Multi-granularity learning of explicit geometric constraint and contrast for label-efficient medical image segmentation and differentiable clinical function assessment.

Automated segmentation is a challenging task in medical image analysis that usually requires a large amount of manually labeled data. However, most current supervised learning based algorithms suffer from insufficient manual annotations, posing a significant difficulty for accurate and robust segmentation. In addition, most current semi-supervised methods lack explicit representations of geometric structure and semantic information, restricting segmentation accuracy. In this work, we propose a hybrid framework to learn polygon vertices, region masks, and their boundaries in a weakly/semi-supervised manner that significantly advances geometric and semantic representations. Firstly, we propose multi-granularity learning of explicit geometric structure constraints via polygon vertices (PolyV) and pixel-wise region (PixelR) segmentation masks in a semi-supervised manner. Secondly, we propose eliminating boundary ambiguity by using an explicit contrastive objective to learn a discriminative feature space of boundary contours at the pixel level with limited annotations. Thirdly, we exploit the task-specific clinical domain knowledge to differentiate the clinical function assessment end-to-end. The ground truth of clinical function assessment, on the other hand, can serve as auxiliary weak supervision for PolyV and PixelR learning. We evaluate the proposed framework on two tasks, including optic disc (OD) and cup (OC) segmentation along with vertical cup-to-disc ratio (vCDR) estimation in fundus images; left ventricle (LV) segmentation at end-diastolic and end-systolic frames along with ejection fraction (LVEF) estimation in two-dimensional echocardiography images. Experiments on nine large-scale datasets of the two tasks under different label settings demonstrate our model's superior performance on segmentation and clinical function assessment.

Linking Vascular Structure and Function: Image-Based Virtual Populations of the Retina.

Purpose: This study explored the relationship among microvascular parameters as delineated by optical coherence tomography angiography (OCTA) and retinal perfusion. Here, we introduce a versatile framework to examine the interplay between the retinal vascular structure and function by generating virtual vasculatures from central retinal vessels to macular capillaries. Also, we have developed a hemodynamics model that evaluates the associations between vascular morphology and retinal perfusion. Methods: The generation of the vasculature is based on the distribution of four clinical parameters pertaining to the dimension and blood pressure of the central retinal vessels, constructive constrained optimization, and Voronoi diagrams. Arterial and venous trees are generated in the temporal retina and connected through three layers of capillaries at different depths in the macula. The correlations between total retinal blood flow and macular flow fraction and vascular morphology are derived as Spearman rank coefficients, and uncertainty from input parameters is quantified. Results: A virtual cohort of 200 healthy vasculatures was generated. Means and standard deviations for retinal blood flow and macular flow fraction were 20.80 ± 7.86 µL/min and 15.04% ± 5.42%, respectively. Retinal blood flow was correlated with vessel area density, vessel diameter index, fractal dimension, and vessel caliber index. The macular flow fraction was not correlated with any morphological metrics. Conclusions: The proposed framework is able to reproduce vascular networks in the macula that are morphologically and functionally similar to real vasculature. The framework provides quantitative insights into how macular perfusion can be affected by changes in vascular morphology delineated on OCTA.

Using Deep Learning to Segment Retinal Vascular Leakage and Occlusion in Retinal Vasculitis.

PURPOSE: Retinal vasculitis (RV) is characterised by retinal vascular leakage, occlusion or both on fluorescein angiography (FA). There is no standard scheme available to segment RV features. We aimed to develop a deep learning model to segment both vascular leakage and occlusion in RV. METHODS: Four hundred and sixty-three FA images from 82 patients with retinal vasculitis were used to develop a deep learning model, in 60:20:20 ratio for training:validation:testing. Parameters, including deep learning architectures (DeeplabV3+, UNet++ and UNet), were altered to find the best binary segmentation model separately for retinal vascular leakage and occlusion, using a Dice score to determine the reliability of each model. RESULTS: Our best model for vascular leakage had a Dice score of 0.6279 (95% confidence interval (CI) 0.5584-0.6974). For occlusion, the best model achieved a Dice score of 0.6992 (95% CI 0.6109-0.7874). CONCLUSION: Our RV segmentation models could perform reliable segmentation for retinal vascular leakage and occlusion in FAs of RV patients.

Presumed topiramate-induced retinopathy in a 58-year-old woman.

We present a case of presumed topiramate-induced retinopathy in a 58-year-old woman who presented with progressive, bilateral visual loss following a 3- to 4-year history of oral topiramate intake for migraine. She reported difficulty with light adaptation, hemeralopia, and color desaturation. Her best-corrected visual acuity was 1/60 (20/1200) in the right eye and 6/18 (20/60) in the left eye, and she performed poorly on Ishihara color plate testing. Anterior segment examination was normal; dilated funduscopy showed mild macular pigmentary changes. Optical coherence tomography revealed subtle thinning and reduced reflectivity of the subfoveal ellipsoid zone and interdigitation zone bilaterally, associated with increased foveal autofluorescence. Humphrey visual field 24-2 revealed central defects. Electrodiagnostic testing showed a reduced and delayed b-wave and a normal a-wave on photopic full-field electroretinogram (ERG), with normal scotopic responses; multifocal ERG revealed reduced responses in the inner 10° in both eyes. She underwent extensive investigations including whole-body computed tomography and positron emission tomography scan, magnetic resonance imaging of the brain, uveitis screening, retinal autoantibody testing, and genetic testing on the retinal dystrophy panel to rule-out other causes for her presentation, all of which were normal or negative.

Clinical outcomes of treated macular neovascularisation secondary to inherited retinal diseases: a literature review.

Many inherited retinal diseases (IRD) can be associated with, or be secondarily complicated by, macular neovascularisation (MNV), which has been variably treated with intravitreal antivascular endothelial growth factor, steroids, laser and surgery. In this article, we aim to present a consolidated literature review of management of IRD-related MNV.

RETINAL VASCULITIS SEVERITY ASSESSMENT: Intraobserver and Interobserver Reliability of a New Scheme for Grading Wide-Field Fluorescein Angiograms in Retinal Vasculitis.

PURPOSE: Wide-field fluorescein angiography is commonly used to assess retinal vasculitis (RV), which manifests as vascular leakage and occlusion. Currently, there is no standard grading scheme for RV severity. The authors propose a novel RV grading scheme and assess its reliability and reproducibility. METHODS: A grading scheme was developed to assess both leakage and occlusion in RV. Wide-field fluorescein angiography images from 50 patients with RV were graded by four graders, and one grader graded them twice. An intraclass correlation coefficient (ICC) was used to determine intraobserver-interobserver reliability. Generalized linear models were calculated to associate the scoring with visual acuity. RESULTS: Repeated grading by the same grader showed good intraobserver reliability for both leakage (ICC = 0.85, 95% CI 0.78-0.89) and occlusion (ICC = 0.82, 95% CI 0.75-0.88) scores. Interobserver reliability among four independent graders showed good agreement for both leakage (ICC = 0.66, 95% CI 0.49-0.77) and occlusion (ICC = 0.75, 95% CI 0.68-0.81) scores. An increasing leakage score was significantly associated with worse concurrent visual acuity (generalized linear models, ß = 0.090, P < 0.01) and at 1-year follow-up (generalized linear models, ß = 0.063, P < 0.01). CONCLUSION: The proposed grading scheme for RV has good to excellent intraobserver and interobserver reliability across a range of graders. The leakage score related to present and future visual acuity.

Can artificial intelligence accelerate the diagnosis of inherited retinal diseases? Protocol for a data-only retrospective cohort study (Eye2Gene).

INTRODUCTION: Inherited retinal diseases (IRD) are a leading cause of visual impairment and blindness in the working age population. Mutations in over 300 genes have been found to be associated with IRDs and identifying the affected gene in patients by molecular genetic testing is the first step towards effective care and patient management. However, genetic diagnosis is currently slow, expensive and not widely accessible. The aim of the current project is to address the evidence gap in IRD diagnosis with an AI algorithm, Eye2Gene, to accelerate and democratise the IRD diagnosis service. METHODS AND ANALYSIS: The data-only retrospective cohort study involves a target sample size of 10 000 participants, which has been derived based on the number of participants with IRD at three leading UK eye hospitals: Moorfields Eye Hospital (MEH), Oxford University Hospital (OUH) and Liverpool University Hospital (LUH), as well as a Japanese hospital, the Tokyo Medical Centre (TMC). Eye2Gene aims to predict causative genes from retinal images of patients with a diagnosis of IRD. For this purpose, 36 most common causative IRD genes have been selected to develop a training dataset for the software to have enough examples for training and validation for detection of each gene. The Eye2Gene algorithm is composed of multiple deep convolutional neural networks, which will be trained on MEH IRD datasets, and externally validated on OUH, LUH and TMC. ETHICS AND DISSEMINATION: This research was approved by the IRB and the UK Health Research Authority (Research Ethics Committee reference 22/WA/0049) 'Eye2Gene: accelerating the diagnosis of IRDs' Integrated Research Application System (IRAS) project ID: 242050. All research adhered to the tenets of the Declaration of Helsinki. Findings will be reported in an open-access format.

UK Biobank retinal imaging grading: methodology, baseline characteristics and findings for common ocular diseases.

BACKGROUND/OBJECTIVES: This study aims to describe the grading methods and baseline characteristics for UK Biobank (UKBB) participants who underwent retinal imaging in 2009-2010, and to characterise individuals with retinal features suggestive of age-related macular degeneration (AMD), glaucoma and retinopathy. METHODS: Non-mydriatic colour fundus photographs and macular optical coherence tomography (OCT) scans were manually graded by Central Administrative Research Facility certified graders and quality assured by clinicians of the Network of Ophthalmic Reading Centres UK. Captured retinal features included those associated with AMD (≥1 drusen, pigmentary changes, geographic atrophy or exudative AMD; either imaging modality), glaucoma (≥0.7 cup-disc ratio, ≥0.2 cup-disc ratio difference between eyes, other abnormal disc features; photographs only) and retinopathy (characteristic features of diabetic retinopathy with or without microaneurysms; either imaging modality). Suspected cases of these conditions were characterised with reference to diagnostic records, physical and biochemical measurements. RESULTS: Among 68,514 UKBB participants who underwent retinal imaging, the mean age was 57.3 years (standard deviation 8.2), 45.7% were men and 90.6% were of White ethnicity. A total of 64,367 participants had gradable colour fundus photographs and 68,281 had gradable OCT scans in at least one eye. Retinal features suggestive of AMD and glaucoma were identified in 15,176 and 2184 participants, of whom 125 (0.8%) and 188 (8.6%), respectively, had a recorded diagnosis. Of 264 participants identified to have retinopathy with microaneurysms, 251 (95.1%) had either diabetes or hypertension. CONCLUSIONS: This dataset represents a valuable addition to what is currently available in UKBB, providing important insights to both ocular and systemic health.

Dual Consistency Enabled Weakly and Semi-Supervised Optic Disc and Cup Segmentation With Dual Adaptive Graph Convolutional Networks.

Glaucoma is a progressive eye disease that results in permanent vision loss, and the vertical cup to disc ratio (vCDR) in colour fundus images is essential in glaucoma screening and assessment. Previous fully supervised convolution neural networks segment the optic disc (OD) and optic cup (OC) from color fundus images and then calculate the vCDR offline. However, they rely on a large set of labeled masks for training, which is expensive and time-consuming to acquire. To address this, we propose a weakly and semi-supervised graph-based network that investigates geometric associations and domain knowledge between segmentation probability maps (PM), modified signed distance function representations (mSDF), and boundary region of interest characteristics (B-ROI) in three aspects. Firstly, we propose a novel Dual Adaptive Graph Convolutional Network (DAGCN) to reason the long-range features of the PM and the mSDF w.r.t. the regional uniformity. Secondly, we propose a dual consistency regularization-based semi-supervised learning paradigm. The regional consistency between the PM and the mSDF, and the marginal consistency between the derived B-ROI from each of them boost the proposed model's performance due to the inherent geometric associations. Thirdly, we exploit the task-specific domain knowledge via the oval shapes of OD & OC, where a differentiable vCDR estimating layer is proposed. Furthermore, without additional annotations, the supervision on vCDR serves as weakly-supervisions for segmentation tasks. Experiments on six large-scale datasets demonstrate our model's superior performance on OD & OC segmentation and vCDR estimation. The implementation code has been made available.https://github.com/smallmax00/Dual_Adaptive_Graph_Reasoning.

Uncovering the burden of hidden ciliopathies in the 100 000 Genomes Project: a reverse phenotyping approach.

BACKGROUND: The 100 000 Genomes Project (100K) recruited National Health Service patients with eligible rare diseases and cancer between 2016 and 2018. PanelApp virtual gene panels were applied to whole genome sequencing data according to Human Phenotyping Ontology (HPO) terms entered by recruiting clinicians to guide focused analysis. METHODS: We developed a reverse phenotyping strategy to identify 100K participants with pathogenic variants in nine prioritised disease genes (BBS1, BBS10, ALMS1, OFD1, DYNC2H1, WDR34, NPHP1, TMEM67, CEP290), representative of the full phenotypic spectrum of multisystemic primary ciliopathies. We mapped genotype data 'backwards' onto available clinical data to assess potential matches against phenotypes. Participants with novel molecular diagnoses and key clinical features compatible with the identified disease gene were reported to recruiting clinicians. RESULTS: We identified 62 reportable molecular diagnoses with variants in these nine ciliopathy genes. Forty-four have been reported by 100K, 5 were previously unreported and 13 are new diagnoses. We identified 11 participants with unreportable, novel molecular diagnoses, who lacked key clinical features to justify reporting to recruiting clinicians. Two participants had likely pathogenic structural variants and one a deep intronic predicted splice variant. These variants would not be prioritised for review by standard 100K diagnostic pipelines. CONCLUSION: Reverse phenotyping improves the rate of successful molecular diagnosis for unsolved 100K participants with primary ciliopathies. Previous analyses likely missed these diagnoses because incomplete HPO term entry led to incorrect gene panel choice, meaning that pathogenic variants were not prioritised. Better phenotyping data are therefore essential for accurate variant interpretation and improved patient benefit.

Long-term Retinal Morphology and Functional Associations in Treated Neovascular Age-Related Macular Degeneration: Findings from the Inhibition of VEGF in Age-Related Choroidal Neovascularisation Trial.

PURPOSE: To describe the frequency of long-term morphologic features and their relationships with visual function in participants who exited the Inhibition of VEGF in Age-Related Choroidal Neovascularisation (IVAN; ISRCTN92166560) trial. DESIGN: Multicenter cohort study up to 7 years after enrollment. PARTICIPANTS: Patients enrolled in the IVAN trial, excluding participants who died or withdrew during the trial. METHODS: Multimodal fundus images, best-corrected visual acuity (BCVA), and low-luminance visual acuity (LLVA) were obtained for a subset of 199 participants who attended a research visit. Clinical sites (n = 20) also provided all visual acuity and clinical information from usual care records for 532 participants and submitted the most recent color, OCT, and other fundus images for 468 participants to a reading center. MAIN OUTCOME MEASURES: Assessed the following from the most recent images: intralesional macular atrophy (ILMA) within the footprint of the neovascular lesion; hyperreflective material (HRM); intraretinal fluid (IRF); subretinal fluid (SRF); pigment epithelial detachment (PED); and disorganized retinal outer layers (DROLs). Cross-sectional relationships between morphologic features and BCVA/LLVA were estimated. RESULTS: Intralesional macular atrophy was present in 31.8% of the study eyes at IVAN exit (mean follow-up, 1.96 years) and 89.5% at the most recent imaging visit (mean follow-up, 6.18 years). Hyperreflective material, IRF, SRF, PED, and DROLs were present in 78.8%, 47.7%, 7.6%, 94.5%, and 55% of the study eyes, respectively. In the subset with complete imaging data, in eyes without DROL, the BCVA was worst in the thinnest outer fovea tertile (thinnest minus middle and thickest tertiles, -19.7 and -19.5 letters, respectively), whereas in eyes with DROL, the BCVA was worst in the thickest (thinnest and middle tertiles minus thickest, 12.5 and 12.2, respectively). Regression models showed that the presence of ILMA and HRM was independently associated with BCVA (22 letters worse [95% confidence interval {CI}, -11.2 to -32.8; P < 0.001] and 9.8 letters worse [95% CI, -0.1 to -19.4; P = 0.047], respectively). Subretinal fluid and foveal PED were associated with better BCVA (5.9 letters [95% CI, -7.9 to 19.7; P = 0.399] and 6.4 letters [95% CI, -1.1 to 14.0; P = 0.094], respectively). The model with LLVA was similar. A sensitivity analysis involving the entire eligible cohort yielded similar estimates. CONCLUSIONS: Macular atrophy and HRM were common after 7 years of follow-up and strongly associated with visual outcomes.

Lessons of the month 2: Retinal vasculitis: a first presentation of Takayasu's arteritis.

Takayasu's arteritis is a chronic, systemic, large-vessel vasculitis affecting the aorta and its primary branches. However, coronary, renal and pulmonary arteries and small vessel involvement has been documented. We describe a rare case of Takayasu's arteritis with extensive supra-aortic arch disease, manifesting with bilateral occlusive retinal vasculitis as a first presentation. This is elicited by fundus findings of vascular sheathing and fundus fluoresceine angiography evidence of retinal vessel occlusion and peripheral capillary non-perfusion.

Accuracy of a Machine-Learning Algorithm for Detecting and Classifying Choroidal Neovascularization on Spectral-Domain Optical Coherence Tomography.

BACKGROUND: To evaluate the performance of a machine-learning (ML) algorithm to detect and classify choroidal neovascularization (CNV), secondary to age-related macular degeneration (AMD) on spectral-domain optical coherence tomography (SD-OCT) images. METHODS: Baseline fluorescein angiography (FA) and SD-OCT images from 1037 treatment-naive study eyes and 531 fellow eyes, without advanced AMD from the phase 3 HARBOR trial (NCT00891735), were used to develop, train, and cross-validate an ML pipeline combining deep-learning-based segmentation of SD-OCT B-scans and CNV classification, based on features derived from the segmentations, in a five-fold setting. FA classification of the CNV phenotypes from HARBOR was used for generating the ground truth for model development. SD-OCT scans from the phase 2 AVENUE trial (NCT02484690) were used to externally validate the ML model. RESULTS: The ML algorithm discriminated CNV absence from CNV presence, with a very high accuracy (area under the receiver operating characteristic [AUROC] = 0.99), and classified occult versus predominantly classic CNV types, per FA assessment, with a high accuracy (AUROC = 0.91) on HARBOR SD-OCT images. Minimally classic CNV was discriminated with significantly lower performance. Occult and predominantly classic CNV types could be discriminated with AUROC = 0.88 on baseline SD-OCT images of 165 study eyes, with CNV from AVENUE. CONCLUSIONS: Our ML model was able to detect CNV presence and CNV subtypes on SD-OCT images with high accuracy in patients with neovascular AMD.

Topographical Response of Retinal Neovascularization to Aflibercept or Panretinal Photocoagulation in Proliferative Diabetic Retinopathy: Post Hoc Analysis of the CLARITY Randomized Clinical Trial.

IMPORTANCE: Eyes with proliferative diabetic retinopathy have a variable response to treatment with panretinal photocoagulation (PRP) or anti-vascular endothelial growth factor agents. The location of neovascularization (NV) is associated with outcomes (eg, patients with disc NV [NVD] have poorer visual prognosis than those with NV elsewhere [NVE]). OBJECTIVE: To investigate the distribution of NV in patients with proliferative diabetic retinopathy and the topographical response of NV to treatment with aflibercept or PRP. DESIGN, SETTING, AND PARTICIPANTS: This post hoc analysis of the phase 2b randomized clinical single-masked multicenter noninferiority Clinical Efficacy and Mechanistic Evaluation of Aflibercept for Proliferative Diabetic Retinopathy (CLARITY) trial was conducted from November 1, 2019, to September 1, 2020, among 120 treatment-naive patients with proliferative diabetic retinopathy to evaluate the topography of NVD and NVE in 4 quadrants of the retina on color fundus photography at baseline and at 12 and 52 weeks after treatment. EXPOSURES: In the CLARITY trial, patients were randomized to receive intravitreal aflibercept (2 mg/0.05 mL at baseline, 4 weeks, and 8 weeks, and as needed from 12 weeks onward) or PRP (completed in initial fractionated sessions and then on an as-needed basis when reviewed every 8 weeks). MAIN OUTCOMES AND MEASURES: Main outcomes were per-retinal quadrant frequencies of NV at baseline and frequencies of patterns of regression, recurrence, and new occurrence at 12-week and 52-week unmasked follow-up. RESULTS: The study included 120 treatment-naive patients (75 men; mean [SD] age, 54.8 [14.6] years) with proliferative diabetic retinopathy (there was a 1:1 ratio of eyes to patients). At baseline, NVD with or without NVE was observed in 42 eyes (35.0%), and NVE only was found in 78 eyes (65.0%); NVE had a predilection for the nasal quadrant (64 [53.3%]). Rates of regression with treatment were higher among eyes with NVE (89 of 102 [87.3%]) compared with eyes with NVD (23 of 43 [53.5%]) by 52 weeks, with NVD being more resistant to either treatment with higher rates of persistence than NVE (20 of 39 [51.3%] vs 29 of 100 [29.0%]). Considering NVE, the regression rate in the temporal quadrant was lowest (32 of 42 [76.2%]). Eyes treated with aflibercept showed higher rates of regression of NVE compared with those treated with PRP (50 of 52 [96.2%] vs 39 of 50 [78.0%]; difference, 18.2% [95% CI, 5.5%-30.8%]; P = .01), but no difference was found for NVD (11 of 17 [64.7%] vs 12 of 26 [46.2%]; difference, 18.6% [95% CI, -11.2% to 48.3%]; P = .23). CONCLUSIONS AND RELEVANCE: This post hoc analysis found that NVD is less frequent but is associated with more resistance to currently available treatments than NVE. Aflibercept was superior to PRP for treating NVE, but neither treatment was particularly effective against NVD by 52 weeks. Future treatments are needed to better target NVD, which has poorer visual prognosis. TRIAL REGISTRATION: isrctn.org Identifier: ISRCTN32207582.

Cooperative Low-Rank Models for Removing Stripe Noise From OCTA Images.

Optical coherence tomography angiography (OCTA) is an emerging non-invasive imaging technique for imaging the microvasculature of the eye based on phase variance or amplitude decorrelation derived from repeated OCT images of the same tissue area. Stripe noise occurs during the OCTA acquisition process due to the involuntary movement of the eye. To remove the stripe noise (or 'destriping') effectively, we propose two novel image decomposition models to simultaneously destripe all the OCTA images of the same eye cooperatively: cooperative uniformity destriping (CUD) model and cooperative similarity destriping (CSD) model. Both the models consider stripe noise by low-rank constraint but in different ways: the CUD model assumes that stripe noise is identical across all the layers while the CSD model assumes that the stripe noise at different layers are different and have to be considered in the model. Compared to the CUD model, CSD is a more general solution for real OCTA images. An efficient solution (CSD+) is developed for model CSD to reduce the computational complexity. The models were extensively evaluated against state-of-the-art methods on both synthesized and real OCTA datasets. The experiments demonstrated not only the effectiveness of the CSD and CSD+ models in terms of peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) and CSD+ is twice faster than CSD, but also their beneficiary effect on the vessel segmentation of OCTA images. We expect our models will become a powerful tool for clinical applications.