripapillary RNFL thickness of clock-hour 12 despite obtaining a p-value (p = 0.023; w = 228.5) close to the acceptance zone (p ≥ 0.05). The mean ± standard deviation with vessel and without vessel were 70.95 ± 24.35 and 88.46 ± 23.96, respectively. No differences were found between the mean values of threshold sensitivity to the presence or absence of blood vessels in each of the sectors considered.

CONCLUSIONS

Our findings do not allow us to affirm that there is an association between the presence of a vessel and protection against glaucomatous damage in subjects with an advanced manifestation of the disease. In the future, more extensive studies are needed to study this relationship in subjects with early glaucoma.

Hybrid Neural Autoencoders for Stimulus Encoding in Visual and Other Sensory Neuroprostheses

Sensory neuroprostheses are emerging as a promising technology to restore lost sensory function or augment human capabilities. However, sensations elicited by current devices often appear artificial and distorted. Although current models can predict the neural or perceptual response to an electrical stimulus, an optimal stimulation strategy solves the inverse problem: what is the required stimulus to produce a desired response? Here, we frame this as an end-to-end optimization problem, where a deep neural network stimulus encoder is trained to invert a known and fixed forward model that approximates the underlying biological system. As a proof of concept, we demonstrate the effectiveness of this hybrid neural autoencoder (HNA) in visual neuroprostheses. We find that HNA produces high-fidelity patient-specific stimuli representing handwritten digits and segmented images of everyday objects, and significantly outperforms conventional encoding strategies across all simulated patients. Overall this is an important step towards the long-standing challenge of restoring high-quality vision to people living with incurable blindness and may prove a promising solution for a variety of neuroprosthetic technologies.

Moving Stimulus Perimetry: A New Functional Test for Glaucoma

Purpose

Static pointwise perimetric sensitivities of less than approximately 19 dB are unreliable in glaucoma owing to excessive variability. We propose using moving stimuli to increase detectability, decrease variability, and hence increase this dynamic range.

Methods

A moving stimulus was designed to travel parallel to the average nerve fiber bundle orientation at each location, and compared against an otherwise identical static stimulus. To assess dynamic range, psychometric functions were measured at 4 locations of each of 10 subjects. To assess clinically realistic test-retest variability, 34 locations of 94 subjects with glaucoma and glaucoma suspects were tested twice, 6 months apart. Pointwise sensitivity estimates were compared using generalized estimating equation regression models. The test-retest limits of agreement for each stimulus were assessed, adjusted for within-eye clustering.

Results

Using static stimuli, 9 of the 40 psychometric functions had less than a 90% maximum response probability, suggesting being beyond the dynamic range. Eight of those locations had asymptotic maximum of more than 90% with moving stimuli. Sensitivities were higher for moving stimuli (P < 0.001); the difference increased as sensitivity decreased (P < 0.001). Test-retest limits of agreement were narrower for moving stimuli (-6.35 to +6.48 dB) than static stimuli (-12.7 to +7.81 dB). Sixty-two percent of subjects preferred using moving stimuli versus 19% who preferred static stimuli.

Conclusions

Using a moving stimulus increases perimetric sensitivities in regions of glaucomatous loss. This extends the effective dynamic range, allowing reliable testing later into the disease. Results are more repeatable, and the test is preferred by most subjects.

Translational Relevance

Moving stimuli allow reliable testing in patients with more severe glaucoma than currently possible.

Factors affecting two-point discrimination in Argus II patients

Two of the main obstacles to the development of epiretinal prosthesis technology are electrodes that require current amplitudes above safety limits to reliably elicit percepts, and a failure to consistently elicit pattern vision. Here, we explored the causes of high current amplitude thresholds and poor spatial resolution within the Argus II epiretinal implant. We measured current amplitude thresholds and two-point discrimination (the ability to determine whether one or two electrodes had been stimulated) in 3 blind participants implanted with Argus II devices. Our data and simulations show that axonal stimulation, lift and retinal damage all play a role in reducing performance in the Argus 2, by either limiting sensitivity and/or reducing spatial resolution. Understanding the relative role of these various factors will be critical for developing and surgically implanting devices that can successfully subserve pattern vision.

Test of a Retinal Nerve Fiber Bundle Trajectory Model Using Eyes With Glaucomatous Optic Neuropathy

Purpose

To test a model of retinal nerve fiber bundle trajectories that predicts the arcuate-shaped patterns seen on optical coherence tomography (OCT) retinal nerve fiber layer (RNFL) probability/deviation maps (p-maps) in glaucomatous eyes.

Methods

Thirty-one glaucomatous eyes from a database of 250 eyes had clear arcuate-shaped patterns on RNFL p-maps derived from an OCT cube scan. The borders of the arcuate patterns were extracted from the RNFL p-maps. Next, the trajectories from an arcuate model were compared against these borders via a normalized root-mean-square difference analysis. The model's parameter, β, was varied, and the best-fitting, initial clock-hour position of the trajectory to the border was found for each β. Finally, the regions, as determined by the arcuate border's best-fit, initial clock-hour positions, were compared against the abnormal regions on the circumpapillary retinal nerve fiber layer (cpRNFL) profile.

Results

The arcuate model's mean β_Sup and β_Inf parameters minimized large differences between the trajectories and the arcuate borders on the RNFL p-maps. Furthermore, on average, 68% of the cpRNFL regions defined by the arcuate border's best-fit, initial clock-hour positions were abnormal (i.e., below the ≤5% threshold).

Conclusions

The arcuate model performed well in predicting the borders of arcuate patterns seen on RNFL p-maps. It also predicted the associated abnormal regions of the cpRNFL thickness plots.

Translational Relevance

This model should prove useful in helping clinicians understand topographical comparisons among different OCT representations and should improve structure-structure, as well as structure-function agreement analyses.

Elucidating macular structure-function correlations in glaucoma

Correlation between structural data from optical coherence tomography and functional data from the visual field may be suboptimal because of poor mapping of OCT measurement locations to VF stimuli. We tested the hypothesis that stronger structure–function correlations in the macula can be achieved with fundus-tracking perimetery, by precisely mapping OCT measurements to VF sensitivity at the same location. The conventional 64 superpixel (3° × 3°) OCT grid was mapped to VF sensitivities averaged in 40 corresponding VF units with standard automated perimetry (conventional mapped approach, CMA) in 38 glaucoma patients and 10 healthy subjects. Similarly, a 144 superpixel (2° × 2°) OCT grid was mapped to each of the 68 locations with fundus-tracking perimetry (localized mapped approach, LMA). For each approach, the correlation between sensitivity at each VF unit and OCT superpixel was computed. Vector maps showing the maximum correlation between each VF unit and OCT pixel was generated. CMA yielded significantly higher structure–function correlations compared to LMA. Only 20% of the vectors with CMA and < 5% with LMA were within corresponding mapped OCT superpixels, while most were directed towards loci with structural damage. Measurement variability and patterns of structural damage more likely impact correlations compared to precise mapping of VF stimuli.

Human-in-the-loop optimization of visual prosthetic stimulation

Retinal prostheses are a promising strategy to restore sight to patients with retinal degenerative diseases. These devices compensate for the loss of photoreceptors by electrically stimulating neurons in the retina. Currently, the visual function that can be recovered with such devices is very limited. This is due, in part, to current spread, unintended axonal activation, and the limited resolution of existing devices. Here we show, using a recent model of prosthetic vision, that optimizing how visual stimuli are encoded by the device can help overcome some of these limitations, leading to dramatic improvements in visual perception.

APPROACH

We propose a strategy to do this in practice, using patients' feedback in a visual task. The main challenge of our approach comes from the fact that, typically, one only has access to a limited number of noisy responses from patients. We propose two ways to deal with this: first, we use a model of prosthetic vision to constrain and simplify the optimization. We show that, if one knew the parameters of this model for a given patient, it would be possible to greatly improve their perceptual performance. Second we propose a preferential Bayesian optimization to efficiently learn these model parameters for each patient, using minimal trials.

MAIN RESULTS

To test our approach, we presented healthy subjects with visual stimuli generated by a recent model of prosthetic vision, to replicate the perceptual experience of patients fitted with an implant. Our optimization procedure led to significant and robust improvements in perceived image quality, that transferred to increased performance in other tasks.

SIGNIFICANCE

Importantly, our strategy is agnostic to the type of prosthesis and thus could readily be implemented in existing implants.

Improving the spatial resolution of artificial vision using midget retinal ganglion cell populations modelled at the human fovea

OBJECTIVE

Retinal prostheses seek to create artificial vision by stimulating surviving retinal neurons of patients with profound vision impairment. Notwithstanding tremendous research efforts, the performance of all implants tested to date has remained rudimentary, incapable of overcoming the threshold for legal blindness. To maximize the perceptual efficacy of retinal prostheses, a device must be capable of controlling retinal neurons with greater spatiotemporal precision. Most studies of retinal stimulation were derived from either non-primate species or the peripheral primate retina. We investigated if artificial stimulation could leverage the high spatial resolution afforded by the neural substrates at the primate fovea and surrounding regions to achieve improved percept qualities.

APPROACH

We began by developing a new computational model capable of generating anatomically accurate retinal ganglion cell (RGC) populations within the human central retina. Next, multiple RGC populations across the central retina were stimulated in-silico to compare clinical and recently proposed neurostimulation configurations based on their ability to improve perceptual efficacy and reduce activation thresholds.

MAIN RESULTS

Our model uniquely upholds eccentricity-dependent characteristics such as RGC density and dendritic field diameter, whilst incorporating anatomically accurate features such as axon projection and three-dimensional RGC layering, features often forgone in favor of reduced computational complexity. Following epiretinal stimulation, the RGCs in our model produced response patterns in shapes akin to the complex percepts reported in clinical trials. Our results also demonstrated that even within the neuron-dense central retina, epiretinal stimulation using a multi-return hexapolar electrode arrangement could reliably achieve spatially focused RGC activation and could achieve single-cell excitation in 74% of all tested locations.

SIGNIFICANCE

This study establishes an anatomically accurate three-dimensional model of the human central retina and demonstrates the potential for an epiretinal hexapolar configuration to achieve consistent, spatially confined retinal responses, even within the neuron-dense foveal region. Our results promote the prospect and optimization of higher spatial resolution in future epiretinal implants.

Cortical Atrophy Predicts Visual Performance in Long-Term Central Retinal Disease; GCL, pRNFL and Cortical Thickness Are Key Biomarkers

Purpose

The aim of this study was to assess both retinal and cortical structure in a cohort of patients with long-term acquired central retinal disease in order to identify potential disease biomarkers and to explore the relationship between the anterior and posterior visual pathways.

Methods

Fourteen participants diagnosed with long-term central retinal disease underwent structural assessments of the retina using spectral-domain optical coherence tomography, including macular ganglion cell layer (GCL) and peripapillary retinal nerve fiber layer (pRNFL) thickness. Structural magnetic resonance imaging was used to measure visual cortex, including cortical volume of the entire occipital lobe and cortical thickness of the occipital pole and calcarine sulcus, representing the central and peripheral retina, respectively.

Results

Mean thickness was significantly reduced in both the macular GCL and the inferior temporal pRNFL across patients. Cortical thickness was significantly reduced in both the occipital pole and calcarine sulcus, representing the central and peripheral retina, respectively. Disease duration significantly correlated with GCL thickness with a large effect size, whereas a medium effect size suggests the possibility that cortical thickness in the occipital pole may correlate with visual acuity.

Conclusions

Long-term central retinal disease is associated with significant structural changes to both the retina and the brain. Exploratory analysis suggests that monitoring GCL thickness may be a sensitive biomarker of disease progression and reductions in visual cortical thickness may be associated with reduced visual acuity. Although this study is limited by its heterogeneous population, larger cohort studies would be needed to better establish some of the relationships detected between disease dependent structural properties of the anterior and posterior visual pathway given the effect sizes reported in our exploratory analysis.

Detecting glaucoma with only OCT: Implications for the clinic, research, screening, and AI development

A method for detecting glaucoma based only on optical coherence tomography (OCT) is of potential value for routine clinical decisions, for inclusion criteria for research studies and trials, for large-scale clinical screening, as well as for the development of artificial intelligence (AI) decision models. Recent work suggests that the OCT probability (p-) maps, also known as deviation maps, can play a key role in an OCT-based method. However, artifacts seen on the p-maps of healthy control eyes can resemble patterns of damage due to glaucoma. We document in section 2 that these glaucoma-like artifacts are relatively common and are probably due to normal anatomical variations in healthy eyes. We also introduce a simple anatomical artifact model based upon known anatomical variations to help distinguish these artifacts from actual glaucomatous damage. In section 3, we apply this model to an OCT-based method for detecting glaucoma that starts with an examination of the retinal nerve fiber layer (RNFL) p-map. While this method requires a judgment by the clinician, sections 4 and 5 describe automated methods that do not. In section 4, the simple model helps explain the relatively poor performance of commonly employed summary statistics, including circumpapillary RNFL thickness. In section 5, the model helps account for the success of an AI deep learning model, which in turn validates our focus on the RNFL p-map. Finally, in section 6 we consider the implications of OCT-based methods for the clinic, research, screening, and the development of AI models.

Policy-Driven, Multimodal Deep Learning for Predicting Visual Fields from the Optic Disc and Optical Coherence Tomography Imaging

PURPOSE

To develop and validate a deep learning (DL) system for predicting each point on visual fields (VF) from disc and optical coherence tomography (OCT) imaging and derive a structure-function mapping.

DESIGN

Retrospective, cross-sectional database study PARTICIPANTS: 6437 patients undergoing routine care for glaucoma in three clinical sites in the UK.

METHODS

OCT and infrared reflectance (IR) optic disc imaging was paired with the closest VF within 7 days. Efficient-Net B2 was used to train two single modality DL models to predict each of the 52 sensitivity points on the 24-2 VF pattern. A policy DL model was designed and trained to fuse the two model predictions.

MAIN OUTCOME MEASURES

Pointwise Mean Absolute Error (PMAE) RESULTS: A total of 5078 imaging to VF pairs were used as a held-out test set to measure the final performance. The improvement in PMAE with the policy model was 0.485 [0.438, 0.533] dB compared to the IR image of the disc alone and 0.060 [0.047, 0.073] dB compared to the OCT alone. The improvement with the policy fusion model was statistically significant (p < 0.0001). Occlusion masking shows that the DL models learned the correct structure function mapping in a data-driven, feature agnostic fashion.

CONCLUSIONS

The multimodal, policy DL model performed the best; it provided explainable maps of its confidence in fusing data from single modalities and provides a pathway for probing the structure-function relationship in glaucoma.

Identifying the Retinal Layers Linked to Human Contrast Sensitivity Via Deep Learning

Purpose

Luminance contrast is the fundamental building block of human spatial vision. Therefore contrast sensitivity, the reciprocal of contrast threshold required for target detection, has been a barometer of human visual function. Although retinal ganglion cells (RGCs) are known to be involved in contrast coding, it still remains unknown whether the retinal layers containing RGCs are linked to a person's contrast sensitivity (e.g., Pelli-Robson contrast sensitivity) and, if so, to what extent the retinal layers are related to behavioral contrast sensitivity. Thus the current study aims to identify the retinal layers and features critical for predicting a person's contrast sensitivity via deep learning.

Methods

Data were collected from 225 subjects including individuals with either glaucoma, age-related macular degeneration, or normal vision. A deep convolutional neural network trained to predict a person's Pelli-Robson contrast sensitivity from structural retinal images measured with optical coherence tomography was used. Then, activation maps that represent the critical features learned by the network for the output prediction were computed.

Results

The thickness of both ganglion cell and inner plexiform layers, reflecting RGC counts, were found to be significantly correlated with contrast sensitivity (r = 0.26 ∼ 0.58, Ps < 0.001 for different eccentricities). Importantly, the results showed that retinal layers containing RGCs were the critical features the network uses to predict a person's contrast sensitivity (an average R2 = 0.36 ± 0.10).

Conclusions

The findings confirmed the structure and function relationship for contrast sensitivity while highlighting the role of RGC density for human contrast sensitivity.

Combining vascular and nerve fiber layer thickness measurements to model glaucomatous focal visual field loss

We compare the focal structure-function correlation of structural measurements of peripapillary retinal nerve fiber layer thickness (RNFL-T) using optical coherence tomography (OCT), capillary density (CD) measurements using OCT-angiography (OCT-A), or a combination of both, with visual field deviation (VFD) in early to advanced glaucoma. Primary open angle glaucoma patients (n = 46, mean ± SD age: 67 ± 10 years; VF mean deviation: -10.41 ± 6.76 dB) were included in this cross-sectional study. We performed 30-2 standard automated perimetry OCT (3.5-mm diameter ring scan) and 15°×15° OCT-A (superficial vascular complex slab). Based on a nerve fiber trajectory model, each VF test spot was assigned to an OCT-A wedge and an OCT ring-sector. Two univariate linear models (M_v and M_t ) using either CD-based vascular (M_v ) or RNFL-T-based thickness information (M_t ) and one multivariate model using both (M_v:t ) were compared in their associations with measured focal VFD, which were higher for the multivariate model M_v:t (mean ± SD correlation coefficient: 0.710 ± 0.086) than for either nested model (0.627 ± 0.078 for M_v and 0.578 ± 0.095 for M_t ). Using a focal visual field approach, the combination of RNFL-T and CD showed better structure-function correlations than thickness or vascular information only.

Automatic retinal nerve fiber bundle tracing based on large field of view polarization sensitive OCT data

A technique to accurately estimate trajectories of retinal nerve fiber bundles (RNFB) in a large field of view (FOV) image covering 45° is described. The method utilizes stitched projections of polarization-sensitive optical coherence tomography (PS-OCT) data, as well as a mathematical model of average RNFB trajectories as prior. The fully automatic process was applied to data recorded in healthy subjects and glaucoma patients and automatically detected individual RNFB trajectories are compared to manual traces.

Combining OCT and OCTA for Focal Structure-Function Modeling in Early Primary Open-Angle Glaucoma

Purpose

To investigate modeling of the focal visual field (VF) loss by combining structural measurements and vascular measurements in eyes with early primary open-angle glaucoma (POAG).

Methods

In this cross-sectional study, subjects with early glaucoma (VF mean deviation, ≥−6 dB) underwent optical coherence tomography (OCT) imaging, optical coherence tomography angiography (OCTA) imaging, and Humphrey 24-2 VF tests. Capillary perfusion densities (CPDs) were calculated after the removal of large vessels in the OCTA images. Focal associations between VF losses at the individual VF test locations, circumpapillary retinal nerve fiber layer (RNFL) thickness measurements from OCT, and CPDs were determined using nerve fiber trajectory tracings. Linear mixed models were used to model focal VF losses at each VF test location.

Results

Ninety-seven eyes with early POAG (VF mean deviation, −2.47 ± 1.64 dB) of 71 subjects were included. Focal VF modeling using a combined RNFL–CPD approach resulted in a median adjusted R² value of 0.30 (interquartile range [IQR], 0.13–0.55), whereas the RNFL-only and CPD-only approaches resulted in median values of 0.22 (IQR, 0.10–0.51) and 0.26 (IQR, 0.10–0.52), respectively. Seventeen VF locations with the combined approach had an adjusted R² value greater than 0.50. Likelihood testing at each VF test location showed that the combined approach performed significantly better at the superior nasal VF regions of the eyes compared with the univariate approaches.

Conclusions

Modeling of focal VF losses showed improvements when structural thickness and vascular parameters were included in tandem. Evaluation of VF defects in early glaucoma may benefit from considering both RNFL and OCTA characteristics.

A Computational Model of Phosphene Appearance for Epiretinal Prostheses

Retinal neuroprostheses are the only FDA-approved treatment option for blinding degenerative diseases. A major outstanding challenge is to develop a computational model that can accurately predict the elicited visual percepts (phosphenes) across a wide range of electrical stimuli. Here we present a phenomenological model that predicts phosphene appearance as a function of stimulus amplitude, frequency, and pulse duration. The model uses a simulated map of nerve fiber bundles in the retina to produce phosphenes with accurate brightness, size, orientation, and elongation. We validate the model on psychophysical data from two independent studies, showing that it generalizes well to new data, even with different stimuli and on different electrodes. Whereas previous models focused on either spatial or temporal aspects of the elicited phosphenes in isolation, we describe a more comprehensive approach that is able to account for many reported visual effects. The model is designed to be flexible and extensible, and can be fit to data from a specific user. Overall this work is an important first step towards predicting visual outcomes in retinal prosthesis users across a wide range of stimuli.

Hierarchical Censored Bayesian Analysis of Visual Field Progression

Purpose

To develop a Bayesian model (BM) for visual field (VF) progression accounting for the hierarchical, censored and heteroskedastic nature of the data.

Methods

Three versions of a hierarchical BM were developed: a simple linear (Hi-linear); censored at 0 dB (Hi-censored); heteroskedastic censored (Hi-HSK). For the latter, we modeled the test variability according to VF sensitivity using a large test-retest cohort (1396 VFs, 146 eyes with glaucoma). We analyzed a large cohort of 44,371 VF tests from 3352 eyes from five glaucoma clinics. We quantified the bias in the estimated rate-of-progression, the detection of progression (Hit-rate [HR]), the median time-to-progression and the prediction error of future observations (mean absolute error [MAE]). HR and time-to-progression were compared at matched false-positive-rate (FPR), quantified using permutations of a separate test-retest cohort (360 tests, 30 eyes with glaucoma). BMs were compared to simple linear regression and Permutation-Analyses-of Pointwise-Linear-Regression. Differences in time-to-progression were tested using survival analysis.

Results

Censored models showed the smallest bias in the rate-of-progression. The three BMs performed very similarly in terms of HR and time-to-progression and always better than the other methods. The average reduction in time-to-progression was 37% with the BMs (P < 0.001) at 5% FPR. MAE for prediction was very similar among methods.

Conclusions

Bayesian hierarchical models improved the detection of VF progression. Accounting for censoring improves the precision of the estimates, but minimal effect is provided by accounting for heteroskedasticity.

Translational Relevance

These results are relevant for quantification of VF progression in practice and for clinical trials.

Comparison of the 24-2 and 24-2C Visual Field Grids in Determining the Macular Structure-Function Relationship in Glaucoma

PRECIS

The 24-2C grid showed significantly greater structure-function associations in the global and regional sectors than the 24-2 VF grid in open-angle glaucoma eyes.

PURPOSE

To compare a Humphrey Field Analyzer (HFA) Swedish Interactive Threshold Algorithm-Faster (SITA-faster) 24-2C grid to a conventional HFA 24-2 grid regarding macular structure-function (S-F) relationships.

MATERIALS AND METHODS

The macular ganglion cell-inner plexiform layer thickness (mGCIPLT) was measured at different parafoveal sectors using spectral-domain optical coherence tomography in 150 eyes from 150 healthy, preperimetric and perimetric glaucoma subjects. The central visual field mean sensitivity (VFMS) on the decibel and 1/L scales and the parafoveal mGCIPLT were matched topographically in four sectors and the strengths of the S-F relationships were assessed using 'weighted' correlation coefficients and compared between 24-2C and 24-2 VF test grids.

RESULTS

There were significant global and sectoral correlations between the mGCIPLT and VFMS using both VF grids. The S-F correlations between the average/hemimacular mGCIPLT and the corresponding VFMS using a 24-2C grid were however significantly greater in both the entire and PG groups (P<0.05), except for the average mGCIPLT of the PG group in the 1/L scale (P=0.065). The 24-2C grid showed significantly greater S-F associations in the superotemporal and inferotemporal parafoveal sectors than the 24-2 VF grid (both P<0.05).

CONCLUSIONS

A 24-2C grid may offer an advantage over the conventional 24-2 VF grid in assessing macular S-F relationships.

Identification and quantification of fibrotic areas in the human retina using polarization-sensitive OCT

Subretinal fibrosis is one of the most prevalent causes of blindness in the elderly population, but a true gold standard to objectively diagnose fibrosis is still lacking. Since fibrotic tissue is birefringent, it can be detected by polarization-sensitive optical coherence tomography (PS-OCT). We present a new algorithm to automatically detect, segment, and quantify fibrotic lesions within 3D data sets recorded by PS-OCT. The algorithm first compensates for the birefringence of anterior ocular tissues and then uses the uniformity of the birefringent optic axis as an indicator to identify fibrotic tissue, which is then segmented and quantified. The algorithm was applied to 3D volumes recorded in 57 eyes of 57 patients with neovascular age-related macular degeneration using a spectral domain PS-OCT system. The results of fibrosis detection were compared to the clinical diagnosis based on color fundus photography (CFP), and the precision of fibrotic area measurement was assessed by three repeated measurements in a sub-set of 15 eyes. The average standard deviation of the fibrotic area obtained in eyes with a lesion area > 0.7 mm2 was 15%. Fibrosis detection by CFP and PS-OCT agreed in 48 cases, discrepancies were only observed in cases of lesion area < 0.7 mm2. These remaining discrepancies are discussed, and a new method to treat ambiguous cases is presented.

Developing a normative database for retinal perfusion using optical coherence tomography angiography

Visualizing and characterizing microvascular abnormalities with optical coherence tomography angiography (OCTA) has deepened our understanding of ocular diseases, such as glaucoma, diabetic retinopathy, and age-related macular degeneration. Two types of microvascular defects can be detected by OCTA: focal decrease because of localized absence and collapse of retinal capillaries, which is referred to as the non-perfusion area in OCTA, and diffuse perfusion decrease usually detected by comparing with healthy case-control groups. Wider OCTA allows for insights into peripheral retinal vascularity, but the heterogeneous perfusion distribution from the macula, parapapillary area to periphery hurdles the quantitative assessment. A normative database for OCTA could estimate how much individual's data deviate from the normal range, and where the deviations locate. Here, we acquired OCTA images using a swept-source OCT system and a 12×12 mm protocol in healthy subjects. We automatically segmented the large blood vessels with U-Net, corrected for anatomical factors such as the relative position of fovea and disc, and segmented the capillaries by a moving window scheme. A total of 195 eyes were included and divided into 4 age groups: < 30 (n=24) years old, 30-49 (n=28) years old, 50-69 (n=109) years old and >69 (n=34) years old. This provides an age-dependent normative database for characterizing retinal perfusion abnormalities in 12×12 mm OCTA images. The usefulness of the normative database was tested on two pathological groups: one with diabetic retinopathy; the other with glaucoma.

Compartmental Differences in Macular Retinal Ganglion Cell Function

Purpose

The purpose of this study was to investigate local differences of macular retinal ganglion cell (RGC) function by means of the steady-state pattern electroretinogram (SS-PERG).

Methods

SS-PERGs were recorded in healthy subjects (n = 43) in response to gratings (1.6 c/deg, 15.63 reversals/s, and 98% contrast) presented on an LED display (800 cd/m², 12.5 degrees eccentricity at 30 cm viewing distance) partitioned in triangular sectors (inferior [I]; nasal [N]; superior [S]; and temporal [T]) or concentric regions (central [C] and annulus [A]). For each partition, response amplitude (nV), amplitude adaptation (% change over recording time), phase/latency (deg/ms), and oscillatory potentials (OPs) amplitude (root mean square [RMS] nV) were measured. Data were analyzed with Generalized Estimating Equation (GEE) statistics.

Results

Amplitude differed (P < 0.001) between sectors (I: 254 nV; N: 328 nV; S: 275 nV; T: 264 nV; and N>T, I) as well as concentrically (C: 684 nV; A: 323 nV; and C>A). Latency did not differ between sectors (range = 53–54 ms, P = 0.45) or concentrically (range = 51–51 ms, P = 0.7). Adaptation did not differ (P = 0.66) concentrically (C: −19% and A: −22%) but differed (P = 0.004) between sectors (I: +25% and S: −29%). The OP amplitude did not differ (P = 0.5) between sectors (range = 63–73 nV) as well as concentrically (range = 82–90 nV, P = 0.3).

Conclusions

Amplitude profiles paralleled RGC densities from histological studies. Adaptation profile suggested greater autoregulatory challenge in the inferior retina. Latency profile may reflect axonal conduction time to the optic nerve head assuming a direct relationship between axon length and its size/velocity. Location-independent OPs may reflect preganglionic activity.

Translational Relevance

Normal macular RGC function displays local differences that may be related to local vulnerability in optic nerve disorders.

Improving Personalized Structure to Function Mapping From Optic Nerve Head to Visual Field

Purpose

Maps are required to relate visual field locations to optic nerve head regions. We compare individualized structure-to-function mapping (CUSTOM-MAP) to a population-derived mapping schema (POP-MAP).

Methods

Maps were compared for 118 eyes with glaucomatous field loss, circumpapillary retinal nerve fiber layer (cpRNFL) thickness measured using spectral domain optical coherence tomography (OCT), and two landmarks: the optic nerve head (ONH) position relative to the fovea and the temporal raphe angle. Locations with visual field damage (total deviation < −6 dB) were mapped to 30° ONH sectors centered on the angle given by each mapping schema. The concordance between damaged function and damaged structure was determined per location for various cpRNFL damage probability levels, with the number of concordant locations divided by the total number of damaged field locations providing a concordance ratio per eye.

Results

For the strictest concordance criteria (minimum cpRNFL thickness < 1% of normal), CUSTOM-MAP had higher mean concordance ratio than POP-MAP (60.5% c.f. 57.0% paired Wilcoxon, P = 0.005), with CUSTOM-MAP having a higher ratio in 43 eyes and POP-MAP having a higher ratio in 21 eyes. For all cpRNFL probability levels <20% of normal, more locations concorded for CUSTOM-MAP than POP-MAP. Inspection of the spatial patterns of differences revealed that CUSTOM-MAP often performed better in the arcuate regions, whereas POP-MAP had benefits inferior to the macula.

Conclusions

Anatomic parameters required for individualized structure-function mapping are readily measured with OCT and can provide improved concordance for some eyes.

Translational Relevance

Personalizing structure-function mapping may improve concordance between these measures. We provide a web-based tool for creating customized maps.

Focal Structure-Function Relationships in Primary Open-Angle Glaucoma Using OCT and OCT-A Measurements

Purpose

To evaluate the focal structure-function associations among visual field (VF) loss, optical coherence tomography angiography (OCT-A) vascular measurements, and optical coherence tomography (OCT) structural measurements in glaucoma.

Methods

In this cross-sectional study, subjects underwent standard automated perimetry, OCT-based nerve fiber thickness measurements, and OCT-A imaging. Mappings of focal VF test locations with OCT and OCT-A measurements were defined using anatomically adjusted nerve fiber trajectories and were studied using multivariate mixed-effects analysis. Segmented regression analysis was used to determine the presence of breakpoints in the structure-function associations.

Results

The study included 119 eyes from 86 Chinese subjects with primary open-angle glaucoma (POAG). VF mean deviation was significantly associated with global capillary perfusion density (β = 0.13 ± 0.08) and global retinal nerve fiber layer thickness (β = 0.09 ± 0.02). Focal capillary density (FCD) was significantly associated with VF losses at 34 VF test locations (66.7% of 24-2 VF), with 24 of the 34 locations being within 20° of retinal eccentricity. Focal nerve layer (FNL) thickness was significantly associated with 16 VF test locations (31.4% of 24-2 VF; eight locations within 20° eccentricity). For VF test locations in the central 10° VF, VF losses below the breakpoint were significantly associated with FCD (slope, 0.89 ± 0.12, P < 0.001), but not with FNL thickness (slope, 0.57 ± 0.39, P = 0.15).

Conclusions

Focal capillary densities were significantly associated with a wider range of visual field losses and in a larger proportion of the visual field compared to nerve fiber thickness.

Detection of functional deterioration in glaucoma by trend analysis using comprehensive overlapping clusters of locations

Detecting rapid visual field deterioration is crucial for individuals with glaucoma. Cluster trend analysis detects visual field deterioration with higher sensitivity than global analyses by using predefined non-overlapping subsets of visual field locations. However, it may miss small defects that straddle cluster borders. This study introduces a comprehensive set of overlapping clusters, and assesses whether this further improves progression detection. Clusters were defined as locations from where ganglion cell axons enter the optic nerve head within a θ° wide sector, centered at 1º intervals, for various θ. Deterioration in eyes with or at risk of glaucomatous visual field loss was “detected” if ≥ N_θ clusters had deteriorated with p < p_Cluster, chosen empirically to give 95% specificity based on permuting the series. N_θ was chosen to minimize the time to detect subsequently-confirmed deterioration in ≥ 1/3rd of eyes. Times to detect deterioration were compared using Cox survival models. Biannual series were available for 422 eyes of 214 participants. Predefined non-overlapping clusters detected subsequently-confirmed change in ≥ 1/3rd of eyes in 3.41 years (95% confidence interval 2.75–5.48 years). After equalizing specificity, no criteria based on comprehensive overlapping clusters detected deterioration significantly sooner. The quickest was 3.13 years (2.69–4.65) for θ° = 20° and N_θ = 25, but the comparison with non-overlapping clusters had p = 0.672. Any improvement in sensitivity for detecting deterioration when using a comprehensive set of overlapping clusters was negated by the need to maintain equal specificity. The existing cluster trend analysis using predefined non-overlapping clusters provides a useful tool for monitoring visual field progression.

Anatomical Location of the Raphe and Extended Raphe in the Human Retina: Implications for Assessment of the Optic Nerve with OCT

Purpose

To determine the location of (1) the superior–inferior watershed between the fovea and optic disc (extended raphe) at the peripapillary optical coherence tomography (OCT) measurement circle and (2) the raphe, temporal to the fovea.

Methods

We used existing data consisting of 2285 traced retinal nerve fiber bundle trajectories from 83 fundus images. For localization of the extended raphe at the 3.46-mm-diameter OCT measurement circle, trajectories were classified as belonging to the superior or inferior hemiretina, using predefined criteria. For the raphe, we localized the endings of trajectories coming from the superior and inferior arcuate bundles.

Results

At the measurement circle, the extended raphe is located 14° (range, 12°–16°) inferiorly to a horizontal line through the optic disc center. The raphe follows a horizontal line at the latitude of the fovea if the disc is assumed to be located 15° nasal to and 2° above the fovea.

Conclusions

At the measurement circle, OCT brands use either the 9 o'clock location or a straight line connecting the center of the optic disc and the fovea as a reference for separating the hemiretinas. This results, on average, in a 14° and 6° misalignment with respect to the anatomical watershed, respectively. For the macular area, the commonly used line through the center of the optic disc and the fovea fails to describe the raphe adequately.

Translational Relevance

An unbiased asymmetry assessment of the optic nerve requires a detailed knowledge of the shape and location of the (extended) raphe.

Improving the Detection of Glaucoma and Its Progression: A Topographical Approach

Glaucoma is typically defined as a progressive optic neuropathy characterized by a specific (arcuate) pattern of visual field (VF) and anatomic changes. Therefore, we should be comparing arcuate patterns of damage seen on VFs with those seen on optical coherence tomography (OCT) maps. Instead, clinicians often use summary metrics such as VF pattern standard deviation, OCT retinal nerve fiber (RNF) global thickness, etc. There are 2 major impediments to topographically comparing patterns of damage on VF and OCT maps. First, until recently, it was not easy to make these comparisons with commercial reports. While recent reports do make it easier to compare VF and OCT maps, they have shortcomings. In particular, the 24-2 VF covers a larger retinal region than the commercial OCT scans, and, further, it is not easy to understand the topographical relationship among the different maps/plots within the current OCT reports. Here we show how a model of RNF bundles can overcome these problems. The second major impediment is the lack of a quantitative, and automated, method for comparing patterns of damage seen on VF and OCT maps. However, it is now possible to objectively and automatically quantify this agreement. Together, the RNF bundle model and the automated structure-function method should improve the power of topographical methods for detecting glaucoma and its progression. This should prove useful in clinical studies and trials, as well as for training and validating artificial intelligence/deep learning approaches for these purposes.

Localization of Anatomical Features in Vascular-enhanced Enface OCT Images

Optical coherence tomography (OCT) allows in vivo volumetric imaging of the eye. Identification and localization of anatomical features in enface OCT are important steps in OCT-based image analysis. However the visibility of anatomical features in both structural OCT or vascular OCT angiography is limited. In this paper, we propose to use vascular-enhanced enface OCT image for the concurrent detection of anatomical features, using a FasterRCNN object detection framework based on convolutional networks. Transfer learning was applied to adapt pre-trained models as the backbone networks. Models were evaluated on a dataset of 419 images. The results showed that VGG-FasterRCNN achieved a mean average precision 0.77, with localization errors of 0.18 ± 0.10 mm and 0.24 ± 0.13 mm for the macula and optic disc respectively. The results are promising and suggest that this network could potentially be used to automatically and concurrently detect anatomical features.Clinical Relevance- Localization of anatomical features in enface OCT is needed for the automation of OCT image analysis protocols. The use of fast detection networks could potentially suggest image-based real-time tracking during image acquisition.

Artificial Intelligence Mapping of Structure to Function in Glaucoma

Purpose

To develop an artificial intelligence (AI)-based structure-function (SF) map relating retinal nerve fiber layer (RNFL) damage on spectral domain optical coherence tomography (SDOCT) to functional loss on standard automated perimetry (SAP).

Methods

The study included 26,499 pairs of SAP and SDOCT from 15,173 eyes of 8878 patients with glaucoma or suspected of having the disease extracted from the Duke Glaucoma Registry. The data set was randomly divided at the patient level in training and test sets. A convolutional neural network (CNN) was initially trained and validated to predict the 52 sensitivity threshold points of the 24-2 SAP from the 768 RNFL thickness points of the SDOCT peripapillary scan. Simulated localized RNFL defects of varied locations and depths were created by modifying the normal average peripapillary RNFL profile. The simulated profiles were then fed to the previously trained CNN, and the topographic SF relationships between structural defects and SAP functional losses were investigated.

Results

The CNN predictions had an average correlation coefficient of 0.60 (P < 0.001) with the measured values from SAP and a mean absolute error of 4.25 dB. Simulated RNFL defects led to well-defined arcuate or paracentral visual field losses in the opposite hemifield, which varied according to the location and depth of the simulations.

Conclusions

A CNN was capable of predicting SAP sensitivity thresholds from SDOCT RNFL thickness measurements and generate an SF map from simulated defects.

Translational Relevance

AI-based SF map improves the understanding of how SDOCT losses translate into detectable SAP damage.

[Structure and function of the visual pathway]

Humans receive information from their environment mainly via the visual system. Signals from the photoreceptors of the retina via bipolar and ganglion cells are projected onto specific neuronal subpopulations in the lateral geniculate body and from there are forwarded to appropriate layers of the primary visual cortex. The most important anatomical and functional features of the visual system are explained. For this purpose, a selective literature search was carried out in the databases PubMed (also in Europe PubMed Central), Psychline, Google Scholar, Cochrane Library and Web of Science as well as additional information in relevant books or websites in the fields of (neuro)anatomy, (neuro)physiology, (neuro)ophthalmology and (neuro)otology, among others with the search terms Sehbahn, visual system, visual pathway, receptors, spatial cognition and visual cognition.