Comparison of subjective and objective methods to determine the retinal arterio-venous ratio using fundus photography

Retinal Vessel Caliber Measurement Bias in Fundus Images in the Presence of the Central Light Reflex

Purpose

To investigate the agreement between a fundus camera and a scanning laser ophthalmoscope in retinal vessel caliber measurements and to identify whether the presence of the central light reflex (CLR) explains potential discrepancies.

Methods

For this cross-sectional study, we obtained fundus camera and scanning laser ophthalmoscope images from 85 eyes of 85 healthy individuals (aged 50-65 years) with different blood pressure status. We measured the central retinal artery equivalent (CRAE) and central retinal artery vein equivalent (CRVE) with the Knudtson-Parr-Hubbard algorithm and assessed the CLR using a semiautomatic grading method. We used Bland-Altman plots, 95% limits of agreement, and the two-way mixed effects intraclass correlation coefficient for consistency [ICC(3,1)] to describe interdevice agreement. We used multivariable regression to identify factors associated with differences in between-device measurements.

Results

The between-device difference in CRAE (9.5  µm; 95% confidence interval, 8.0-11.1  µm) was larger than the between-device difference in CRVE (2.9  µm; 95% confidence interval, 1.3-4.5  µm), with the fundus camera yielding higher measurements (both P < 0.001). The 95% fundus camera-scanning laser ophthalmoscope limits of agreement were -4.8 to 23.9  µm for CRAE and -12.0 to 17.8  µm for CRVE. The corresponding ICCs(3,1) were 0.89 (95% confidence interval, 0.83-0.92) and 0.91 (95% confidence interval, 0.86-0.94). The between-device CRAE difference was positively associated with the presence of a CLR (P = 0.002).

Conclusions

Fundus cameras and scanning laser ophthalmoscopes yield correlated but not interchangeable caliber measurements. The CLR induces bias in arteriolar caliber in fundus camera images, compared with scanning laser ophthalmoscope images.

Translational Relevance

Refined measurements could yield better estimates of the association between retinal vessel caliber and ophthalmic or systemic disease.

A narrative review of retinal vascular parameters and the applications (Part I): Measuring methods

The retina is often used to evaluate the vascular health status of eyes and the whole body directly and noninvasively in vivo. Retinal vascular parameters included caliber, tortuosity and fractal dimension. These variables represent the density or geometric characteristics of the vascular network apart from reflecting structural changes in the retinal vessel system. Currently, these parameters are often used as indicators of retinal disease, cardiovascular and cerebrovascular disease. Advanced digital fundus photography apparatus and computer-assisted analysis techniques combined with artificial intelligence, make the quantitative calculation of these parameters easier, objective, and labor-saving.

MSGANet-RAV: A multiscale guided attention network for artery-vein segmentation and classification from optic disc and retinal images

BACKGROUND

Retinal and optic disc images are used to assess changes in the retinal vasculature. These can be changes associated with diseases such as diabetic retinopathy and glaucoma or induced using ophthalmodynamometry to measure arterial and venous pressure. Key steps toward automating the assessment of these changes are the segmentation and classification of the veins and arteries. However, such segmentation and classification are still required to be manually labelled by experts. Such automated labelling is challenging because of the complex morphology, anatomical variations, alterations due to disease and scarcity of labelled data for algorithm development. We present a deep machine learning solution called the multiscale guided attention network for retinal artery and vein segmentation and classification (MSGANet-RAV).

METHODS

MSGANet-RAV was developed and tested on 383 colour clinical optic disc images from LEI-CENTRAL, constructed in-house and 40 colour fundus images from the AV-DRIVE public dataset. The datasets have a mean optic disc occupancy per image of 60.6% and 2.18%, respectively. MSGANet-RAV is a U-shaped encoder-decoder network, where the encoder extracts multiscale features, and the decoder includes a sequence of self-attention modules. The self-attention modules explore, guide and incorporate vessel-specific structural and contextual feature information to segment and classify central optic disc and retinal vessel pixels.

RESULTS

MSGANet-RAV achieved a pixel classification accuracy of 93.15%, sensitivity of 92.19%, and specificity of 94.13% on LEI-CENTRAL, outperforming several reference models. It similarly performed highly on AV-DRIVE with an accuracy, sensitivity and specificity of 95.48%, 93.59% and 97.27%, respectively.

CONCLUSION

The results show the efficacy of MSGANet-RAV for identifying central optic disc and retinal arteries and veins. The method can be used in automated systems designed to assess vascular changes in retinal and optic disc images quantitatively.

Relationship of Retinal Vessel Caliber with Age-Related Macular Degeneration

Purpose

Evaluate the relationship between retinal vascular caliber and age-related macular degeneration (AMD) severity or progression.

Methods

A retrospective secondary analysis of 1172 fundus photographs and clinical data from the prospective Age-Related Eye Disease Study (AREDS). Central retinal artery equivalent (CRAE), central retinal vein equivalent (CRVE), and arteriole-to-venule ratio (AVR) were measured using the Parr-Hubbard-Knudtson formula. Univariate and multivariate regressions were used to determine the association of CRAE, CRVE, and AVR with age, sex, smoking status, presence of cilioretinal artery, and AMD severity at baseline and 5 years using the 9-step AMD severity score.

Results

Only CRAE and CRVE were higher in men (P < 0.001), current smokers (P < 0.001), and the eyes with a cilioretinal artery (P=0.009 - 0.043). AMD severity was greater in older patients (P=0.001), current smokers (P=0.012), the eyes without a cilioretinal artery (P=0.001), and lower AVR (P=0.034) on multivariate regression but was not influenced by CRAE or CRVE (P=0.240 - 0.500). Choroidal neovascularization (CNV) presence was associated with older age (P=0.003) and absence of a cilioretinal artery (P=0.009), while central geographic atrophy (CGA) was associated with narrower CRAE (P=0.002) and possibly AVR (P=0.046). None of the retinal vessel parameters were predictive of AMD severity score or new onset of CNV or CGA at 5 years.

Conclusion

A lower arteriole-to-venule ratio may be associated with AMD severity, with narrower arterioles seen in the eyes with geographic atrophy, suggesting a role of the retinal vasculature in AMD pathophysiology. This trial is registered with ClinicalTrials.gov Identifier: NCT00000145.

The application of arterio-venous ratio (AVR) cut-off values in clinic to stratify cardiovascular risk in patients

INTRODUCTION

Cardiovascular risk calculators are a useful tool for identifying at-risk individuals. There are standardised methods for assessing the retinal microcirculation which alters as a consequence of cardiovascular disease (CVD). This study aimed to explore if a standardised retinal vessel assessment conducted in primary optometric care reflects current cardiovascular risk, as measured using two validated CVD risk calculators (QRISK 2; Mayo Clinic).

METHODS

A total of 120 subjects were included in the analyses. Following a routine eye examination, participants had disc-centred retinal photographs and systemic blood pressure taken. Retinal vessel parameters (central retinal artery and vein equivalent and arterio-venous ratio (AVR)) were calculated using semi-automated software. Participants were then grouped into AVR quintiles as defined by the Atherosclerosis Risk in Communities Study (ARIC). Cardiovascular risk was calculated with the validated QRISK and Mayo Clinic health calculators.

RESULTS

Systolic blood pressure was significantly greater in those with an AVR value falling in the lowest quintile compared to the highest quintile (150.65 mmHg vs. 132.21 mmHg [p = 0.001]). Similarly, CVD risk was significantly higher in those with the lowest AVR compared to the highest (QRISK: 14.28% vs. 9.87% [p = 0.05]; MAYO risk: 36.35% vs. 19.21% [p = 0.01]). Chi squared analyses showed a significant difference in the number of hypertensives in the lowest AVR quintile compared to those in the highest [p = 0.02].

CONCLUSION

Whilst the ARIC population is not directly comparable to the population used to develop the QRISK calculator, it has been shown that its application could help to identify at risk individuals using retinal vessel analyses.

Retinal Vascular Caliber Changes in Early Type 2 Diabetic Patients without Retinopathy

Purpose: To evaluate the retinal vascular caliber in early type 2 diabetic patients without retinopathy.Methods: This retrospective study reviewed the medical records of 85 patients who were diagnosed with type 2 diabetes (age, 10-48 years) and sex-matched healthy controls. Dilated fundus examinations were performed to confirm the absence of diabetic retinopathy. Fundus photographs were acquired and analyzed using software that semi-automatically measured retinal vascular caliber. Central retinal arteriolar equivalent (CRAE), central retinal venular equivalent (CRVE), and arteriovenous ratio (AVR) were determined using the revised Parr-Hubbard formula. Healthy controls were examined using the same approach.Results: There were no significant differences between patients and controls in terms of CRAE (149.92 ± 21.29 µm and 144.85 ± 17.14 µm, respectively; p = 0.137) or CRVE (195.85 ± 22.94 µm and 196.69 ± 18.43 µm, respectively; p = 0.829). However, AVR was significantly higher in the patient group (0.768 ± 0.088 vs. 0.737 ± 0.061; p = 0.031). In the patient group, AVR was significantly correlated with hemoglobin A1c (r = 0.325, p = 0.006), but not with body mass index (p = 0.478).Conclusions: An increase in peripapillary AVR may be an early feature of retinal vessels observed in patients with type 2 diabetes, which is closely related to elevated blood sugar.

Expressions of Serum lncRNAs in Diabetic Retinopathy - A Potential Diagnostic Tool

With increasing incidence of diabetes worldwide, there is an ever-expanding number of patients with chronic diabetic complications such as diabetic retinopathy (DR), one of the leading causes of blindness in the working age population. Early screening for the onset and severity of DR is essential for timely intervention. With recent advancements in genomic technologies, epigenetic alterations in DR are beginning to unravel. Long non-coding RNAs (lncRNAs), which are key epigenetic mediators, have demonstrated implications in several (DR) related processes. Based on the previous research, we have developed a serum-based, multi-panel PCR test using 9 lncRNAs (ANRIL, MALAT1, WISPER, ZFAS1, H19, HOTAIR, HULC, MEG3, and MIAT) to identify and validate whether this panel could be used as a diagnostic and prognostic tool for DR. We initially used a cell culture model (human retinal endothelial cells) and confirmed that 25 mM glucose induces upregulations of ANRIL, HOTAIR, HULC, MALAT1, and ZFAS1, and downregulation of H19 compared to 5 mM glucose controls. Then as an initial proof-of-concept, we tested vitreous humor and serum samples from a small cohort of non-diabetic (N=10) and diabetic patients with proliferative retinopathy (PDR, N=11) and measured the levels of the 9 lncRNAs. Differential expressions of lncRNAs were found in the vitreous and serum of patients and showed significant correlations. We expanded our approach and assessed the same lncRNAs using samples from a larger cohort of diabetic (n= 59; M/F:44/15) and non-diabetic patients (n= 11; M/F:4/7). Significant increased lncRNA expressions of ANRIL, H19, HOTAIR, HULC, MIAT, WISPER and ZFAS1 were observed in the serum of diabetic patients (with varying stages of DR) compared to non-diabetics. No significant correlations were demonstrated between lncRNA expressions and creatinine or glycated hemoglobin (HbA1C) levels. Using ROC and further analyses, we identified distinct lncRNA phenotype combinations, which may be used to identify patients with DR. Data from this study indicate that a panel of serum lncRNAs may be used for a potential screening test for DR. Further large-scale studies are needed to validate this notion.

Performance and Limitation of Machine Learning Algorithms for Diabetic Retinopathy Screening: Meta-analysis

Background

Diabetic retinopathy (DR), whose standard diagnosis is performed by human experts, has high prevalence and requires a more efficient screening method. Although machine learning (ML)–based automated DR diagnosis has gained attention due to recent approval of IDx-DR, performance of this tool has not been examined systematically, and the best ML technique for use in a real-world setting has not been discussed.

Objective

The aim of this study was to systematically examine the overall diagnostic accuracy of ML in diagnosing DR of different categories based on color fundus photographs and to determine the state-of-the-art ML approach.

Methods

Published studies in PubMed and EMBASE were searched from inception to June 2020. Studies were screened for relevant outcomes, publication types, and data sufficiency, and a total of 60 out of 2128 (2.82%) studies were retrieved after study selection. Extraction of data was performed by 2 authors according to PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses), and the quality assessment was performed according to the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2). Meta-analysis of diagnostic accuracy was pooled using a bivariate random effects model. The main outcomes included diagnostic accuracy, sensitivity, and specificity of ML in diagnosing DR based on color fundus photographs, as well as the performances of different major types of ML algorithms.

Results

The primary meta-analysis included 60 color fundus photograph studies (445,175 interpretations). Overall, ML demonstrated high accuracy in diagnosing DR of various categories, with a pooled area under the receiver operating characteristic (AUROC) ranging from 0.97 (95% CI 0.96-0.99) to 0.99 (95% CI 0.98-1.00). The performance of ML in detecting more-than-mild DR was robust (sensitivity 0.95; AUROC 0.97), and by subgroup analyses, we observed that robust performance of ML was not limited to benchmark data sets (sensitivity 0.92; AUROC 0.96) but could be generalized to images collected in clinical practice (sensitivity 0.97; AUROC 0.97). Neural network was the most widely used method, and the subgroup analysis revealed a pooled AUROC of 0.98 (95% CI 0.96-0.99) for studies that used neural networks to diagnose more-than-mild DR.

Conclusions

This meta-analysis demonstrated high diagnostic accuracy of ML algorithms in detecting DR on color fundus photographs, suggesting that state-of-the-art, ML-based DR screening algorithms are likely ready for clinical applications. However, a significant portion of the earlier published studies had methodology flaws, such as the lack of external validation and presence of spectrum bias. The results of these studies should be interpreted with caution.

Relationship between Retinal Microvasculature, Cardiovascular Risk and Silent Brain Infarction in Hypertensive Patients

Objective: The aims of this study are to analyze the role of artery-vein ratio AVR assessment using VesselMap 2 software (Imedos Systems) and cardiovascular risk evaluation by means of REGICOR in the prediction of silent brain infarction (SBI) in middle-age hypertensive patients from the ISSYS study. Material and Methods: A cross-sectional study with 695 patients with hypertension aged 50 to 70 years who participated in the project Investigating Silent Strokes in HYpertensives: a Magnetic Resonance Imaging Study (ISSYS), was conducted in two Primary Care Centres of Barcelona. Participants agreed to a retinography and an MRI to detect silent brain infarction (SBI). The IMEDOS software was used for the semiautomatic caliber measurement of retinal arteries and veins, and the AVR was considered abnormal when <0.66. The REGICOR score was calculated for all patients. Results: Multivariate logistic regression analysis was used to evaluate the impact of AVR and REGICOR scores on SBI. The OR (odds ratio) for a high REGICOR score and an abnormal AVR were 3.16 and 4.45, respectively. When analysing the interaction of both factors, the OR of an abnormal AVR and moderate REGICOR score was 3.27, whereas with a high REGICOR score it reached 13.07. Conclusions: The measurement of AVR in patients with hypertension and with a high REGICOR score can contribute to the detection of silent brain infarction.

Age-related impairment of neurovascular coupling responses: a dynamic vessel analysis (DVA)-based approach to measure decreased flicker light stimulus-induced retinal arteriolar dilation in healthy older adults

Aging is a major risk factor for vascular cognitive impairment and dementia (VCID). Recent studies demonstrate that cerebromicrovascular dysfunction plays a causal role in the development of age-related cognitive impairment, in part via disruption of neurovascular coupling (NVC) responses. NVC (functional hyperemia) is responsible for adjusting cerebral blood flow to the increased energetic demands of activated neurons, and in preclinical animal models of aging, pharmacological restoration of NVC is associated with improved cognitive performance. To translate these findings, there is an increasing need to develop novel and sensitive tools to assess cerebromicrovascular function and NVC to assess risk for VCID and evaluate treatment efficacy. Due to shared developmental origins, anatomical features, and physiology, assessment of retinal vessel function may serve as an important surrogate outcome measure to study neurovascular dysfunction. The present study was designed to compare NVC responses in young (< 45 years of age; n = 18) and aged (> 65 years of age; n = 11) healthy human subjects by assessing flicker light-induced changes in the diameter of retinal arterioles using a dynamic vessel analyzer (DVA)-based approach. We found that NVC responses in retinal arterioles were significantly decreased in older adults as compared with younger subjects. We propose that the DVA-based approach can be used to assess NVC, as a surrogate cerebromicrovascular outcome measure, to evaluate the effects of therapeutic interventions in older individuals.

Precision of Epithelial Defect Measurements

PURPOSE

To ensure optimal care of patients, cornea specialists measure corneal features, including epithelial defects (ED), with slit-lamp calipers. However, caliper measurements are subject to interphysician variability. We examined the extent of variability in ED measurements between cornea specialists and discuss the potential clinical impact.

METHODS

A total of 48 variably sized EDs were created in pig eyes. Three cornea specialists measured the maximum vertical and horizontal ED lengths to the nearest 10th of a millimeter using slit-lamp microscopy. An absolute difference in ED measurement between cornea specialists of 0.5 mm was chosen to be the a priori threshold for clinical significance and was evaluated by the Wilcoxon signed-rank test. Interrater reliability was assessed by intraclass correlation coefficients.

RESULTS

The average absolute difference in the vertical ED length between pairs of examiners ranged from 0.54 to 0.63 mm, and that of the horizontal ED length ranged from 0.44 to 0.46 mm. These differences in ED measurement were not significantly different from 0.5 mm (all P > 0.06). However, pairs of examiners differed in vertical ED length measurements by >0.5 mm in 44% to 52% of EDs and by >1.0 mm in 13% to 17% of EDs. Pairs of examiners differed in horizontal ED length measurements by >0.5 mm in 31% to 40% of EDs and by >1.0 mm in 10% to 15% of EDs. The intraclass correlation coefficient was 0.85 (95% confidence interval, 0.77-0.91) for vertical and 0.84 (95% confidence interval, 0.74-0.90) for horizontal ED measurements.

CONCLUSIONS

Cornea specialists showed good reliability in the measured EDs; however, depending on the threshold for clinical significance, a nontrivial percentage of cases have high interexaminer clinical variability.

Inter-method agreement in retinal blood vessels diameter analysis between Dynamic Vessel Analyzer and optical coherence tomography

PURPOSE

To analyze the inter-methods agreement in arteriovenous ratio (AVR) evaluation between spectral-domain optical coherence tomography (SD-OCT) and Dynamic Vessel Analyzer (DVA).

METHODS

Healthy volunteers underwent DVA and SD-OCT examination. AVR was measured by SD-OCT using the four external lines of the optic nerve head-centered 7-line cube and by DVA using an automated AVR estimation. The mean AVR was calculated, twice, separately by two independent readers for each tool.

RESULTS

Twenty-two eyes of 11 healthy subjects (five women and six men, mean age 35) were included. AVR analysis by DVA showed high inter-observer agreement between reader 1 and 2, and high intra-observer agreement for both reader 1 and reader 2. With regard to AVR analysis on SD-OCT, we found high inter-observer agreement between reader 1 and 2, and low intra-observer agreement for reader 2 but high intra-observer agreement for reader 1. Overall, the mean AVR measured on SD-OCT turned out to be significantly higher than mean AVR measured through DVA (reader 1, 0.9023 ± 0.06 vs 0.8036 ± 0.08; p < 0.001, and reader 2, 0.9067 ± 0.06 vs 0.8083 ± 0.05; p= 0.003).

CONCLUSIONS

No inter-method agreement in AVR could be detected in the present study due to bias in measurements (shift between DVA and SD-OCT). We found significant difference in the two noninvasive methods for AVR measurement, with a tendency for SD-OCT to overestimate retinal vascular caliber in comparison to DVA. This may be useful for achieving greater accuracy in the evaluation of retinal vessel in ocular as well as systemic diseases.