Developing a Continuous Severity Scale for Macular Telangiectasia Type 2 Using Deep Learning and Implications for Disease Grading.

PURPOSE: Deep learning (DL) models have achieved state-of-the-art medical diagnosis classification accuracy. Current models are limited by discrete diagnosis labels, but could yield more information with diagnosis in a continuous scale. We developed a novel continuous severity scaling system for macular telangiectasia (MacTel) type 2 by combining a DL classification model with uniform manifold approximation and projection (UMAP). DESIGN: We used a DL network to learn a feature representation of MacTel severity from discrete severity labels and applied UMAP to embed this feature representation into 2 dimensions, thereby creating a continuous MacTel severity scale. PARTICIPANTS: A total of 2003 OCT volumes were analyzed from 1089 MacTel Project participants. METHODS: We trained a multiview DL classifier using multiple B-scans from OCT volumes to learn a previously published discrete 7-step MacTel severity scale. The classifiers' last feature layer was extracted as input for UMAP, which embedded these features into a continuous 2-dimensional manifold. The DL classifier was assessed in terms of test accuracy. Rank correlation for the continuous UMAP scale against the previously published scale was calculated. Additionally, the UMAP scale was assessed in the κ agreement against 5 clinical experts on 100 pairs of patient volumes. For each pair of patient volumes, clinical experts were asked to select the volume with more severe MacTel disease and to compare them against the UMAP scale. MAIN OUTCOME MEASURES: Classification accuracy for the DL classifier and κ agreement versus clinical experts for UMAP. RESULTS: The multiview DL classifier achieved top 1 accuracy of 63.3% (186/294) on held-out test OCT volumes. The UMAP metric showed a clear continuous gradation of MacTel severity with a Spearman rank correlation of 0.84 with the previously published scale. Furthermore, the continuous UMAP metric achieved κ agreements of 0.56 to 0.63 with 5 clinical experts, which was comparable with interobserver κ values. CONCLUSIONS: Our UMAP embedding generated a continuous MacTel severity scale, without requiring continuous training labels. This technique can be applied to other diseases and may lead to more accurate diagnosis, improved understanding of disease progression, and key imaging features for pathologic characteristics. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Integrating human iPSC-derived macrophage progenitors into retinal organoids to generate a mature retinal microglial niche.

In the retina, microglia are resident immune cells that are essential for development and function. Retinal microglia play a central role in mediating pathological degeneration in diseases such as glaucoma, retinitis pigmentosa, age-related neurodegeneration, ischemic retinopathy, and diabetic retinopathy. Current models of mature human retinal organoids (ROs) derived from iPS cell (hiPSC) do not contain resident microglia integrated into retinal layers. Increasing cellular diversity in ROs by including resident microglia would more accurately represent the native retina and better model diseases in which microglia play a key role. In this study, we develop a new 3D in vitro tissue model of microglia-containing retinal organoids by co-culturing ROs and hiPSC-derived macrophage precursor cells (MPCs). We optimized the parameters for successful integration of MPCs into retinal organoids. We show that while in the ROs, MPCs migrate to the equivalent of the outer plexiform layer where retinal microglia cells reside in healthy retinal tissue. While there, they develop a mature morphology characterized by small cell bodies and long branching processes which is only observed in vivo. During this maturation process these MPCs cycle through an activated phase followed by a stable mature microglial phase as seen by the down regulation of pro-inflammatory cytokines and upregulation of anti-inflammatory cytokines. Finally, we characterized mature ROs with integrated MPCs using RNAseq showing an enrichment of cell-type specific microglia markers. We propose that this co-culture system may be useful for understanding the pathogenesis of retinal diseases involving retinal microglia and for drug discovery directly in human tissue.

An allosteric peptide inhibitor of HIF-1α regulates hypoxia-induced retinal neovascularization.

Retinal neovascularization (NV), a leading cause of vision loss, results from localized hypoxia that stabilizes the hypoxia-inducible transcription factors HIF-1α and HIF-2α, enabling the expression of angiogenic factors and genes required to maintain homeostasis under conditions of oxygen stress. HIF transcriptional activity depends on the interaction between its intrinsically disordered C-terminal domain and the transcriptional coactivators CBP/p300. Much effort is currently directed at disrupting protein-protein interactions between disease-associated transcription factors like HIF and their cellular partners. The intrinsically disordered protein CITED2, a direct product of HIF-mediated transcription, functions as a hypersensitive negative regulator that attenuates the hypoxic response by competing allosterically with HIF-1α for binding to CBP/p300. Here, we show that a peptide fragment of CITED2 is taken up by retinal cells and efficiently regulates pathological angiogenesis in murine models of ischemic retinopathy. Both vaso-obliteration (VO) and NV were significantly inhibited in an oxygen-induced retinopathy (OIR) model following intravitreal injection of the CITED2 peptide. The CITED2 peptide localized to retinal neurons and glia, resulting in decreased expression of HIF target genes. Aflibercept, a commonly used anti-VEGF therapy for retinal neovascular diseases, rescued NV but not VO in OIR. However, a combination of the CITED2 peptide and a reduced dose of aflibercept significantly decreased both NV and VO. In contrast to anti-VEGF agents, the CITED2 peptide can rescue hypoxia-induced retinal NV by modulating the hypoxic response through direct competition with HIF for CBP/p300, suggesting a dual targeting strategy for treatment of ischemic retinal diseases and other neovascular disorders.

Deletion of Tgfß signal in activated microglia prolongs hypoxia-induced retinal neovascularization enhancing Igf1 expression and retinal leukostasis.

Retinal neovascularization (NV) is the major cause of severe visual impairment in patients with ischemic eye diseases. While it is known that retinal microglia contribute to both physiological and pathological angiogenesis, the molecular mechanisms by which these glia regulate pathological NV have not been fully elucidated. In this study, we utilized a retinal microglia-specific Transforming Growth Factor-ß (Tgfß) receptor knock out mouse model and human iPSC-derived microglia to examine the role of Tgfß signaling in activated microglia during retinal NV. Using a tamoxifen-inducible, microglia-specific Tgfß receptor type 2 (Tgfßr2) knockout mouse [Tgfßr2 KO (ΔMG)] we show that Tgfß signaling in microglia actively represses leukostasis in retinal vessels. Furthermore, we show that Tgfß signaling represses expression of the pro-angiogenic factor, Insulin-like growth factor 1 (Igf1), independent of Vegf regulation. Using the mouse model of oxygen-induced retinopathy (OIR) we show that Tgfß signaling in activated microglia plays a role in hypoxia-induced NV where a loss in Tgfß signaling microglia exacerbates and prolongs retinal NV in OIR. Using human iPSC-derived microglia cells in an in vitro assay, we validate the role of Transforming Growth Factor-ß1 (Tgfß1) in regulating Igf1 expression in hypoxic conditions. Finally, we show that Tgfß signaling in microglia is essential for microglial homeostasis and that the disruption of Tgfß signaling in microglia exacerbates retinal NV in OIR by promoting leukostasis and Igf1 expression.

Serine and Lipid Metabolism in Macular Disease and Peripheral Neuropathy.

BACKGROUND: Identifying mechanisms of diseases with complex inheritance patterns, such as macular telangiectasia type 2, is challenging. A link between macular telangiectasia type 2 and altered serine metabolism has been established previously. METHODS: Through exome sequence analysis of a patient with macular telangiectasia type 2 and his family members, we identified a variant in SPTLC1 encoding a subunit of serine palmitoyltransferase (SPT). Because mutations affecting SPT are known to cause hereditary sensory and autonomic neuropathy type 1 (HSAN1), we examined 10 additional persons with HSAN1 for ophthalmologic disease. We assayed serum amino acid and sphingoid base levels, including levels of deoxysphingolipids, in patients who had macular telangiectasia type 2 but did not have HSAN1 or pathogenic variants affecting SPT. We characterized mice with low serine levels and tested the effects of deoxysphingolipids on human retinal organoids. RESULTS: Two variants known to cause HSAN1 were identified as causal for macular telangiectasia type 2: of 11 patients with HSAN1, 9 also had macular telangiectasia type 2. Circulating deoxysphingolipid levels were 84.2% higher among 125 patients with macular telangiectasia type 2 who did not have pathogenic variants affecting SPT than among 94 unaffected controls. Deoxysphingolipid levels were negatively correlated with serine levels, which were 20.6% lower than among controls. Reduction of serine levels in mice led to increases in levels of retinal deoxysphingolipids and compromised visual function. Deoxysphingolipids caused photoreceptor-cell death in retinal organoids, but not in the presence of regulators of lipid metabolism. CONCLUSIONS: Elevated levels of atypical deoxysphingolipids, caused by variant SPTLC1 or SPTLC2 or by low serine levels, were risk factors for macular telangiectasia type 2, as well as for peripheral neuropathy. (Funded by the Lowy Medical Research Institute and others.).

Description of a patient cohort with Hereditary Sensory Neuropathy type 1 without retinal disease Macular Telangiectasia type 2 - implications for retinal screening in HSN1.

Pathogenic variants in the genes encoding serine palmitoyl transferase (SPTLC1 or SPTLC2) are the most common causes of the rare peripheral nerve disorder Hereditary Sensory Neuropathy Type 1 (HSN1). Macular telangiectasia type 2 (MacTel), a retinal disorder associated with disordered serine-glycine metabolism, has been described in some patients with HSN1. This study aims to further investigate this association in a cohort of people with HSN1. Fourteen patients with a clinically and genetically confirmed diagnosis of HSN1 from the National Hospital for Neurology and Neurosurgery (NHNN, University College London Hospitals NHS Foundation Trust, London, United Kingdom) were recruited to the MacTel Registry, between July 2018 and April 2019. Two additional patients were identified from the dataset of the international clinical registry study (www.lmri.net). Ocular examination included fundus autofluorescence, blue light and infrared reflectance, macular pigment optical density mapping and optical coherence tomography. Twelve patients had a pathogenic variant in the SPTLC1 gene, with p.Cys133Trp in 11 cases (92%) and p.Cys133Tyr in one case (8%). Four patients had a variant in the SPTLC2 gene. None of the patients showed clinical evidence of MacTel. The link between HSN1 and MacTel seems more complex than can solely be explained by the genetic variants. An extension of the spectrum of SPTLC1/2-related disease with phenotypic pleiotropy is proposed. HSN1 patients should be screened for visual symptoms and referred for specialist retinal screening, but the association of the two diseases is likely to be variable and remains unexplained.

Interferon-γ inhibits retinal neovascularization in a mouse model of ischemic retinopathy.

Interferon-γ (IFNG) is one of the key cytokines that regulates both innate and adaptive immune responses in the body. However, the role of IFNG in the regulation of vascularization, especially in the context of Vascular endothelial growth factor A (VEGFa)-induced angiogenesis is not clarified. Here, we report that IFNG shows potent anti-angiogenic potential against VEGFa-induced angiogenesis. IFNG significantly inhibited proliferation, migration, and tube formation of Human umbilical vein endothelial cells (HUVECs) both under basal and VEGFa-treated conditions. Intriguingly, Knockdown (KD) of STAT1 abolished the inhibitory effect of IFNG on VEGFa-induced angiogenic processes in HUVECs. Furthermore, IFNG exhibited potent anti-angiogenic efficacy in the mouse model of oxygen-induced retinopathy (OIR), an in vivo model for hypoxia-induced retinal neovascularization, without induction of functional side effects. Taken together, these results show that IFNG plays a crucial role in the regulation of VEGFa-dependent angiogenesis, suggesting its potential therapeutic applicability in neovascular diseases.

HYPERREFLECTIVITY ON OPTICAL COHERENCE TOMOGRAPHY IN MACULAR TELANGIECTASIA TYPE 2.

PURPOSE: To define, characterize, and classify hyperreflectivity on optical coherence tomography and report its prevalence in macular telangiectasia Type 2. METHODS: In a primary cross-sectional analysis, multimodal imaging data were retrospectively analyzed. The definition of hyperreflectivity and neovascularization on optical coherence tomography followed optical coherence tomography angiography-based criteria. Eyes were graded for the presence of hyperreflectivity and neovascularization and further categorized into three classes based on position and extent of hyperreflectivity. In a secondary analysis, eyes were reviewed for ≥24 months using optical coherence tomography imaging. RESULTS: Three hundred and twenty-two eyes from 161 patients were analyzed in the cross-sectional analysis. Hyperreflectivity was found in 177 (55%) and neovascular membranes in 49 (15%) eyes. Hyperreflectivity correlated significantly with parameters indicative of disease progression. In the longitudinal analysis, 206 eyes from 103 patients were reviewed over a mean of 35.6 months. 17/86 eyes (20%) showed a de novo development of hyperreflectivity. 8/29 eyes (28%) with preexistent intraretinal hyperreflectivity developed outer retinal hyperreflectivity. A high proportion of eyes with outer retinal hyperreflectivity (17/52 [33%]) developed neovascular membranes. CONCLUSION: Hyperreflectivity represents a common finding in macular telangiectasia Type 2 but lacks a uniform definition. We propose a hyperreflectivity grading scale that may help to estimate disease progression and identify eyes at risk for developing neovascular membranes.

Beyond Performance Metrics: Automatic Deep Learning Retinal OCT Analysis Reproduces Clinical Trial Outcome.

PURPOSE: To validate the efficacy of a fully automatic, deep learning-based segmentation algorithm beyond conventional performance metrics by measuring the primary outcome of a clinical trial for macular telangiectasia type 2 (MacTel2). DESIGN: Evaluation of diagnostic test or technology. PARTICIPANTS: A total of 92 eyes from 62 participants with MacTel2 from a phase 2 clinical trial (NCT01949324) randomized to 1 of 2 treatment groups METHODS: The ellipsoid zone (EZ) defect areas were measured on spectral domain OCT images of each eye at 2 time points (baseline and month 24) by a fully automatic, deep learning-based segmentation algorithm. The change in EZ defect area from baseline to month 24 was calculated and analyzed according to the clinical trial protocol. MAIN OUTCOME MEASURE: Difference in the change in EZ defect area from baseline to month 24 between the 2 treatment groups. RESULTS: The difference in the change in EZ defect area from baseline to month 24 between the 2 treatment groups measured by the fully automatic segmentation algorithm was 0.072±0.035 mm2 (P = 0.021). This was comparable to the outcome of the clinical trial using semiautomatic measurements by expert readers, 0.065±0.033 mm2 (P = 0.025). CONCLUSIONS: The fully automatic segmentation algorithm was as accurate as semiautomatic expert segmentation to assess EZ defect areas and was able to reliably reproduce the statistically significant primary outcome measure of the clinical trial. This approach, to validate the performance of an automatic segmentation algorithm on the primary clinical trial end point, provides a robust gauge of its clinical applicability.

WITHDRAWN: Hyper-reflectivity on optical coherence tomography in macular telangiectasia type 2.

Ahead of Print article withdrawn by publisher.

miR-30a-5p inhibition promotes interaction of Fas+ endothelial cells and FasL+ microglia to decrease pathological neovascularization and promote physiological angiogenesis.

Ischemia-induced angiogenesis contributes to various neuronal and retinal diseases, and often results in neurodegeneration and visual impairment. Current treatments involve the use of anti-VEGF agents but are not successful in all cases. In this study we determined that miR-30a-5p is another important mediator of retinal angiogenesis. Using a rodent model of ischemic retinopathy, we show that inhibiting miR-30a-5p reduces neovascularization and promotes tissue repair, through modulation of microglial and endothelial cell cross-talk. miR-30a-5p inhibition results in increased expression of the death receptor Fas and CCL2, to decrease endothelial cell survival and promote microglial migration and phagocytic function in focal regions of ischemic injury. Our data suggest that miR-30a-5p inhibition accelerates tissue repair by enhancing FasL-Fas crosstalk between microglia and endothelial cells, to promote endothelial cell apoptosis and removal of dead endothelial cells. Finally, we found that miR-30a levels were increased in the vitreous of patients with proliferative diabetic retinopathy. Our study identifies a role for miR-30a in the pathogenesis of neovascular retinal disease by modulating microglial and endothelial cell function, and suggests it may be a therapeutic target to treat ischemia-mediated conditions.

Effect of Ciliary Neurotrophic Factor on Retinal Neurodegeneration in Patients with Macular Telangiectasia Type 2: A Randomized Clinical Trial.

PURPOSE: To test the effects of an encapsulated cell-based delivery of a neuroprotective agent, ciliary neurotrophic factor (CNTF), on progression of macular telangiectasia type 2, a neurodegenerative disease with no proven effective therapy. DESIGN: Randomized sham-controlled clinical trial. PARTICIPANTS: Ninety-nine study eyes of 67 eligible participants were enrolled. METHODS: Single-masked randomized clinical trial of 24 months' duration conducted from May 2014 through April 2017 in 11 clinical centers of retinal specialists in the United States and Australia. Participants were randomized 1:1 to surgical implantation of intravitreal sustained delivery of human CNTF versus a sham procedure. MAIN OUTCOME MEASURES: The primary outcome was the difference in the area of neurodegeneration as measured in the area of the ellipsoid zone disruption (or photoreceptor loss) measured on spectral-domain (SD) OCT images at 24 months from baseline between the treated and untreated groups. Secondary outcomes included comparison of visual function changes between treatment groups. RESULTS: Among the 67 participants who were randomized (mean age, 62±8.9 years; 41 women [61%]; 58 white persons [86%]), 65 (97%) completed the study. Two participants (3 study eyes) died and 3 participants (4 eyes) were found ineligible. The eyes receiving sham treatment had 31% greater progression of neurodegeneration than the CNTF-treated eyes. The difference in mean area of photoreceptor loss was 0.05±0.03 mm2 (P = 0.04) at 24 months. Retinal sensitivity changes, measured using microperimetry, were correlated highly with the changes in the area of photoreceptor loss (r = 0.86; P < 0.0001). The mean retinal sensitivity loss of the sham group was 45% greater than that of the treated group (decrease, 15.81±8.93 dB; P = 0.07). Reading speed deteriorated in the sham group (-13.9 words per minute) with no loss in the treated group (P = 0.02). Serious adverse ocular effects were found in 2 of 51 persons (4%) in the sham group and 2 of 48 persons (4%) in the treated group. CONCLUSIONS: In participants with macular telangiectasia type 2, a surgical implant that released CNTF into the vitreous cavity, compared with a sham procedure, slowed the progression of retinal degeneration. Further research is needed to assess longer-term clinical outcomes and safety.

Antibody-Mediated Inhibition of Tspan12 Ameliorates Vasoproliferative Retinopathy Through Suppression of ß-Catenin Signaling.

BACKGROUND: Anti-angiogenic biologicals represent an important concept for the treatment of vasoproliferative diseases. However, the need for continued treatment, the presence of nonresponders, and the risk of long-term side effects limit the success of existing therapeutic agents. Although Tspan12 has been shown to regulate retinal vascular development, nothing is known about its involvement in neovascular disease and its potential as a novel therapeutic target for the treatment of vasoproliferative diseases. METHODS: Rodent models of retinal neovascular disease, including the mouse model of oxygen-induced retinopathy and the very low density lipoprotein receptor knockout mouse model were analyzed for Tspan/ß-catenin regulation. Screening of a phage display of a human combinatorial antibody (Ab) library was used for the development of a high-affinity Ab against Tspan12. Therapeutic effects of the newly developed Ab on vascular endothelial cells were tested in vitro and in vivo in the oxygen-induced retinopathy and very low density lipoprotein receptor knockout mouse model. RESULTS: The newly developed anti-Tspan12 Ab exhibited potent inhibitory effects on endothelial cell migration and tube formation. Mechanistic studies confirmed that the Ab inhibited the interaction between Tspan12 and Frizzled-4 and effectively modulates ß-catenin levels and target genes in vascular endothelial cells. Tspan12/ß-catenin signaling was activated in response to acute and chronic stress in the oxygen-induced retinopathy and very low density lipoprotein receptor mouse model of proliferative retinopathy. Intravitreal application of the Ab showed significant therapeutic effects in both models without inducing negative side effects on retina function. Moreover, combined intravitreal injection of the Ab with a known vascular endothelial growth factor inhibitor, Aflibercept, resulted in significant enhancement of the therapeutic efficacy of each monotherapy. Combination therapy with the Tspan12 blocking antibody can be used to reduce anti-vascular endothelial growth factor doses, thus decreasing the risk of long-term off-target effects. CONCLUSIONS: Tspan12/ß-catenin signaling is critical for the progression of vasoproliferative disease. The newly developed anti-Tspan12 antibody has therapeutic effects in vasoproliferative retinopathy and can enhance the potency of existing anti- vascular endothelial growth factor agents.

Special Commentary: Early Clinical Development of Cell Replacement Therapy: Considerations for the National Eye Institute Audacious Goals Initiative.

The National Eye Institute launched the Audacious Goals Initiative (AGI) in 2013 with the aim "to restore vision through the regeneration of neurons and neural connections in the eye and visual system." An AGI Town Hall held at the Association for Research in Vision and Ophthalmology Annual Meeting in 2016 brought together basic, translational, and clinical scientists to address the clinical implications of the AGI, with a particular emphasis on diseases amenable to regenerative medicine and strategies to deal with barriers to progess. An example of such a barrier is that replacement of lost neurons may be insufficient because damage to other neurons and non-neuronal cells is common in retinal and optic nerve disease. Reparative processes such as gliosis and fibrosis also can make it difficult to replenish and regenerate neurons. Other issues include choice of animal models, selecting appropriate endpoints, ethics of informed consent, and regulatory issues. Another area critical to next steps in the AGI is the choice of target diseases and the stage at which early development studies should be focused. For example, an advantage of doing clinical trials in patients with early disease is that supporting cellular and structural constituents are still likely to be present. However, regenerative studies in patients with late disease make it easier to detect the effects of replacement therapy against the background of severe visual loss, whereas it may be harder to detect incremental improvement in visual function in those with early disease and considerable remaining visual function. Achieving the goals of the AGI also requires preclinical advances, new imaging techniques, and optimizing translational issues. The work of the AGI is expected to take at least 10 years but should eventually result in therapies to restore some degree of vision to the blind.

Hypoxia-inducible factor (HIF)/vascular endothelial growth factor (VEGF) signaling in the retina.

Over a span of two decades, it has become increasingly clear that vascular endothelial growth factor (VEGF) plays an important role in the pathogenesis of retinal diseases including age-related macular degeneration (AMD) and diabetic retinopathy (DR). Based on these observations, anti-VEGF therapies are being developed and approved for clinical use in the treatment of neovascular eye diseases. Hypoxia-inducible factors (HIFs) are transcriptional factors that are stabilized and activated under hypoxic conditions and induce expression of gene products, including VEGF, that are required for cell survival under hypoxia. Here we discuss recent findings from our lab and others that define roles of the HIF-VEGF axis in the retina.

Utilizing stem cell-derived RPE cells as a therapeutic intervention for age-related macular degeneration.

PURPOSE: Degeneration or dysfunction of the retinal pigment epithelium (RPE) can induce secondary photoreceptor atrophy and catastrophic vision loss in patients with age-related macular degeneration (AMD). AMD is the leading cause of vision loss in the elderly in industrialized countries and no cure exists for the "dry" or atrophic form to date. However, recent pre-clinical data from several groups suggests that embryonic stem cell-derived RPE cell transplantation may prevent photoreceptor degeneration in animal models of RPE degeneration. Another approach may be to derive RPE cells from autologous induced pluripotent stem cells (iPSCs) reprogrammed from dermal tissue. However, the safety of this approach has been questioned on several levels. In this chapter we will summarize work reported by several groups, including our own, that clearly demonstrate that transplanted RPE cells can provide anatomical and functional photoreceptor rescue in animal models of retinal degeneration. We will also discuss some of the prevailing concerns and challenges associated with this technique.

Applications of Deep Learning: Automated Assessment of Vascular Tortuosity in Mouse Models of Oxygen-Induced Retinopathy.

Objective: To develop a generative adversarial network (GAN) to segment major blood vessels from retinal flat-mount images from oxygen-induced retinopathy (OIR) and demonstrate the utility of these GAN-generated vessel segmentations in quantifying vascular tortuosity. Design: Development and validation of GAN. Subjects: Three datasets containing 1084, 50, and 20 flat-mount mice retina images with various stains used and ages at sacrifice acquired from previously published manuscripts. Methods: Four graders manually segmented major blood vessels from flat-mount images of retinas from OIR mice. Pix2Pix, a high-resolution GAN, was trained on 984 pairs of raw flat-mount images and manual vessel segmentations and then tested on 100 and 50 image pairs from a held-out and external test set, respectively. GAN-generated and manual vessel segmentations were then used as an input into a previously published algorithm (iROP-Assist) to generate a vascular cumulative tortuosity index (CTI) for 20 image pairs containing mouse eyes treated with aflibercept versus control. Main Outcome Measures: Mean dice coefficients were used to compare segmentation accuracy between the GAN-generated and manually annotated segmentation maps. For the image pairs treated with aflibercept versus control, mean CTIs were also calculated for both GAN-generated and manual vessel maps. Statistical significance was evaluated using Wilcoxon signed-rank tests (P ≤ 0.05 threshold for significance). Results: The dice coefficient for the GAN-generated versus manual vessel segmentations was 0.75 ± 0.27 and 0.77 ± 0.17 for the held-out test set and external test set, respectively. The mean CTI generated from the GAN-generated and manual vessel segmentations was 1.12 ± 0.07 versus 1.03 ± 0.02 (P = 0.003) and 1.06 ± 0.04 versus 1.01 ± 0.01 (P < 0.001), respectively, for eyes treated with aflibercept versus control, demonstrating that vascular tortuosity was rescued by aflibercept when quantified by GAN-generated and manual vessel segmentations. Conclusions: GANs can be used to accurately generate vessel map segmentations from flat-mount images. These vessel maps may be used to evaluate novel metrics of vascular tortuosity in OIR, such as CTI, and have the potential to accelerate research in treatments for ischemic retinopathies. Financial Disclosures: The author(s) have no proprietary or commercial interest in any materials discussed in this article.

Development of a Semi-automated Computer-based Tool for the Quantification of Vascular Tortuosity in the Murine Retina.

Purpose: The murine oxygen-induced retinopathy (OIR) model is one of the most widely used animal models of ischemic retinopathy, mimicking hallmark pathophysiology of initial vaso-obliteration (VO) resulting in ischemia that drives neovascularization (NV). In addition to NV and VO, human ischemic retinopathies, including retinopathy of prematurity (ROP), are characterized by increased vascular tortuosity. Vascular tortuosity is an indicator of disease severity, need to treat, and treatment response in ROP. Current literature investigating novel therapeutics in the OIR model often report their effects on NV and VO, and measurements of vascular tortuosity are less commonly performed. No standardized quantification of vascular tortuosity exists to date despite this metric's relevance to human disease. This proof-of-concept study aimed to apply a previously published semi-automated computer-based image analysis approach (iROP-Assist) to develop a new tool to quantify vascular tortuosity in mouse models. Design: Experimental study. Subjects: C57BL/6J mice subjected to the OIR model. Methods: In a pilot study, vasculature was manually segmented on flat-mount images of OIR and normoxic (NOX) mice retinas and segmentations were analyzed with iROP-Assist to quantify vascular tortuosity metrics. In a large cohort of age-matched (postnatal day 12 [P12], P17, P25) NOX and OIR mice retinas, NV, VO, and vascular tortuosity were quantified and compared. In a third experiment, vascular tortuosity in OIR mice retinas was quantified on P17 following intravitreal injection with anti-VEGF (aflibercept) or Immunoglobulin G isotype control on P12. Main Outcome Measures: Vascular tortuosity. Results: Cumulative tortuosity index was the best metric produced by iROP-Assist for discriminating between OIR mice and NOX controls. Increased vascular tortuosity correlated with disease activity in OIR. Treatment of OIR mice with aflibercept rescued vascular tortuosity. Conclusions: Vascular tortuosity is a quantifiable feature of the OIR model that correlates with disease severity and may be quickly and accurately quantified using the iROP-Assist algorithm. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Self-Supervised Learning for Improved Optical Coherence Tomography Detection of Macular Telangiectasia Type 2.

Importance: Deep learning image analysis often depends on large, labeled datasets, which are difficult to obtain for rare diseases. Objective: To develop a self-supervised approach for automated classification of macular telangiectasia type 2 (MacTel) on optical coherence tomography (OCT) with limited labeled data. Design, Setting, and Participants: This was a retrospective comparative study. OCT images from May 2014 to May 2019 were collected by the Lowy Medical Research Institute, La Jolla, California, and the University of Washington, Seattle, from January 2016 to October 2022. Clinical diagnoses of patients with and without MacTel were confirmed by retina specialists. Data were analyzed from January to September 2023. Exposures: Two convolutional neural networks were pretrained using the Bootstrap Your Own Latent algorithm on unlabeled training data and fine-tuned with labeled training data to predict MacTel (self-supervised method). ResNet18 and ResNet50 models were also trained using all labeled data (supervised method). Main Outcomes and Measures: The ground truth yes vs no MacTel diagnosis is determined by retinal specialists based on spectral-domain OCT. The models' predictions were compared against human graders using accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), area under precision recall curve (AUPRC), and area under the receiver operating characteristic curve (AUROC). Uniform manifold approximation and projection was performed for dimension reduction and GradCAM visualizations for supervised and self-supervised methods. Results: A total of 2636 OCT scans from 780 patients with MacTel and 131 patients without MacTel were included from the MacTel Project (mean [SD] age, 60.8 [11.7] years; 63.8% female), and another 2564 from 1769 patients without MacTel from the University of Washington (mean [SD] age, 61.2 [18.1] years; 53.4% female). The self-supervised approach fine-tuned on 100% of the labeled training data with ResNet50 as the feature extractor performed the best, achieving an AUPRC of 0.971 (95% CI, 0.969-0.972), an AUROC of 0.970 (95% CI, 0.970-0.973), accuracy of 0.898%, sensitivity of 0.898, specificity of 0.949, PPV of 0.935, and NPV of 0.919. With only 419 OCT volumes (185 MacTel patients in 10% of labeled training dataset), the ResNet18 self-supervised model achieved comparable performance, with an AUPRC of 0.958 (95% CI, 0.957-0.960), an AUROC of 0.966 (95% CI, 0.964-0.967), and accuracy, sensitivity, specificity, PPV, and NPV of 90.2%, 0.884, 0.916, 0.896, and 0.906, respectively. The self-supervised models showed better agreement with the more experienced human expert graders. Conclusions and Relevance: The findings suggest that self-supervised learning may improve the accuracy of automated MacTel vs non-MacTel binary classification on OCT with limited labeled training data, and these approaches may be applicable to other rare diseases, although further research is warranted.