Retinal vascular caliber and the development of hypertension: a meta-analysis of individual participant data

AutoMorph: Automated Retinal Vascular Morphology Quantification Via a Deep Learning Pipeline

Purpose

To externally validate a deep learning pipeline (AutoMorph) for automated analysis of retinal vascular morphology on fundus photographs. AutoMorph has been made publicly available, facilitating widespread research in ophthalmic and systemic diseases.

Methods

AutoMorph consists of four functional modules: image preprocessing, image quality grading, anatomical segmentation (including binary vessel, artery/vein, and optic disc/cup segmentation), and vascular morphology feature measurement. Image quality grading and anatomical segmentation use the most recent deep learning techniques. We employ a model ensemble strategy to achieve robust results and analyze the prediction confidence to rectify false gradable cases in image quality grading. We externally validate the performance of each module on several independent publicly available datasets.

Results

The EfficientNet-b4 architecture used in the image grading module achieves performance comparable to that of the state of the art for EyePACS-Q, with an F1-score of 0.86. The confidence analysis reduces the number of images incorrectly assessed as gradable by 76%. Binary vessel segmentation achieves an F1-score of 0.73 on AV-WIDE and 0.78 on DR HAGIS. Artery/vein scores are 0.66 on IOSTAR-AV, and disc segmentation achieves 0.94 in IDRID. Vascular morphology features measured from the AutoMorph segmentation map and expert annotation show good to excellent agreement.

Conclusions

AutoMorph modules perform well even when external validation data show domain differences from training data (e.g., with different imaging devices). This fully automated pipeline can thus allow detailed, efficient, and comprehensive analysis of retinal vascular morphology on color fundus photographs.

Translational Relevance

By making AutoMorph publicly available and open source, we hope to facilitate ophthalmic and systemic disease research, particularly in the emerging field of oculomics.

A deep-learning system for the assessment of cardiovascular disease risk via the measurement of retinal-vessel calibre

Retinal blood vessels provide information on the risk of cardiovascular disease (CVD). Here, we report the development and validation of deep-learning models for the automated measurement of retinal-vessel calibre in retinal photographs, using diverse multiethnic multicountry datasets that comprise more than 70,000 images. Retinal-vessel calibre measured by the models and by expert human graders showed high agreement, with overall intraclass correlation coefficients of between 0.82 and 0.95. The models performed comparably to or better than expert graders in associations between measurements of retinal-vessel calibre and CVD risk factors, including blood pressure, body-mass index, total cholesterol and glycated-haemoglobin levels. In retrospectively measured prospective datasets from a population-based study, baseline measurements performed by the deep-learning system were associated with incident CVD. Our findings motivate the development of clinically applicable explainable end-to-end deep-learning systems for the prediction of CVD on the basis of the features of retinal vessels in retinal photographs.

Retinal vascular abnormalities and their associations with cardiovascular and cerebrovascular diseases: a Study in rural southwestern Harbin, China

Background

Limited data is available on retinal vessel morphology in the north China. The study aimed to evaluate the prevalence of retinal vascular abnormalities (RVAs) and investigate their associations with the self-reported diagnosis of cardiovascular and cerebrovascsular diseases (CCVds) in a rural adult population of northeast China.

Methods

A population-based, cross-sectional study was conducted, using the cluster random sampling method. One eye of each participant was photographed with a non-mydriatic fundus camera. RVAs including focal and general arteriolar narrowing (FAN and GAN), arteriovenous nicking (AVN), arteriolar sheathing (AS), and retinopathy were evaluated. Data on self-reported diagnosis of cardiovascular and cerebrovascular diseases and status of smoking and alcohol drinking were obtained from questionnaires.

Results

Among the 6267 participants with an age ≥ 50 years, photographs were obtained of 99.2%, with quality sufficient to perform retinal evaluations in 82.5%. The prevalence of FAN, AVN, AS, retinopathy and GAN were 9.1, 8.9, 5.0, 6.6 and 6.2%, respectively. All the retinal lesions were associated with hypertension (all P < 0.01). After adjusting for age, gender, and left/right eyes, hypertension, hyperlipidaemia, diabetes mellitus, habits of past or current smoking and alcohol consumption, AVN was strongly associated with the self-reported diagnosis histories of coronary heart diseases(CHD) (OR, 1.44; 95% CI, 1.09, 1.89) and retinopathy was significantly associated with a self-reported diagnosis of stroke (OR, 2.05; 95% CI, 1.18, 3.57).

Conclusions

The overall prevalence of retinal microvascular abnormalities in this population was relatively higher than that reported in other regions of the world. Retinopathy is associated with the self-reported diagnosis of stroke while AVN was associated with the self-reported diagnosis of CHD, but the remaining retinal lesions were not consistently associated with CCVds. Thus, an examination of retinal microvascular characteristics may offer clues to CCVds and could be a potentially novel biomarkers for CCVds risk.

Automatic optic disc detection in colour fundus images by means of multispectral analysis and information content

The optic disc (OD) in retinal fundus images is widely used as a reference in computer-based systems for the measurement of the severity of retinal disease. A number of algorithms have been published in the past 5 years to locate and measure the OD in digital fundus images. Our proposed algorithm, automatically: (i) uses the three channels (RGB) of the digital colour image to locate the region of interest (ROI) where the OD lies, (ii) measures the Shannon information content per channel in the ROI, to decide which channel is most appropriate for searching for the OD centre using the circular Hough transform. A series of evaluations were undertaken to test our hypothesis that using the three channels gives a better performance than a single channel. Three different databases were used for evaluation purposes with a total of 2,371 colour images giving a misdetection error of 3% in the localisation of the centre of the OD. We find that the area determined by our algorithm which assumes that the OD is circular, is similar to that found by other algorithms that detected the shape of the OD. Five metrics were measured for comparison with other recent studies. Combining the two databases where expert delineation of the OD is available (1,240 images), the average results for our multispectral algorithm are: TPR = 0.879, FPR = 0.003, Accuracy = 0.994, Overlap = 80.6% and Dice index = 0.878.

Risk factors for hypertensive retinopathy in a Chinese population with hypertension: The Beijing Eye study

Hypertensive retinopathy (HRP) is a clinical feature and systemic manifestation of hypertension. There have been few reports on the risk factors for hypertensive retinopathy in China. The aim of the present study was to describe risk factors associated with HRP in a population-based sample of Chinese patients with hypertension in Beijing. The clinical data of 228 hypertensive patients was retrospectively studied, including 112 patients with retinopathy for the HRP group and 116 patients without retinopathy for the NO-HRP group. Basic clinical data and plasma clinical indicators of the two groups were compared. Logistic regression models were used to assess risk factors associated with HRP. Duration of hypertension, smoking habits, family history of hypertension, plasma level of endothelin-1 (ET-1) and systolic and diastolic blood pressure were significantly increased in the HRP group compared with the NO-HRP group (P<0.05). No significant differences were identified between the two groups for other factors. Logistic regression models indicated that hypertension duration (P<0.001, 95% CI: 0.962-0.988) and ET-1 level (P<0.001, 95% CI: 1.144-1.278) were significantly associated with HRP. The diagnostic threshold of ET-1 to diagnosis HRP was 43.5 ng/l. Of the factors studied, longer hypertension duration and elevated ET-1 level were identified to be risk factors for HRP in patients with hypertension from Beijing. Detecting the plasma level of ET-1 in patients with hypertension may be a useful diagnostic indicator for HRP.

Association between blood pressure and retinal arteriolar and venular diameters in Chinese early adolescent children, and whether the association has gender difference: a cross-sectional study

Background

To establish the independent association between blood pressure (BP) and retinal vascular caliber, especially the retinal venular caliber, in a population of 12-year-old Chinese children.

Methods

We have examined 1501 students in the 7th grade with mean age of 12.7 years. A non-mydriatic fundus camera (Canon CR-2, Tokyo, Japan) was used to capture 45⁰ fundus images of the right eyes. Retinal vascular caliber was measured using a computer-based program (IVAN). BP was measured using an automated sphygmomanometer (HEM-907, Omron, Kyoto, Japan).

Results

The mean retinal arteriolar caliber was 145.3 μm (95% confidence interval [CI], 110.6–189.6 μm) and the mean venular caliber was 212.7 μm (95% CI, 170.6–271.3 μm). After controlling for age, sex, axial length, BMI, waist, spherical equivalent, birth weight, gestational age and fellow retinal vessel caliber, children in the highest quartile of BP had significantly narrower retinal arteriolar caliber than those with lower quartiles (P for trend< 0.05). Each 10-mmHg increase in BP was associated with narrowing of the retinal arterioles by 3.00 μm (multivariable-adjusted P < 0.001), and the results were consist in three BP measurements. The association between BP measures and retinal venular caliber did not persist after adjusting for fellow arteriolar caliber. And there was no significant interaction between BP and sex, age, BMI, and birth status.

Conclusions

In a large population of adolescent Chinese children, higher BP was found to be associated with narrower retinal arterioles, but not with retinal venules. Sex and other confounding factors had no effect on the relationship of BP and retinal vessel diameter.

Retinal Microvasculature Is Associated With Long-Term Survival in the General Adult Dutch Population

Retinal vascular diameters are associated with (sub)clinical cardiovascular disease and short-term cardiovascular mortality, but their association with long-term mortality is uncertain. We studied the association of retinal vascular diameters with cause-specific mortality in the general adult Dutch population during 25 years of follow-up. From 1990 to 1993, arteriolar and venular diameters were measured semiautomatically on digitized images in 5674 persons (mean age 68.0 years, 59% women) from the population-based Rotterdam study. Follow-up for mortality was complete till March 2015. Associations between vascular diameters and mortality were examined using Cox proportional hazards models, adjusting for age, sex, cardiovascular risk factors, and the fellow vessel diameter. During 85 770 person-years (mean±SD: 15.1±6.67), 3794 (66.8%) persons died, of whom 1034 due to cardiovascular causes. We found that narrower arterioles and wider venules were associated with higher risk of mortality (adjusted hazard ratio [95% confidence interval] per SD decrease 1.04 [1.00–1.08] and increase 1.07 [1.03–1.12], respectively). For arterioles, these associations were strongest for cardiovascular mortality, whereas venules showed consistent associations for cardiovascular and noncardiovascular mortality. Importantly, these associations remained unchanged after excluding the first 10 years of follow-up as immortal person-time. We found evidence for effect modification with stronger associations in persons <70 years (venules only) and smokers ( P value for interaction<0.01). We replicated our findings in another independent cohort from the Rotterdam Study of 3106 persons with 19 880 person-years of follow-up and 144 deaths (hazard ratio for venules 1.22 [1.00–1.49]). Markers of retinal microvasculature are associated with long-term mortality in the general adult Dutch population.