Development of a deep learning model to distinguish the cause of optic disc atrophy using retinal fundus photography

The differential diagnosis for optic atrophy can be challenging and requires expensive, time-consuming ancillary testing to determine the cause. While Leber's hereditary optic neuropathy (LHON) and optic neuritis (ON) are both clinically significant causes for optic atrophy, both relatively rare in the general population, contributing to limitations in obtaining large imaging datasets. This study therefore aims to develop a deep learning (DL) model based on small datasets that could distinguish the cause of optic disc atrophy using only fundus photography. We retrospectively reviewed fundus photographs of 120 normal eyes, 30 eyes (15 patients) with genetically-confirmed LHON, and 30 eyes (26 patients) with ON. Images were split into a training dataset and a test dataset and used for model training with ResNet-18. To visualize the critical regions in retinal photographs that are highly associated with disease prediction, Gradient-Weighted Class Activation Map (Grad-CAM) was used to generate image-level attention heat maps and to enhance the interpretability of the DL system. In the 3-class classification of normal, LHON, and ON, the area under the receiver operating characteristic curve (AUROC) was 1.0 for normal, 0.988 for LHON, and 0.990 for ON, clearly differentiating each class from the others with an overall total accuracy of 0.93. Specifically, when distinguishing between normal and disease cases, the precision, recall, and F1 scores were perfect at 1.0. Furthermore, in the differentiation of LHON from other conditions, ON from others, and between LHON and ON, we consistently observed precision, recall, and F1 scores of 0.8. The model performance was maintained until only 10% of the pixel values of the image, identified as important by Grad-CAM, were preserved and the rest were masked, followed by retraining and evaluation.

Deep learning prediction of steep and flat corneal curvature using fundus photography in post-COVID telemedicine era

Recently, fundus photography (FP) is being increasingly used. Corneal curvature is an essential factor in refractive errors and is associated with several pathological corneal conditions. As FP-based examination systems have already been widely distributed, it would be helpful for telemedicine to extract information such as corneal curvature using FP. This study aims to develop a deep learning model based on FP for corneal curvature prediction by categorizing corneas into steep, regular, and flat groups. The EfficientNetB0 architecture with transfer learning was used to learn FP patterns to predict flat, regular, and steep corneas. In validation, the model achieved a multiclass accuracy of 0.727, a Matthews correlation coefficient of 0.519, and an unweighted Cohen's κ of 0.590. The areas under the receiver operating characteristic curves for binary prediction of flat and steep corneas were 0.863 and 0.848, respectively. The optic nerve and its peripheral areas were the main focus of the model. The developed algorithm shows that FP can potentially be used as an imaging modality to estimate corneal curvature in the post-COVID-19 era, whereby patients may benefit from the detection of abnormal corneal curvatures using FP in the telemedicine setting.

Development of a generative deep learning model to improve epiretinal membrane detection in fundus photography

BACKGROUND

The epiretinal membrane (ERM) is a common retinal disorder characterized by abnormal fibrocellular tissue at the vitreomacular interface. Most patients with ERM are asymptomatic at early stages. Therefore, screening for ERM will become increasingly important. Despite the high prevalence of ERM, few deep learning studies have investigated ERM detection in the color fundus photography (CFP) domain. In this study, we built a generative model to enhance ERM detection performance in the CFP.

METHODS

This deep learning study retrospectively collected 302 ERM and 1,250 healthy CFP data points from a healthcare center. The generative model using StyleGAN2 was trained using single-center data. EfficientNetB0 with StyleGAN2-based augmentation was validated using independent internal single-center data and external datasets. We randomly assigned healthcare center data to the development (80%) and internal validation (20%) datasets. Data from two publicly accessible sources were used as external validation datasets.

RESULTS

StyleGAN2 facilitated realistic CFP synthesis with the characteristic cellophane reflex features of the ERM. The proposed method with StyleGAN2-based augmentation outperformed the typical transfer learning without a generative adversarial network. The proposed model achieved an area under the receiver operating characteristic (AUC) curve of 0.926 for internal validation. AUCs of 0.951 and 0.914 were obtained for the two external validation datasets. Compared with the deep learning model without augmentation, StyleGAN2-based augmentation improved the detection performance and contributed to the focus on the location of the ERM.

CONCLUSIONS

We proposed an ERM detection model by synthesizing realistic CFP images with the pathological features of ERM through generative deep learning. We believe that our deep learning framework will help achieve a more accurate detection of ERM in a limited data setting.

Ocular biomarkers of cognitive decline based on deep-learning retinal vessel segmentation

BACKGROUND

The current literature shows a strong relationship between retinal neuronal and vascular alterations in dementia. The purpose of the study was to use NFN+ deep learning models to analyze retinal vessel characteristics for cognitive impairment (CI) recognition.

METHODS

We included 908 participants from a community-based cohort followed for over 15 years (the prospective KaiLuan Study) who underwent brain magnetic resonance imaging (MRI) and fundus photography between 2021 and 2022. The cohort consisted of both cognitively healthy individuals (N = 417) and those with cognitive impairment (N = 491). We employed the NFN+ deep learning framework for retinal vessel segmentation and measurement. Associations between Retinal microvascular parameters (RMPs: central retinal arteriolar / venular equivalents, arteriole to venular ratio, fractal dimension) and CI were assessed by Pearson correlation. P < 0.05 was considered statistically significant. The correlation between the CI and RMPs were explored, then the correlation coefficients between CI and RMPs were analyzed. Random Forest nonlinear classification model was used to predict whether one having cognitive decline or not. The assessment criterion was the AUC value derived from the working characteristic curve.

RESULTS

The fractal dimension (FD) and global vein width were significantly correlated with the CI (P < 0.05). Age (0.193), BMI (0.154), global vein width (0.106), retinal vessel FD (0.099), and CRAE (0.098) were the variables in this model that were ranked in order of feature importance. The AUC values of the model were 0.799.

CONCLUSIONS

Establishment of a predictive model based on the extraction of vascular features from fundus images has a high recognizability and predictive power for cognitive function and can be used as a screening method for CI.

Using a dual-stream attention neural network to characterize mild cognitive impairment based on retinal images

Mild cognitive impairment (MCI) is a critical transitional stage between normal cognition and dementia, for which early detection is crucial for timely intervention. Retinal imaging has been shown as a promising potential biomarker for MCI. This study aimed to develop a dual-stream attention neural network to classify individuals with MCI based on multi-modal retinal images. Our approach incorporated a cross-modality fusion technique, a variable scale dense residual model, and a multi-classifier mechanism within the dual-stream network. The model utilized a residual module to extract image features and employed a multi-level feature aggregation method to capture complex context information. Self-attention and cross-attention modules were utilized at each convolutional layer to fuse features from optical coherence tomography (OCT) and fundus modalities, resulting in multiple output losses. The neural network was applied to classify individuals with MCI, Alzheimer's disease, and control participants with normal cognition. Through fine-tuning the pre-trained model, we classified community-dwelling participants into two groups based on cognitive impairment test scores. To identify retinal imaging biomarkers associated with accurate prediction, we used the Gradient-weighted Class Activation Mapping technique. The proposed method achieved high precision rates of 84.96% and 80.90% in classifying MCI and positive test scores for cognitive impairment, respectively. Notably, changes in the optic nerve head on fundus photographs or OCT images among patients with MCI were not used to discriminate patients from the control group. These findings demonstrate the potential of our approach in identifying individuals with MCI and emphasize the significance of retinal imaging for early detection of cognitive impairment.

The automatic detection of diabetic kidney disease from retinal vascular parameters combined with clinical variables using artificial intelligence in type-2 diabetes patients

BACKGROUND

Diabetic kidney disease (DKD) has become the largest cause of end-stage kidney disease. Early and accurate detection of DKD is beneficial for patients. The present detection depends on the measurement of albuminuria or the estimated glomerular filtration rate, which is invasive and not optimal; therefore, new detection tools are urgently needed. Meanwhile, a close relationship between diabetic retinopathy and DKD has been reported; thus, we aimed to develop a novel detection algorithm for DKD using artificial intelligence technology based on retinal vascular parameters combined with several easily available clinical parameters in patients with type-2 diabetes.

METHODS

A total of 515 consecutive patients with type-2 diabetes mellitus from Xiangyang Central Hospital were included. Patients were stratified by DKD diagnosis and split randomly into either the training set (70%, N = 360) or the testing set (30%, N = 155) (random seed = 1). Data from the training set were used to develop the machine learning algorithm (MLA), while those from the testing set were used to validate the MLA. Model performances were evaluated.

RESULTS

The MLA using the random forest classifier presented optimal performance compared with other classifiers. When validated, the accuracy, sensitivity, specificity, F1 score, and AUC for the optimal model were 84.5%(95% CI 83.3-85.7), 84.5%(82.3-86.7), 84.5%(82.7-86.3), 0.845(0.831-0.859), and 0.914(0.903-0.925), respectively.

CONCLUSIONS

A new machine learning algorithm for DKD diagnosis based on fundus images and 8 easily available clinical parameters was developed, which indicated that retinal vascular changes can assist in DKD screening and detection.

Artificial intelligence in retinal image analysis: Development, advances, and challenges

Modern advances in diagnostic technologies offer the potential for unprecedented insight into ophthalmic conditions relating to the retina. We discuss the current landscape of artificial intelligence in retina with respect to screening, diagnosis, and monitoring of retinal pathologies such as diabetic retinopathy, diabetic macular edema, central serous chorioretinopathy, and age-related macular degeneration. We review the methods used in these models and evaluate their performance in both research and clinical contexts and discuss potential future directions for investigation, use of multiple imaging modalities in artificial intelligence algorithms, and challenges in the application of artificial intelligence in retinal pathologies.

Retinal imaging technologies in cerebral malaria: a systematic review

BACKGROUND

Cerebral malaria (CM) continues to present a major health challenge, particularly in sub-Saharan Africa. CM is associated with a characteristic malarial retinopathy (MR) with diagnostic and prognostic significance. Advances in retinal imaging have allowed researchers to better characterize the changes seen in MR and to make inferences about the pathophysiology of the disease. The study aimed to explore the role of retinal imaging in diagnosis and prognostication in CM; establish insights into pathophysiology of CM from retinal imaging; establish future research directions.

METHODS

The literature was systematically reviewed using the African Index Medicus, MEDLINE, Scopus and Web of Science databases. A total of 35 full texts were included in the final analysis. The descriptive nature of the included studies and heterogeneity precluded meta-analysis.

RESULTS

Available research clearly shows retinal imaging is useful both as a clinical tool for the assessment of CM and as a scientific instrument to aid the understanding of the condition. Modalities which can be performed at the bedside, such as fundus photography and optical coherence tomography, are best positioned to take advantage of artificial intelligence-assisted image analysis, unlocking the clinical potential of retinal imaging for real-time diagnosis in low-resource environments where extensively trained clinicians may be few in number, and for guiding adjunctive therapies as they develop.

CONCLUSIONS

Further research into retinal imaging technologies in CM is justified. In particular, co-ordinated interdisciplinary work shows promise in unpicking the pathophysiology of a complex disease.

Diabetic retinopathy screenings in West Virginia: an assessment of teleophthalmology implementation

BACKGROUND

The prevalence of diabetes in the state of West Virginia (WV) is amongst the highest in the United States, making diabetic retinopathy (DR) and diabetic macular edema (DME) a major epidemiological concern within the state. Several challenges exist regarding access to eye care specialists for DR screening in this rural population. A statewide teleophthalmology program has been implemented. We analyzed real-world data acquired via these systems to explore the concordance between image findings and subsequent comprehensive eye exams and explore the impact of age on image gradeability and patient distance from the West Virginia University (WVU) Eye Institute on follow-up.

METHODS

Nonmydriatic fundus images of diabetic eyes acquired at primary care clinics throughout WV were reviewed by retina specialists at the WVU Eye Institute. Analysis included the concordance between image interpretations and dilated examination findings, hemoglobin A1c (HbA1c) levels and DR presence, image gradeability and patient age, and distance from the WVU Eye Institute and follow-up compliance.

RESULTS

From the 5,512 fundus images attempted, we found that 4,267 (77.41%) were deemed gradable.  Out of the 289 patients whose image results suggested DR, 152 patients (52.6%) followed up with comprehensive eye exams-finding 101 of these patients to truly have DR/DME and allowing us to determine a positive predictive value of 66.4%. Patients within the HbA1c range of 9.1-14.0% demonstrated significantly greater prevalence of DR/DME (p < 0.01).  We also found a statistically significant decrease in image gradeability with increased age.  When considering distance from the WVU Eye Institute, it was found that patients who resided within 25 miles demonstrated significantly greater compliance to follow-up (60% versus 43%, p < 0.01).

CONCLUSIONS

The statewide implementation of a telemedicine program intended to tackle the growing burden of DR in WV appears to successfully bring concerning patient cases to the forefront of provider attention.  Teleophthalmology addresses the unique rural challenges of WV, but there is suboptimal compliance to essential follow-up with comprehensive eye exams. Obstacles remain to be addressed if these systems are to effectively improve outcomes in DR/DME patients and diabetic patients at risk of developing these sight-threatening pathologies.

Eye-brain connections revealed by multimodal retinal and brain imaging genetics in the UK Biobank

As an anatomical extension of the brain, the retina of the eye is synaptically connected to the visual cortex, establishing physiological connections between the eye and the brain. Despite the unique opportunity retinal structures offer for assessing brain disorders, less is known about their relationship to brain structure and function. Here we present a systematic cross-organ genetic architecture analysis of eye-brain connections using retina and brain imaging endophenotypes. Novel phenotypic and genetic links were identified between retinal imaging biomarkers and brain structure and function measures derived from multimodal magnetic resonance imaging (MRI), many of which were involved in the visual pathways, including the primary visual cortex. In 65 genomic regions, retinal imaging biomarkers shared genetic influences with brain diseases and complex traits, 18 showing more genetic overlaps with brain MRI traits. Mendelian randomization suggests that retinal structures have bidirectional genetic causal links with neurological and neuropsychiatric disorders, such as Alzheimer's disease. Overall, cross-organ imaging genetics reveals a genetic basis for eye-brain connections, suggesting that the retinal images can elucidate genetic risk factors for brain disorders and disease-related changes in intracranial structure and function.

Fundus Photography Methodologies to Assess RP Patients

The development of fundus photography and imaging has improved our ability to diagnose and monitor inherited retinal degenerations. Nowadays, color fundus photography has become a staple in evaluating patients with retinitis pigmentosa (RP). Other important multimodal forms of fundus photography used today include red-free fundus photography, short-wavelength autofluorescence, and near-infrared autofluorescence. These photography methodologies provide valuable information on the natural history of disease progression, which in turn can lead to the identification of viable outcome measurements for current and future therapeutic trials. Further advances and developments in the field of fundus imaging will help in our understanding of RP and allied disorders.

Multimodal diagnostic imaging in primary vitreoretinal lymphoma

Background

Primary vitreoretinal lymphoma (PVRL) is an aggressive lymphoma that may present with protean features and represents a diagnostic challenge. Given that patients with PVRL are at high risk of CNS involvement with a high mortality and morbidity rate, prompt diagnosis is crucial to initiate treatment early in the disease course. A multimodality imaging approach including fundus photography, fundus autofluorescence (FAF), optical coherence tomography (OCT), fluorescein and indocyanine angiography, and electroretinography (ERG) can provide information to establish a diagnosis and provide objective measures for management. We review key findings seen via these imaging modalities in patients with PVRL.

Observations

Fundus photography can highlight commonly seen patterns of PVRL including vitritis, subretinal disease, retinal pigment epithelial (RPE) abnormalities, optic nerve edema, retinal detachment, and less typical retinitis-like lesions. FAF can identify characteristic patterns of hyper- and hypoautofluorescent signal abnormalities in the macula. Spectral-domain OCT will demonstrate vitreous cells, RPE nodularity, and hyperreflectivity of the outer retina. The presence of a hyper-reflective band in the subretinal space and infiltrates between the RPE and Bruch’s membrane can assist in distinguishing PVRL from choroidal lymphoma. Vertical hyperreflective columns (VHRLs) are another pertinent finding that may represent microinfiltrates of the tumor. OCT has proven to be a particularly useful modality in assessing the progress of treatment in PVRL. Fluorescein angiography can show RPE changes, which include granularity, late staining at the RPE level, and blockage. Indocyanine green angiography (ICGA) primarily shows hypocyanescence, which corresponds to PVRL lesions on fundus photography and may occur secondary to loss of RPE and choriocapillaris.

Conclusion

While PVRL remains a challenging disease to diagnose and follow, the use of a multimodality imaging approach may assist in establishing a diagnosis. Because of the anatomic spaces PVRL may affect, fundus photography, OCT, FAF, angiography, and ERG can identify key characteristics of the disease, differentiate PVRL from other diseases, and provide baseline information for targeted systemic and local therapies. Further assessment of anatomic and functional targets will aid our clinical application of multimodal imaging in the management of PVRL.

Machine learning predicting myopic regression after corneal refractive surgery using preoperative data and fundus photography

PURPOSE

Myopic regression after surgery is the most common long-term complication of refractive surgery, but it is difficult to identify myopic regression without long-term observation. This study aimed to develop machine learning models to identify high-risk patients for refractive regression based on preoperative data and fundus photography.

METHODS

This retrospective study assigned subjects to the training (n = 1606 eyes) and validation (n = 403 eyes) datasets with chronological data splitting. Machine learning models with ResNet50 (for image analysis) and XGBoost (for integration of all variables and fundus photography) were developed based on subjects who underwent corneal refractive surgery. The primary outcome was the predictive performance for the presence of myopic regression at 4 years of follow-up examination postoperatively.

RESULTS

By integrating all factors and fundus photography, the final combined machine learning model showed good performance to predict myopic regression of more than 0.5 D (area under the receiver operating characteristic curve [ROC-AUC], 0.753; 95% confidence interval [CI], 0.710-0.793). The performance of the final model was better than the single ResNet50 model only using fundus photography (ROC-AUC, 0.673; 95% CI, 0.627-0.716). The top-five most important input features were fundus photography, preoperative anterior chamber depth, planned ablation thickness, age, and preoperative central corneal thickness.

CONCLUSION

Our machine learning algorithm provides an efficient strategy to identify high-risk patients with myopic regression without additional labor, cost, and time. Surgeons might benefit from preoperative risk assessment of myopic regression, patient counseling before surgery, and surgical option decisions.

Detection of Posterior Segment Eye Disease in Rural Eye Camps in South India: A Nonrandomized Cluster Trial

PURPOSE

Rural screening camps in India historically have focused on detection of cataract and uncorrected refractive error. This study aimed to increase detection, referral, and follow-up for posterior segment diseases (PSDs) in rural eye camps using a novel technology-driven eye camp model.

DESIGN

A clustered nonrandomized trial in the catchment area of Aravind Eye Care System (AECS) Pondicherry, to compare 2 eye camp models: the traditional AECS eye camp model and the novel, technology-driven, eye camp model.

PARTICIPANTS

Patients 40 to 75 years of age who attended free camps conducted by AECS Pondicherry. Those with corneal pathologic features were excluded because this precluded an adequate view of the posterior segment to screen for PSD.

METHODS

The clinical protocols in the 2 arms were standardized and the same study team was used in both study arms. The unit of allocation to the 2 study arms was at the level of the eye camp, rather than the level of the individual study participant.

MAIN OUTCOME MEASURES

The primary study outcome was detection of suspected PSD (glaucoma, diabetic retinopathy, age-related macular degeneration, other PSDs). Secondary outcomes included: (1) the proportion of referred participants who underwent an examination at the base hospital and (2) the proportion with confirmed PSD on examination at the base hospital.

RESULTS

The study included 11 traditional and 18 novel eye camps with a total of 3048 participants (50% in each study arm). The mean age of all participants was 58.4 ± 9.1 years and 1434 participants (47%) were men. The proportion receiving a referral for PSD was significantly greater in the novel (8.3%) compared with the traditional (3.6%) eye camp (P < 0.001; risk ratio, 2.31; 95% confidence interval, 2.30-2.34). Among the 183 participants referred from the camps for PSD, 73 (39.9%) followed up for further evaluation at the base hospital.

CONCLUSIONS

In a resource-constrained setting, use of digital fundus photography in novel eye camps resulted in increased detection of and referral for PSD. Further research is needed to determine whether this intervention is cost effective and may contribute to prevention of avoidable blindness and visual impairment in South India. Further research also is needed to improve follow-up of patients referred from camps for suspicion of PSD.

A novel multiple subdivision-based algorithm for quantitative assessment of retinal vascular tortuosity

Vascular tortuosity as an indicator of retinal vascular morphological changes can be quantitatively analyzed and used as a biomarker for the early diagnosis of relevant disease such as diabetes. While various methods have been proposed to evaluate retinal vascular tortuosity, the main obstacle limiting their clinical application is the poor consistency compared with the experts' evaluation. In this research, we proposed to apply a multiple subdivision-based algorithm for the vessel segment vascular tortuosity analysis combining with a learning curve function of vessel curvature inflection point number, emphasizing the human assessment nature focusing not only global but also on local vascular features. Our algorithm achieved high correlation coefficients of 0.931 for arteries and 0.925 for veins compared with clinical grading of extracted retinal vessels. For the prognostic performance against experts' prediction in retinal fundus images from diabetic patients, the area under the receiver operating characteristic curve reached 0.968, indicating a good consistency with experts' predication in full retinal vascular network evaluation.

[Diagnostics of diseases of the optic nerve head in times of artificial intelligence and big data]

Hintergrund

Der Einsatz von künstlicher Intelligenz (KI) ist unter anderem in der automatischen Bildsegmentierung, -analyse und Klassifikation interessant und bereits für verschiedene Bereiche der Augenheilkunde beschrieben.

Fragestellung

Diese Arbeit soll einen Überblick über aktuelle Ansätze und Fortschritte bei der Anwendung von Big Data und KI bei verschiedenen Erkrankungen des Sehnervenkopfes geben.

Material und Methode

Es wurde eine PubMed-Recherche durchgeführt. Gesucht wurde nach Studien, die klinische Fragestellungen mithilfe von Big-Data-Ansätzen beantworteten oder klassische Methoden des maschinellen Lernens bei der Analyse von multimodaler Bildgebung des Sehnervenkopfes verwendeten.

Ergebnisse

Big Data kann bei Volkskrankheiten wie dem Glaukom helfen, klinische Fragestellungen zu beantworten. KI findet sowohl bei der Segmentierung von multimodaler Bildgebung des Sehnervenkopfes als auch bei der Klassifikation von Erkrankungen wie dem Glaukom oder der Stauungspapille auf diesen Bilddaten Anwendung.

Schlussfolgerung

Mithilfe von Big Data und KI können Zusammenhänge besser erkannt und die Diagnostik und Verlaufsbeurteilung von Erkrankungen des Sehnervenkopfes erleichtert oder automatisiert werden. Eine Voraussetzung für die klinische Anwendung ist in Europa die CE-Kennzeichnung als ein Medizinprodukt und in den USA die Zulassung durch die Food and Drug Administration.

Ocular Fundus Abnormalities in Acute Subarachnoid Hemorrhage: The FOTO-ICU Study

BACKGROUND

Ocular fundus abnormalities, especially intraocular hemorrhage, may represent a clinically useful prognostic marker in patients with acute subarachnoid hemorrhage (SAH).

OBJECTIVE

To evaluate associations between ocular fundus abnormalities and clinical outcomes in acute SAH.

METHODS

Prospective evaluation of acute SAH patients with ocular fundus photography at bedside. Multivariable logistic models were used to evaluate associations between fundus abnormalities and poor outcome (inpatient death, care withdrawal, or discharge Glasgow Outcome Score <4) and intensive care unit (ICU) and hospital lengths-of-stay, controlling for APACHE II score, respiratory failure at ICU admission, Hunt & Hess score, aneurysmal etiology, age, and sex.

RESULTS

Fundus abnormalities were present in 29/79 patients with acute SAH (35.4%), and 20/79 (25.3%) had intraocular hemorrhage. In univariate analyses, poor outcomes were more likely among patients with fundus abnormalities vs without (15/28 [53.6%] vs 15/51 [29.4%], P = .03); median length of ICU stay was longer in patients with intraocular hemorrhage than without (18 d [interquartile range (IQR) 12-25] vs 11 [IQR 7-17], P = .03). Logistic regression with fundus abnormality as predictor of interest showed that male sex (odds ratio [OR] 5.33 [95% CI 1.09-26.0], P = .045), higher APACHE II (OR, per 1-point increase, 1.35 [95% CI 1.08-1.78], P = .01), and aneurysmal etiology (OR 4.35 [95% CI 1.01-22.9], P = .048), but not fundus abnormalities (OR 1.56 [95% CI 0.43-5.65], P = .49) or intraocular hemorrhage (OR 1.28 [95% CI 0.26-5.59], P = .75) were associated with poor outcome.

CONCLUSION

Although ocular fundus abnormalities are associated with disease severity in SAH, they do not add value to patients' acute management beyond other risk factors already in use.

Peripheral bright streaks in tuberous sclerosis

Purpose

To describe the finding of bright hyperautofluorescent streaks in the peripheral retina in tuberous sclerosis.

Observations

A woman with a pathogenic TSC1 mutation and cutaneous manifestations of tuberous sclerosis underwent fundus examination and was found to have a cluster of thin, yellowish streaks in the inferior peripheral fundus of her left eye. The streaks were hyperautofluorescent in blue light and associated with irregular thickening of the photoreceptor-pigment epithelium complex on optical coherence tomography.

Conclusions and importance

The cluster of outer retinal abnormalities in a sector of the peripheral retina in one eye of a TSC1 patient has features in common with the more centrally located and less numerous lesions called achromatic patches. The resemblance of the streak pattern with the pattern of hypoautofluorescence in X-linked retinopathies suggests that the streaks may represent a clone of cells derived from a single somatic mutation in TSC1. The identification of this lesion type expands the scope of conditions that can be diagnosed by fundus imaging.

A 15 month experience with a primary care-based telemedicine screening program for diabetic retinopathy

Background

Using telemedicine for diabetic retinal screening is becoming popular especially amongst at-risk urban communities with poor access to care. The goal of the diabetic telemedicine project at Temple University Hospital is to improve cost-effective access to appropriate retinal care to those in need of close monitoring and/or treatment.

Methods

This will be a retrospective review of 15 months of data from March 2016 to May 2017. We will investigate how many patients were screened, how interpretable the photographs were, how often the photographs generated a diagnosis of diabetic retinopathy (DR) based on the screening photo, and how many patients followed-up for an exam in the office, if indicated.

Results

Six-hundred eighty-nine (689) digital retinal screening exams on 1377 eyes of diabetic patients were conducted in Temple’s primary care clinic. The majority of the photographs were read to have no retinopathy (755, 54.8%). Among all of the screening exams, 357 (51.8%) triggered a request for a referral to ophthalmology. Four-hundred forty-nine (449, 32.6%) of the photos were felt to be uninterpretable by the clinician. Referrals were meant to be requested for DR found in one or both eyes, inability to assess presence of retinopathy in one or both eyes, or for suspicion of a different ophthalmic diagnosis. Sixty-seven patients (9.7%) were suspected to have another ophthalmic condition based on other findings in the retinal photographs. Among the 34 patients that were successfully completed a referral visit to Temple ophthalmology, there was good concordance between the level of DR detected by their screening fundus photographs and visit diagnosis.

Conclusions

Although a little more than half of the patients did not have diabetic eye disease, about half needed a referral to ophthalmology. However, only 9.5% of the referral-warranted exams actually received an eye exam. Mere identification of referral-warranted diabetic retinopathy and other ophthalmic conditions is not enough. A successful telemedicine screening program must close the communication gap between screening and diagnosis by reviewer to provide timely follow-up by eye care specialists.

Evaluation of a novel artificial intelligence-based screening system for diabetic retinopathy in community of China: a real-world study

PURPOSE

To evaluate the performance of an AI-based diabetic retinopathy (DR) grading model in real-world community clinical setting.

METHODS

Participants with diabetes on record in the chosen community were recruited by health care staffs in a primary clinic of Zhengzhou city, China. Retinal images were prospectively collected during December 2018 and April 2019 based on intent-to-screen principle. A pre-validated AI system based on deep learning algorithm was deployed to screen DR graded according to the International Clinical Diabetic Retinopathy scale. Kappa value of DR severity, the sensitivity, specificity of detecting referable DR (RDR) and any DR were generated based on the standard of the majority manual grading decision of a retina specialist panel.

RESULTS

Of the 193 eligible participants, 173 (89.6%) were readable with at least one eye image. Mean [SD] age was 69.3 (9.0) years old. Total of 321 eyes (83.2%) were graded both by AI and the specialist panel. The κ value in eye image grading was 0.715. The sensitivity, specificity and area under curve for detection of RDR were 84.6% (95% CI: 54.6- 98.1%), 98.0% (95% CI: 94.3-99.6%) and 0.913 (95% CI: 0.797-1.000), respectively. For detection of any DR, the upper indicators were 90.0% (95% CI: 68.3-98.8), 96.6% (95% CI: 92.1-98.9) and 0.933 (95% CI: 0.933-1.000), respectively.

CONCLUSION

The AI system showed relatively good consistency with ophthalmologist diagnosis in DR grading, high specificity and acceptable sensitivity for identifying RDR and any DR.

TRANSLATIONAL RELEVANCE

It is feasible to apply AI-based DR screening in community.

PRECIS

Deployed in community real-world clinic setting, AI-based DR screening system showed high specificity and acceptable sensitivity in identifying RDR and any DR. Good DR diagnostic consistency was found between AI and manual grading. These prospective evidences were essential for regulatory approval.

Ocular Imaging for Enhancing the Understanding, Assessment, and Management of Age-Related Macular Degeneration

Age-related macular degeneration (AMD) is a progressive neuro-retinal disease and the leading cause of central vision loss among elderly individuals in the developed countries. Modern ocular imaging technologies constitute an essential component of the evaluation of these patients and have contributed extensively to our understanding of the disease. A challenge with any review of ocular imaging technologies is the rapid pace of progress and evolution of these instruments. Nonetheless, for proper and optimal use of these technologies, it is essential for the user to understand the technical principles underlying the imaging modality and their role in assessing the disease in various settings. Indeed, AMD, like many other retinal diseases, benefits from a multimodal imaging approach to optimally characterize the disease. In this chapter, we will review the various imaging technologies currently used in the assessment and management of AMD.

Retinal Microvascular Signs as Screening and Prognostic Factors for Cardiac Disease: A Systematic Review of Current Evidence

The substantial burden of heart disease promotes an interest in new ways of screening for early disease diagnosis, especially by means of noninvasive imaging. Increasing evidence for association between retinal microvascular signs and heart disease prompted us to systematically investigate the relevant current literature on the subject. We scrutinized the current literature by searching PubMed and Embase databases from 2000 to 2020 for clinical studies of the association between retinal microvascular signs and prevalent or incident heart disease in humans. Following exclusions, we extracted the relevant data from 42 publications (comprising 14 prospective, 26 cross-sectional, and 2 retrospective studies). Our search yielded significant associations between retinal vascular changes, including diameter, tortuosity, and branching, and various cardiac diseases, including acute coronary syndrome, coronary artery disease, heart failure, and conduction abnormalities. The findings of our research suggest that the retinal microvasculature can provide essential data about concurrent cardiac disease status and predict future risk of cardiac-related events.

Deep learning can generate traditional retinal fundus photographs using ultra-widefield images via generative adversarial networks

BACKGROUND AND OBJECTIVE

Retinal imaging has two major modalities, traditional fundus photography (TFP) and ultra-widefield fundus photography (UWFP). This study demonstrates the feasibility of a state-of-the-art deep learning-based domain transfer from UWFP to TFP.

METHODS

A cycle-consistent generative adversarial network (CycleGAN) was used to automatically translate the UWFP to the TFP domain. The model was based on an unpaired dataset including anonymized 451 UWFP and 745 TFP images. To apply CycleGAN to an independent dataset, we randomly divided the data into training (90%) and test (10%) datasets. After automated image registration and masking dark frames, the generator and discriminator networks were trained. Additional twelve publicly available paired TFP and UWFP images were used to calculate the intensity histograms and structural similarity (SSIM) indices.

RESULTS

We observed that all UWFP images were successfully translated into TFP-style images by CycleGAN, and the main structural information of the retina and optic nerve was retained. The model did not generate fake features in the output images. Average histograms demonstrated that the intensity distribution of the generated output images provided a good match to the ground truth images, with an average SSIM level of 0.802.

CONCLUSIONS

Our approach enables automated synthesis of TFP images directly from UWFP without a manual pre-conditioning process. The generated TFP images might be useful for clinicians in investigating posterior pole and for researchers in integrating TFP and UWFP databases. This is also likely to save scan time and will be more cost-effective for patients by avoiding additional examinations for an accurate diagnosis.

Sex judgment using color fundus parameters in elementary school students

PURPOSES

Recently, artificial intelligence has been used to determine sex using fundus photographs alone. We had earlier reported that sex can be distinguished using known factors obtained from color fundus photography (CFP) in adult eyes. However, it is not clear when the sex difference in fundus parameters begins. Therefore, we conducted this study to investigate sex determination based on fundus parameters using binominal logistic regression in elementary school students.

METHODS

This prospective observational cross-sectional study was conducted on 119 right eyes of elementary school students (aged 8 or 9 years, 59 boys and 60 girls). Through CFP, the tessellation fundus index was calculated as R/(R + G + B) using the mean value of red-green-blue intensity in the eight locations around the optic disc. Optic disc ovality ratio, papillomacular angle, retinal artery trajectory, and retinal vessel were quantified based on our earlier reports. Regularized binomial logistic regression was applied to these variables to select the decisive factors. Furthermore, its discriminative performance was evaluated using the leave-one-out cross-validation method. Sex difference in the parameters was assessed using the Mann-Whitney U test.

RESULTS

The optimal model yielded by the Ridge binomial logistic regression suggested that the ovality ratio of girls was significantly smaller, whereas their nasal green and blue intensities were significantly higher, than those of boys. Using this approach, the area under the receiver-operating characteristic curve was 63.2%.

CONCLUSIONS

Although sex can be distinguished using CFP even in elementary school students, the discrimination accuracy was relatively low. Some sex difference in the ocular fundus may begin after the age of 10 years.

Understanding seizure risk with wide field fundus photography: Implications for screening guidelines in the era of COVID-19 and telemedicine

PURPOSE

To report two cases of photosensitive seizures due to fundus photography flash.

OBSERVATIONS

Two patients with seizure history present to a retina clinic for routine follow up. While obtaining imaging, these patients experienced a seizure triggered by fundus camera flash.

CONCLUSIONS

Fundus photography is essential and ubiquitous amongst optometry and ophthalmology practices, especially in the rising era of telemedicine in the setting of the recent COVID-19 pandemic. To our knowledge, there are no other reports in the literature of seizures triggered by fundus photography flash. However, we believe this to be an under-reported phenomenon and suggest that all eye care providers screen patients for a history of seizures or epilepsy prior to fundus photography.

Quantitative Comparison of Fundus Images by 2 Ultra-Widefield Fundus Cameras

PURPOSE

To compare the relative number of retinal pixels and retinal area imaged using the Optos P200DTx (Optos PLC) and Zeiss Clarus 500 (Carl Zeiss Meditec AG) ultra-widefield (UWF) fundus cameras.

DESIGN

Single-center retrospective cross-sectional analysis.

PARTICIPANTS

Seventy-eight eyes of 46 patients.

METHODS

Eyes were imaged with Optos P200DTx, single-capture, and Zeiss Clarus 500, 2 capture montages when possible, UWF fundus cameras. Relative number of pixels encompassing all foveal-centered retinal quadrants were measured. Retinal area was measured with Zeiss Clarus 500 images that were registered to the Optos P200DTx images. Patients and technicians were asked for preferences between the machines. Imaging session times were recorded.

MAIN OUTCOME MEASURES

Relative number of retinal pixels and retina area captured by each fundus camera.

RESULTS

Optos P200DTx consistently captured more relative pixels compared with Zeiss Clarus 500: 510.4 versus 355.6 (P < 0.001) in total with a similarly statistically significant trend in all 4 quadrants (P < 0.001 for each). For area calculation, 70 of the 78 images achieved successful registration. Optos captured a larger total retinal area: 765.6 versus 566.5 mm² (P < 0.001) with a similarly statistically significant trend in all 4 quadrants. In the subset of 52 of 70 registered and montaged Zeiss Clarus 500 images, similar results were found. For peripheral pathologic features, Optos P200DTx captured unique findings in 28 images, and Zeiss Clarus 500 captured unique findings 1 image (P < 0.001). Among the 48 imaging sessions in which technicians preferred Optos P200DTx for 28 sessions (58%) and Zeiss Clarus 500 for 20 (42%; P = 0.15). Among patients who responded with a preference, 24 preferred Optos P200DTx and 20 preferred Zeiss Clarus 500 (P = 0.52). Average imaging session time was 4.6 minutes (standard deviation, 3.0 minutes) for Optos P200DTx and 5.2 minutes (standard deviation, 3.0 minutes) for Zeiss Clarus 500 (P = 0.17).

CONCLUSIONS

In the current study, the Optos P200DTx captured statistically significantly more retinal area in all 4 quadrants compared with the Zeiss Clarus 500. No statistically significant difference was found in patient or technician preference or image acquisition time between devices.

Detection of Progressive Glaucomatous Optic Nerve Damage on Fundus Photographs with Deep Learning

Purpose

To investigate whether predictions of retinal nerve fiber layer (RNFL) thickness obtained from a deep learning model applied to fundus photographs can detect progressive glaucomatous changes over time.

Design

Retrospective cohort study.

Participants

Eighty-six thousand one hundred twenty-three pairs of color fundus photographs and spectral-domain (SD) OCT images collected during 21 232 visits from 8831 eyes of 5529 patients with glaucoma or glaucoma suspects.

Methods

A deep learning convolutional neural network was trained to assess fundus photographs and to predict SD OCT global RNFL thickness measurements. The model then was tested on an independent sample of eyes that had longitudinal follow-up with both fundus photography and SD OCT. The ability to detect eyes that had statistically significant slopes of SD OCT change was assessed by receiver operating characteristic (ROC) curves. The repeatability of RNFL thickness predictions was investigated by measurements obtained from multiple photographs that had been acquired during the same day.

Main Outcome Measures

The relationship between change in predicted RNFL thickness from photographs and change in SD OCT RNFL thickness over time.

Results

The test sample consisted of 33 466 pairs of fundus photographs and SD OCT images collected during 7125 visits from 1147 eyes of 717 patients. Eyes in the test sample were followed up for an average of 5.3 ± 3.3 years, with an average of 6.2 ± 3.8 visits. A significant correlation was found between change over time in predicted and observed RNFL thickness (r = 0.76; 95% confidence interval [CI], 0.70–0.80; P < 0.001). Retinal nerve fiber layer predictions showed an ROC curve area of 0.86 (95% CI, 0.83–0.88) to discriminate progressors from nonprogressors. For detecting fast progressors (slope faster than 2 μm/year), the ROC curve area was 0.96 (95% CI, 0.94–0.98), with a sensitivity of 97% for 80% specificity and 85% for 90% specificity. For photographs obtained at the same visit, the intraclass correlation coefficient was 0.946 (95% CI, 0.940–0.952), with a coefficient of variation of 3.2% (95% CI, 3.1%–3.3%).

Conclusions

A deep learning model was able to obtain objective and quantitative estimates of RNFL thickness that correlated well with SD OCT measurements and potentially could be used to monitor for glaucomatous changes over time.

Effect of color information on the diagnostic performance of glaucoma in deep learning using few fundus images

PURPOSE

The purpose of this study was to evaluate the accuracy of the convolutional neural network (CNN) model in glaucoma identification with three primary colors (red, green, blue; RGB) and split color channels using fundus photographs with a small sample size.

METHODS

The dataset was prepared using color fundus photographs captured with a fundus camera (VX-10i, Kowa Co., Ltd., Tokyo, Japan). The training dataset consisted of 200 images, and the validation dataset contained 60 images. In the preprocessing stage, the color channels for the fundus images were separated into red (red model), green (green model), and blue (blue model) using OpenCV on Windows. All images were resized to squares with a size of 512 × 512 pixels for preprocessing before input into the model, and the model was fine-tuned with VGG16.

RESULTS

The diagnostic performance was significantly higher in the green model [area under the curve (AUC) 0.946; 95% confidence interval (CI) 0.851-0.982] than in the RGB model (AUC 0.800; 95% CI 0.658-0.893; P = 0.006), red model (AUC 0.746; 95% CI 0.601-0.851; P = 0.002), and blue model (AUC 0.558; 95% CI 0.405-0.700; P < 0.001).

CONCLUSION

The present study showed that the green digital filter is useful for structuring CNN models for automatic discrimination of glaucoma using fundus photographs with a small sample size. The present findings suggest that preprocessing, when creating the CNN model, is an important step for the identification of a large number of retinal diseases using color fundus photographs.

CycleGAN-based deep learning technique for artifact reduction in fundus photography

PURPOSE

A low quality of fundus photograph with artifacts may lead to false diagnosis. Recently, a cycle-consistent generative adversarial network (CycleGAN) has been introduced as a tool to generate images without matching paired images. Therefore, herein, we present a deep learning technique that removes the artifacts automatically in a fundus photograph using a CycleGAN model.

METHODS

This study included a total of 2206 anonymized retinal images including 1146 with artifacts and 1060 without artifacts. In this experiment, we applied the CycleGAN model to color fundus photographs with a pixel resolution of 256 × 256 × 3. To apply the CycleGAN to an independent dataset, we randomly divided the data into training (90%) and test (10%) datasets. Additionally, we adopted the automated quality evaluation (AQE) to assess the retinal image quality.

RESULTS

From the results, we observed that the artifacts such as overall haze, edge haze, lashes, arcs, and uneven illumination were successfully reduced by the CycleGAN in the generated images, and the main information of the retina was essentially retained. Further, we observed that most of the generated images exhibited improved AQE grade values when compared with the original images with artifacts.

CONCLUSION

Thus, we could conclude that the CycleGAN technique can effectively reduce the artifacts and improve the quality of fundus photographs, and it may be beneficial for clinicians in analyzing the low-quality fundus photographs. Future studies should improve the quality and resolution of the generated image to provide a more detailed fundus photography.

Ocular biomarkers for cognitive impairment in nonagenarians; a prospective cross-sectional study

Background

Ocular imaging receives much attention as a source of potential biomarkers for dementia. In the present study, we analyze these ocular biomarkers in cognitively impaired and healthy participants in a population aged over 90 years (= nonagenarian), and elucidate the effects of age on these biomarkers.

Methods

For this prospective cross-sectional study, we included individuals from the EMIF-AD 90+ study, consisting of a cognitively healthy (N = 67) and cognitively impaired group (N = 33), and the EMIF-AD PreclinAD study, consisting of cognitively healthy controls aged ≥60 (N = 198). Participants underwent Optical Coherence Tomography (OCT) and fundus photography of both eyes. OCT was used to asses total and individual inner retinal layer thickness in the macular region (Early Treatment Diabetic Retinopathy Study circles) as well as peripapillary retinal nerve fiber layer thickness, fundus images were analyzed with Singapore I Vessel Assessment to obtain 7 retinal vascular parameters. Values for both eyes were averaged. Differences in ocular biomarkers between the 2 nonagenarian groups were analyzed using linear regression, differences between the individual nonagenarian groups and controls were analyzed using generalized estimating equations.

Results

Ocular biomarkers did not differ between the healthy and cognitively impaired nonagenarian groups. 19 out of 22 ocular biomarkers assessed in this study differed between either nonagenarian group and the younger controls.

Conclusion

The ocular biomarkers assessed in this study were not associated with cognitive impairment in nonagenarians, making their use as a screening tool for dementing disorders in this group limited. However, ocular biomarkers were significantly associated with chronological age, which were very similar to those ascribed to occur in Alzheimer’s Disease.

Tumor Environment of Retinoblastoma, Intraocular Cancer

Retinoblastoma, an intraocular cancer primarily affecting children, interacts with surrounding intraocular and extraocular structures in the development and progression. Subretinal and vitreous seeds are characteristic features of retinoblastoma, which result from the interaction between the tumor and its environment at the levels of tissue and microenvironment. The retina and vitreous affect the disease course and responses to treatment options. Also, neighboring cells in the retina and physicochemical properties of the tumor microenvironment are related to the biological activities of retinoblastoma tumors. Researches focusing on the tumor environment of retinoblastoma will lead to the development of more effective treatment options, which can revolutionize the prognosis of patients with retinoblastoma.

Wide-field imaging of sickle retinopathy

Background

Wide-field imaging is a newer retinal imaging technology, capturing up to 200 degrees of the retina in a single photograph. Individuals with sickle cell retinopathy commonly exhibit peripheral retinal ischemia. Patients with proliferative sickle cell retinopathy develop pathologic retinal neovascularization of the peripheral retina which may progress into sight-threatening sequelae of vitreous hemorrhage and/or retinal detachment. The purpose of this review is to provide an overview of current and future applications of wide-field retinal imaging for sickle cell retinopathy, and recommend indications for best use.

Main body

There are several advantages to wide-field imaging in the clinical management of sickle cell disease patients. Retrospective and prospective studies support the success of wide-field imaging in detecting more sickle cell induced retinal microvascular abnormalities than traditional non-wide-field imaging. Clinicians can easily capture a greater extent of the retinal periphery in a patient's clinical baseline imaging to follow the changes at an earlier point and determine the rate of progression over time. Wide-field imaging minimizes patient and photographer burden, necessitating less photos and technical skill to capture the peripheral retina. Minimizing the number of necessary images can be especially helpful for pediatric patients with sickle cell retinopathy. Wide-field imaging has already been successful in identifying new biomarkers and risk factors for the development of proliferative sickle cell retinopathy. While these advantages should be considered, clinicians need to perform a careful risk-benefit analysis before ordering this test. Although wide-field fluorescein angiography successfully detects additional pathologic abnormalities compared to traditional imaging, a recent research study suggests that peripheral changes differentially detected by wide-field imaging may not change clinical management for most sickle cell patients.

Conclusions

While wide-field imaging may not carry a clinically significant direct benefit to all patients, it shows future promise in expanding our knowledge of sickle cell retinopathy. Clinicians may monitor peripheral retinal pathology such as retinal ischemia and retinal neovascularization over progressive time points, and use sequential wide-field retinal images to monitor response to treatment. Future applications for wide-field imaging may include providing data to facilitate machine learning, and potential use in tele-ophthalmology screening for proliferative sickle retinopathy.

Comparison of 1-field, 2-fields, and 3-fields fundus photography for detection and grading of diabetic retinopathy

AIM

To evaluate the sensitivity and specificity of 1-, 2-, and 3-fields, nonmydriatic (NM), 45° color photography compared with mydriatic ophthalmoscopy for detection of diabetic retinopathy (DR).

METHODS

Masked, comparative case series was performed utilizing a group of 128 diabetic patients (256 eyes) with various stages of DR who underwent both 3-fields NM color photography and ophthalmologic examination. In a blinded manner, the same optometrist who read the original 3-fields images for a patient read the 1- and 2-fields photographs on separate dates later.

RESULTS

The sensitivity and specificity of digital retinal photography compared with dilated ophthalmoscopy were, respectively: 88% and 76% for 1-field; 94% and 69% for 2-fields; and 100% and 79% for 3-fields. The proportion of agreement between fundus photography reading and exam DR diagnosis were 58% for 1-field, 58% for 2-fields, and 77% for 3-fields. Kappa and Cramer's V statistics for 1-, 2-, and 3-fields were 0.55 and 0.60, 0.52 and 0.57, and 0.72 and 0.74, respectively. Three-fields measurement of DR was most similar to the dilated ophthalmological exam overall and across all DR severity levels.

CONCLUSIONS

Compared to 1- and 2-fields fundus photography, 3-fields is superior for detecting vision-threatening DR. One- and 2-fields have reasonable sensitivity for DR screening.

Smartphone-based fundus photography for screening of plus-disease retinopathy of prematurity

BACKGROUND

Inadequate screening of treatment-warranted retinopathy of prematurity (ROP) can lead to devastating visual outcomes. Especially in resource-poor communities, the use of an affordable, portable, and easy to use smartphone-based non-contact fundus photography device may prove useful for screening for high-risk ROP. This study evaluates the feasibility of screening for high-risk ROP using a novel smartphone-based fundus photography device, RetinaScope.

METHODS

Retinal images were obtained using RetinaScope on a cohort of prematurely born infants during routine examinations for ROP. Images were reviewed by two masked graders who determined the image quality, the presence or absence of plus disease, and whether there was retinopathy that met predefined criteria for referral. The agreement between image-based assessments was compared to the gold standard indirect ophthalmoscopic assessment.

RESULTS

Fifty-four eyes of 27 infants were included. A wide-field fundus photograph was obtained using RetinaScope. Image quality was acceptable or excellent in 98% and 95% of cases. There was substantial agreement between the gold standard and photographic assessment of presence or absence of plus disease (Cohen's κ = 0.85). Intergrader agreement on the presence of any retinopathy in photographs was also high (κ = 0.92).

CONCLUSIONS

RetinaScope can capture digital retinal photographs of prematurely born infants with good image quality for grading of plus disease.

Bimodal learning via trilogy of skip-connection deep networks for diabetic retinopathy risk progression identification

BACKGROUND

Diabetic Retinopathy (DR) is considered a pathology of retinal vascular complications, which stays in the top causes of vision impairment and blindness. Therefore, precisely inspecting its progression enables the ophthalmologists to set up appropriate next-visit schedule and cost-effective treatment plans. In the literature, existing work only makes use of numerical attributes in Electronic Medical Records (EMR) for acquiring such kind of DR-oriented knowledge through conventional machine learning techniques, which require an exhaustive job of engineering most impactful risk factors.

OBJECTIVE

In this paper, an approach of deep bimodal learning is introduced to leverage the performance of DR risk progression identification.

METHODS

In particular, we further involve valuable clinical information of fundus photography in addition to the aforementioned systemic attributes. Accordingly, a Trilogy of Skip-connection Deep Networks, namely Tri-SDN, is proposed to exhaustively exploit underlying relationships between the baseline and follow-up information of the fundus images and EMR-based attributes. Besides that, we adopt Skip-Connection Blocks as basis components of the Tri-SDN for making the end-to-end flow of signals more efficient during feedforward and backpropagation processes.

RESULTS

Through a 10-fold cross validation strategy on a private dataset of 96 diabetic mellitus patients, the proposed method attains superior performance over the conventional EMR-modality learning approach in terms of Accuracy (90.6%), Sensitivity (96.5%), Precision (88.7%), Specificity (82.1%), and Area Under Receiver Operating Characteristics (88.8%).

CONCLUSIONS

The experimental results show that the proposed Tri-SDN can combine features of different modalities (i.e., fundus images and EMR-based numerical risk factors) smoothly and effectively during training and testing processes, respectively. As a consequence, with impressive performance of DR risk progression recognition, the proposed approach is able to help the ophthalmologists properly decide follow-up schedule and subsequent treatment plans.

[Performance of ultra-wide field retinophotography for screening of diabetic retinopathy]

INTRODUCTION

Early diagnosis of diabetic retinopathy is a public health issue. Fundus retinophotography (FRP) is widely used for the detection of the disease. Recently, ultra-wide field retinophotography (WFRP) device allows imaging of approximately 80% of the retinal surface in a single image. The goal of the present study was to evaluate the efficacy of WFRP in the diagnosis and gradation of diabetic retinopathy compared to the gold standard of FRP.

METHODS

The non-mydriatic Optos P200Tx was used for WFRP imaging. FRP in the 9 positions of gaze was then acquired with the Topcon TRC-NW6S after pupillary dilation. The processing time for each imaging modality was recorded.

RESULTS

One hundred and sixteen eyes of 58 patients were included in this study. Fourteen eyes were excluded from the analysis due to insufficient imaging quality. WFRP sensitivity was 96% and specificity was 92%. Only 6 eyes received a higher severity grade of diabetic retinopathy by WFRP compared to FRP. In these cases, when the WFRP was analyzed in the same field as the FRP, the severity grade was similar for 5 of the 6 eyes. The mean time of acquisition was significantly lower for WFRP compared to FRP.

CONCLUSION

WFRP is fast and effective in screening for diabetic retinopathy. The severity grade of the disease was similar to the gold standard of FRP in most cases. WFRP could thus be used in mass screening for diabetic retinopathy.

Objective ocular torsion outcomes after unilateral horizontal rectus surgery in infantile esotropia

PURPOSE

To analyse objective ocular torsion among patients with infantile esotropia and to determine the effects of unilateral horizontal rectus surgery.

METHODS

Sixty-eight patients (136 eyes) (range 4 to 16 years) who underwent unilateral horizontal rectus surgery for infantile esotropia participated in this retrospective single-centre study. Objective ocular torsion using fundus photography was assessed before surgery and 1 year later. We defined three groups of patients based on preoperative qualitative objective ocular torsion: physiological extorsion and pathological extorsion and intorsion. For each group, the disc-foveal angle was measured and analysed both before and after surgery. We looked for possible correlations between amount of esodeviation and disc-foveal angle size.

RESULTS

Preoperatively, 28 (41%) patients had + 6.73 (± 2.66) degrees of physiological extorsion. Thirty-one (46%) patients had + 12.94 (± 3.67) degrees of pathological extorsion. Nine (13%) patients had - 1.99 (± 2.52) degrees of intorsion. After surgery, the number of subjects with physiological extorsion increased to 45 (66%). The number of patients with pathological extorsion decreased to 17 (25%) and the mean disc-foveal angle was significantly reduced by 1.80°. Six (9%) patients presented intorsion and the mean disc-foveal angle was significantly reduced by 2.28°. For the pathological extorsion group, the size of the disc-foveal angle before surgery was positively correlated to its reduction after surgery. Disc-foveal angle variation and distance esodeviation variation after surgery were positively correlated.

CONCLUSIONS

These results highlight that pathological objective ocular torsion can be frequently found in infantile esotropia and is decreased after unilateral recession-plication surgery.

Comparison of conventional color fundus photography and multicolor imaging in choroidal or retinal lesions

PURPOSE

Our purpose was to compare the characteristics of the retinal and choroidal lesions including choroidal nevus, choroidal melanoma and congenital hypertrophy of the retina pigment epithelium using conventional color fundus photography (CFP) and multicolor imaging (MCI).

METHODS

The paired images of patients with retinal or choroidal lesions were assessed for the visibility of lesion's border, halo and drusen using a grading scale (0-2). The area of the lesion was measured on both imaging modalities. The same grading was also done on the individual color channels of MCI for a further evaluation.

RESULTS

Thirty-three eyes of 33 patients were included. There were no significant differences in the mean border, drusen and halo visibility scores between the two imaging modalities (p = 0.12, p = 0.70, p = 0.35). However, the mean area of the lesion was significantly smaller on MCI than that on CFP (14.9±3.3 versus 18.7±3.4 mm², p = 0.01).

CONCLUSION

The appearance of choroidal and/ or retinal lesions on MCI may be different than that on CFP. Though MCI can provide similar information with CFP for the features of retinal and/ or choroidal lesions including border, halo and drusen; the infrared light reflection on MCI underestimates the extent of the choroidal lesion by 33%.

Fundus Photography vs. Ophthalmoscopy Outcomes in the Emergency Department (FOTO-ED) Phase III: Web-based, In-service Training of Emergency Providers

We evaluated a web-based training aimed at improving the review of fundus photography by emergency providers. 587 patients were included, 12.6% with relevant abnormalities. Emergency providers spent 31 minutes (median) training and evaluated 359 patients. Median post-test score improvement was 6 percentage points (IQR: 2-14; p = 0.06). Pre- vs. post-training, the emergency providers reviewed 45% vs. 43% of photographs; correctly identified abnormals in 67% vs. 57% of cases; and correctly identified normals in 80% vs. 84%. The Fundus photography vs. Ophthalmoscopy Trial Outcomes in the Emergency Department studies have demonstrated that emergency providers perform substantially better with fundus photography than direct ophthalmoscopy, but our web-based, in-service training did not result in further improvements at our institution.

Analysis of Fundus Photography and Fluorescein Angiography in Nonarteritic Anterior Ischemic Optic Neuropathy and Optic Neuritis

Purpose

We evaluated fundus and fluorescein angiography (FAG) findings and characteristics that can help distinguish nonarteritic anterior ischemic optic neuropathy (NAION) from optic neuritis (ON).

Methods

Twenty-three NAION patients and 17 ON with disc swelling patients were enrolled in this study. We performed fundus photography and FAG. The disc-swelling pattern, hyperemia grade, presence of splinter hemorrhages, cotton-wool spots, artery/vein ratio and degree of focal telangiectasia were investigated. The FAG findings for each patient were compared with respect to the following features: the pattern of disc leakage in the early phase, arteriovenous (artery/vein) transit time (second), and the presence and pattern of the filling delay.

Results

Cotton-wool spots, focal telangiectasia, and venous congestion were more common in the affected eyes of NAION patients. Upon FAG, 76.5% of the patients in the ON group exhibited normal choroidal circulation. However, 56.5% of patients in the NAION group demonstrated abnormal filling defects, such as peripapillary, generalized, or watershed zone filling delays.

Conclusions

Fundus findings, including cotton-wool spots, focal telangiectasia, and venous congestion in the affected eye, may be clues that can be used to diagnose NAION. In addition, choroidal insufficiencies on FAG could be also helpful in differentiating NAION from ON.

The effects of fundus photography on the multifocal electroretinogram

PURPOSE

To determine the effect of flash fundus photography (FFP) on the multifocal electroretinogram (mfERG).

METHODS

Ten subjects underwent mfERG testing on three separate dates. Subjects received either mfERG without FFP, mfERG at 5 and 15 min after FFP, or mfERG at 30 and 45 min after FFP on each date. The FFP groups received 10 fundus photographs followed by mfERG testing, first of the right eye then of the left eye 10 min later. Data were averaged and analyzed in six concentric rings at each time point. Average amplitude and implicit times of the N1, P1, and N2 peaks for each concentric ring at each time point after FFP were compared to baseline.

RESULTS

Flash fundus photography did not lead to a significant change of amplitude or implicit times of N1, P1, or N2 at 5 min after light exposure.

CONCLUSIONS

These findings suggest that it is acceptable to perform mfERG testing without delay after performance of FFP.

Clinically detectable drusen domains in fibulin-5-associated age-related macular degeneration (AMD) : Drusen subdomains in fibulin-5 AMD

To evaluate whether drusen of subjects with fibulin-5 mutation-associated age-related macular degeneration (AMD) have clinically demonstrable drusen domains as evidenced by differences between color and fluorescein angiographic profiles. Of seven patients we identified with AMD due to mutations in the fibulin-5 gene (Fib-5 AMD), five had color fundus photography and fluorescein angiography (FA). One had bilateral choroidal neovascularization and no drusen. For each eye, the green channel (GC) of the digital RGB (Red-Green-Blue) color image and hyperfluorescent domain (HD) intensity of the FA image were registered and drusen were manually segmented and measured. Totally 75 small (≤62 μm), 110 intermediate (63-125 μm), and 30 large (>125 μm) drusen were measured in four patients within the 6 × 6 mm central macular areas. All four subjects demonstrated central or paracentral HDs within each drusen perimeter. HDs were found in association with each druse, with a HD/GC ratio of 0.82, 0.76, and 0.72 respectively for small, intermediate, and large drusen (Student T Test: P < 0.01, P < 0.01, P < 0.01). A statistical difference was found for the HD/GC ratios between small- and intermediate-sized drusen and small- and large-sized drusen but not between intermediate-sized and large-sized drusen (P = 0.001, P < 0.001, P > 0.05, respectively). AMD patients with mutations in fibulin-5 share drusen phenotypic structure and have HD/GC ratios that are similar to individuals with cuticular or basal laminar drusen. Drusen substructure may reflect similarities in drusen stage and/or genesis and appear to vary among AMD genotypes.

Hypertensive emergency presenting as blurry vision in a patient with hypertensive chorioretinopathy

A 42-year-old man presented with 3 weeks of blurry vision in the right eye. His exam was significant for decreased vision in the right eye, diffuse retinopathy in both eyes, and serous retinal detachment in the right eye. The patient was found to be hypertensive with blood pressure of 256/160 mmHg. He was diagnosed with hypertensive emergency with end-organ damage due to features of hypertensive chorioretinopathy. He was admitted with abnormal urinalysis, elevated troponin, and abnormal EKG. After appropriate control of his blood pressure, his vision and his labs normalized. Hypertensive emergencies can be manifested first in the eyes. When the choroid is associated, the hypertensive event is often more acute and associated with increased morbidity. It is imperative to obtain a fundus exam in any patient with elevated blood pressure and concomitant vision complaints.

Update on retinal vessel structure measurement with spectral-domain optical coherence tomography

This study was conducted to demonstrate a new scan method for retinal vessel structure measurement in a specific region of fundus (zone B) using spectral-domain optical coherence tomography (SD-OCT), and to assess its reliability. One temporal superior retinal vessel pair passing through a concentric ring (zone B), which was defined between half and one disc distance from the optic disc border, was chosen for the measurement using a volume scan in SD-OCT. On the SD-OCT image, retinal arteriolar outer diameter (RAOD), retinal arteriolar lumen diameter (RALD), retinal venular outer diameter (RVOD) and retinal venular lumen diameter (RVLD) were measured. Retinal vessel diameters on color fundus photographs were also analyzed. Fifty-five healthy individuals were recruited to evaluate intraobserver and interobserver reproducibility between the two examiners. The intraobserver intraclass correlation coefficient (ICC) ranged from 0.972 to 0.981, and the interobserver ICC ranged from 0.968 to 0.980. In the Bland-Altman plot, the 95% limits of interobserver agreement for the RAOD, RALD, RVOD and RVLD were -5.60 to 4.84μm, -5.78 to 5.05μm, -7.52 to 5.62μm and -7.10 to 5.63μm, respectively. The retinal arteriolar and venular lumen diameters on the SD-OCT image were close to the mean arteriolar and venular diameters obtained from the color fundus photographs. Volume scan method produced better images of retinal vessels showing the fine structures of the vessel wall, and provided reliable retinal vessel structure measurement in zone B with good repeatability and reproducibility.

Using ImageJ to evaluate optic disc pallor in traumatic optic neuropathy

Purpose

To evaluate optic disc pallor using ImageJ in traumatic optic neuropathy (TON).

Methods

This study examined unilateral TON patients. The optic disc was divided into 4 quadrants (temporal, superior, nasal, and inferior), consistent with the quadrants on optical coherence tomography (OCT) retinal nerve fiber layer (RNFL) thickness maps. Optic disc photography was performed and disc pallor was quantified using gray scale photographic images imported into ImageJ software. The correlation between optic disc pallor and RNFL thickness was examined in each quadrant.

Results

A total of 35 patients (31 male, 4 female) were enrolled in the study. The mean participant age was 34.8 ± 15.0 years (range, 5 to 63 years). Overall RNFL thickness decreased in 6 patients, with thinning most often occurring in the inferior quadrant (28 of 35 eyes). There was a significant correlation between optic disc pallor and RNFL thickness (superior, rho = -0.358, p = 0.04; inferior, rho = -0.345, p = 0.04; nasal, rho = -0.417, p = 0.01; temporal, rho = -0.390, p = 0.02). The highest level of correspondence between disc pallor and RNFL thickness values outside of the normative 95th percentiles was 39.3% and occurred in the inferior quadrant.

Conclusions

Optic disc pallor in TON was quantified with ImageJ and was significantly correlated with RNFL thickness abnormalities. Thus, ImageJ evaluations of disc pallor may be useful for evaluating RNFL thinning, as verified by OCT RNFL analyses.

Non-mydriatic ocular fundus photography and telemedicine: past, present, and future

Non-mydriatic ocular fundus photography is a promising alternative to direct ophthalmoscopy, particularly when combined with telemedicine. This review discusses these technologies from a longitudinal perspective: past, present, and future. The focus is directed to the role that non-mydriatic fundus photography and telemedicine have played in medical research and patient care, with emphasis on the major advances to date. Also discussed are the challenges to their widespread application and their substantial promise for revitalizing the importance of the ocular fundus examination in patient care, providing improved access to ophthalmic consultative services, and facilitating clinical and epidemiologic research.