A generalizable deep learning regression model for automated glaucoma screening from fundus images.

A plethora of classification models for the detection of glaucoma from fundus images have been proposed in recent years. Often trained with data from a single glaucoma clinic, they report impressive performance on internal test sets, but tend to struggle in generalizing to external sets. This performance drop can be attributed to data shifts in glaucoma prevalence, fundus camera, and the definition of glaucoma ground truth. In this study, we confirm that a previously described regression network for glaucoma referral (G-RISK) obtains excellent results in a variety of challenging settings. Thirteen different data sources of labeled fundus images were utilized. The data sources include two large population cohorts (Australian Blue Mountains Eye Study, BMES and German Gutenberg Health Study, GHS) and 11 publicly available datasets (AIROGS, ORIGA, REFUGE1, LAG, ODIR, REFUGE2, GAMMA, RIM-ONEr3, RIM-ONE DL, ACRIMA, PAPILA). To minimize data shifts in input data, a standardized image processing strategy was developed to obtain 30° disc-centered images from the original data. A total of 149,455 images were included for model testing. Area under the receiver operating characteristic curve (AUC) for BMES and GHS population cohorts were at 0.976 [95% CI: 0.967-0.986] and 0.984 [95% CI: 0.980-0.991] on participant level, respectively. At a fixed specificity of 95%, sensitivities were at 87.3% and 90.3%, respectively, surpassing the minimum criteria of 85% sensitivity recommended by Prevent Blindness America. AUC values on the eleven publicly available data sets ranged from 0.854 to 0.988. These results confirm the excellent generalizability of a glaucoma risk regression model trained with homogeneous data from a single tertiary referral center. Further validation using prospective cohort studies is warranted.

Artificial Intelligence Software for Diabetic Eye Screening: Diagnostic Performance and Impact of Stratification.

AIM: To evaluate the MONA.health artificial intelligence screening software for detecting referable diabetic retinopathy (DR) and diabetic macular edema (DME), including subgroup analysis. METHODS: The algorithm's threshold value was fixed at the 90% sensitivity operating point on the receiver operating curve to perform the disease classification. Diagnostic performance was appraised on a private test set and publicly available datasets. Stratification analysis was executed on the private test set considering age, ethnicity, sex, insulin dependency, year of examination, camera type, image quality, and dilatation status. RESULTS: The software displayed an area under the curve (AUC) of 97.28% for DR and 98.08% for DME on the private test set. The specificity and sensitivity for combined DR and DME predictions were 94.24 and 90.91%, respectively. The AUC ranged from 96.91 to 97.99% on the publicly available datasets for DR. AUC values were above 95% in all subgroups, with lower predictive values found for individuals above the age of 65 (82.51% sensitivity) and Caucasians (84.03% sensitivity). CONCLUSION: We report good overall performance of the MONA.health screening software for DR and DME. The software performance remains stable with no significant deterioration of the deep learning models in any studied strata.

Pointwise Visual Field Estimation From Optical Coherence Tomography in Glaucoma Using Deep Learning.

Purpose: Standard automated perimetry is the gold standard to monitor visual field (VF) loss in glaucoma management, but it is prone to intrasubject variability. We trained and validated a customized deep learning (DL) regression model with Xception backbone that estimates pointwise and overall VF sensitivity from unsegmented optical coherence tomography (OCT) scans. Methods: DL regression models have been trained with four imaging modalities (circumpapillary OCT at 3.5 mm, 4.1 mm, and 4.7 mm diameter) and scanning laser ophthalmoscopy en face images to estimate mean deviation (MD) and 52 threshold values. This retrospective study used data from patients who underwent a complete glaucoma examination, including a reliable Humphrey Field Analyzer (HFA) 24-2 SITA Standard (SS) VF exam and a SPECTRALIS OCT. Results: For MD estimation, weighted prediction averaging of all four individuals yielded a mean absolute error (MAE) of 2.89 dB (2.50-3.30) on 186 test images, reducing the baseline by 54% (MAEdecr%). For 52 VF threshold values' estimation, the weighted ensemble model resulted in an MAE of 4.82 dB (4.45-5.22), representing an MAEdecr% of 38% from baseline when predicting the pointwise mean value. DL managed to explain 75% and 58% of the variance (R2) in MD and pointwise sensitivity estimation, respectively. Conclusions: Deep learning can estimate global and pointwise VF sensitivities that fall almost entirely within the 90% test-retest confidence intervals of the 24-2 SS test. Translational Relevance: Fast and consistent VF prediction from unsegmented OCT scans could become a solution for visual function estimation in patients unable to perform reliable VF exams.

Deep learning on fundus images detects glaucoma beyond the optic disc.

Although unprecedented sensitivity and specificity values are reported, recent glaucoma detection deep learning models lack in decision transparency. Here, we propose a methodology that advances explainable deep learning in the field of glaucoma detection and vertical cup-disc ratio (VCDR), an important risk factor. We trained and evaluated deep learning models using fundus images that underwent a certain cropping policy. We defined the crop radius as a percentage of image size, centered on the optic nerve head (ONH), with an equidistant spaced range from 10-60% (ONH crop policy). The inverse of the cropping mask was also applied (periphery crop policy). Trained models using original images resulted in an area under the curve (AUC) of 0.94 [95% CI 0.92-0.96] for glaucoma detection, and a coefficient of determination (R2) equal to 77% [95% CI 0.77-0.79] for VCDR estimation. Models that were trained on images with absence of the ONH are still able to obtain significant performance (0.88 [95% CI 0.85-0.90] AUC for glaucoma detection and 37% [95% CI 0.35-0.40] R2 score for VCDR estimation in the most extreme setup of 60% ONH crop). Our findings provide the first irrefutable evidence that deep learning can detect glaucoma from fundus image regions outside the ONH.

Pathological myopia classification with simultaneous lesion segmentation using deep learning.

BACKGROUND AND OBJECTIVES: Pathological myopia (PM) is the seventh leading cause of blindness, with a reported global prevalence up to 3%. Early and automated PM detection from fundus images could aid to prevent blindness in a world population that is characterized by a rising myopia prevalence. We aim to assess the use of convolutional neural networks (CNNs) for the detection of PM and semantic segmentation of myopia-induced lesions from fundus images on a recently introduced reference data set. METHODS: This investigation reports on the results of CNNs developed for the recently introduced Pathological Myopia (PALM) dataset, which consists of 1200 images. Our CNN bundles lesion segmentation and PM classification, as the two tasks are heavily intertwined. Domain knowledge is also inserted through the introduction of a new Optic Nerve Head (ONH)-based prediction enhancement for the segmentation of atrophy and fovea localization. Finally, we are the first to approach fovea localization using segmentation instead of detection or regression models. Evaluation metrics include area under the receiver operating characteristic curve (AUC) for PM detection, Euclidean distance for fovea localization, and Dice and F1 metrics for the semantic segmentation tasks (optic disc, retinal atrophy and retinal detachment). RESULTS: Models trained with 400 available training images achieved an AUC of 0.9867 for PM detection, and a Euclidean distance of 58.27 pixels on the fovea localization task, evaluated on a test set of 400 images. Dice and F1 metrics for semantic segmentation of lesions scored 0.9303 and 0.9869 on optic disc, 0.8001 and 0.9135 on retinal atrophy, and 0.8073 and 0.7059 on retinal detachment, respectively. CONCLUSIONS: We report a successful approach for a simultaneous classification of pathological myopia and segmentation of associated lesions. Our work was acknowledged with an award in the context of the "Pathological Myopia detection from retinal images" challenge held during the IEEE International Symposium on Biomedical Imaging (April 2019). Considering that (pathological) myopia cases are often identified as false positives and negatives in glaucoma deep learning models, we envisage that the current work could aid in future research to discriminate between glaucomatous and highly-myopic eyes, complemented by the localization and segmentation of landmarks such as fovea, optic disc and atrophy.

Age and sex affect deep learning prediction of cardiometabolic risk factors from retinal images.

Deep neural networks can extract clinical information, such as diabetic retinopathy status and individual characteristics (e.g. age and sex), from retinal images. Here, we report the first study to train deep learning models with retinal images from 3,000 Qatari citizens participating in the Qatar Biobank study. We investigated whether fundus images can predict cardiometabolic risk factors, such as age, sex, blood pressure, smoking status, glycaemic status, total lipid panel, sex steroid hormones and bioimpedance measurements. Additionally, the role of age and sex as mediating factors when predicting cardiometabolic risk factors from fundus images was studied. Predictions at person-level were made by combining information of an optic disc centred and a macula centred image of both eyes with deep learning models using the MobileNet-V2 architecture. An accurate prediction was obtained for age (mean absolute error (MAE): 2.78 years) and sex (area under the curve: 0.97), while an acceptable performance was achieved for systolic blood pressure (MAE: 8.96 mmHg), diastolic blood pressure (MAE: 6.84 mmHg), Haemoglobin A1c (MAE: 0.61%), relative fat mass (MAE: 5.68 units) and testosterone (MAE: 3.76 nmol/L). We discovered that age and sex were mediating factors when predicting cardiometabolic risk factors from fundus images. We have found that deep learning models indirectly predict sex when trained for testosterone. For blood pressure, Haemoglobin A1c and relative fat mass an influence of age and sex was observed. However, achieved performance cannot be fully explained by the influence of age and sex. In conclusion we confirm that age and sex can be predicted reliably from a fundus image and that unique information is stored in the retina that relates to blood pressure, Haemoglobin A1c and relative fat mass. Future research should focus on stratification when predicting person characteristics from a fundus image.

Systematic Comparison of Heatmapping Techniques in Deep Learning in the Context of Diabetic Retinopathy Lesion Detection.

Purpose: Heatmapping techniques can support explainability of deep learning (DL) predictions in medical image analysis. However, individual techniques have been mainly applied in a descriptive way without an objective and systematic evaluation. We investigated comparative performances using diabetic retinopathy lesion detection as a benchmark task. Methods: The Indian Diabetic Retinopathy Image Dataset (IDRiD) publicly available database contains fundus images of diabetes patients with pixel level annotations of diabetic retinopathy (DR) lesions, the ground truth for this study. Three in advance trained DL models (ResNet50, VGG16 or InceptionV3) were used for DR detection in these images. Next, explainability was visualized with each of the 10 most used heatmapping techniques. The quantitative correspondence between the output of a heatmap and the ground truth was evaluated with the Explainability Consistency Score (ECS), a metric between 0 and 1, developed for this comparative task. Results: In case of the overall DR lesions detection, the ECS ranged from 0.21 to 0.51 for all model/heatmapping combinations. The highest score was for VGG16+Grad-CAM (ECS = 0.51; 95% confidence interval [CI]: [0.46; 0.55]). For individual lesions, VGG16+Grad-CAM performed best on hemorrhages and hard exudates. ResNet50+SmoothGrad performed best for soft exudates and ResNet50+Guided Backpropagation performed best for microaneurysms. Conclusions: Our empirical evaluation on the IDRiD database demonstrated that the combination DL model/heatmapping affects explainability when considering common DR lesions. Our approach found considerable disagreement between regions highlighted by heatmaps and expert annotations. Translational Relevance: We warrant a more systematic investigation and analysis of heatmaps for reliable explanation of image-based predictions of deep learning models.

Accurate prediction of glaucoma from colour fundus images with a convolutional neural network that relies on active and transfer learning.

PURPOSE: To assess the use of deep learning (DL) for computer-assisted glaucoma identification, and the impact of training using images selected by an active learning strategy, which minimizes labelling cost. Additionally, this study focuses on the explainability of the glaucoma classifier. METHODS: This original investigation pooled 8433 retrospectively collected and anonymized colour optic disc-centred fundus images, in order to develop a deep learning-based classifier for glaucoma diagnosis. The labels of the various deep learning models were compared with the clinical assessment by glaucoma experts. Data were analysed between March and October 2018. Sensitivity, specificity, area under the receiver operating characteristic curve (AUC), and amount of data used for discriminating between glaucomatous and non-glaucomatous fundus images, on both image and patient level. RESULTS: Trained using 2072 colour fundus images, representing 42% of the original training data, the trained DL model achieved an AUC of 0.995, sensitivity and specificity of, respectively, 98.0% (CI 95.5%-99.4%) and 91% (CI 84.0%-96.0%), for glaucoma versus non-glaucoma patient referral. CONCLUSIONS: These results demonstrate the benefits of deep learning for automated glaucoma detection based on optic disc-centred fundus images. The combined use of transfer and active learning in the medical community can optimize performance of DL models, while minimizing the labelling cost of domain-specific mavens. Glaucoma experts are able to make use of heat maps generated by the deep learning classifier to assess its decision, which seems to be related to inferior and superior neuroretinal rim (within ONH), and RNFL in superotemporal and inferotemporal zones (outside ONH).

Artery-vein segmentation in fundus images using a fully convolutional network.

Epidemiological studies demonstrate that dimensions of retinal vessels change with ocular diseases, coronary heart disease and stroke. Different metrics have been described to quantify these changes in fundus images, with arteriolar and venular calibers among the most widely used. The analysis often includes a manual procedure during which a trained grader differentiates between arterioles and venules. This step can be time-consuming and can introduce variability, especially when large volumes of images need to be analyzed. In light of the recent successes of fully convolutional networks (FCNs) applied to biomedical image segmentation, we assess its potential in the context of retinal artery-vein (A/V) discrimination. To the best of our knowledge, a deep learning (DL) architecture for simultaneous vessel extraction and A/V discrimination has not been previously employed. With the aim of improving the automation of vessel analysis, a novel application of the U-Net semantic segmentation architecture (based on FCNs) on the discrimination of arteries and veins in fundus images is presented. By utilizing DL, results are obtained that exceed accuracies reported in the literature. Our model was trained and tested on the public DRIVE and HRF datasets. For DRIVE, measuring performance on vessels wider than two pixels, the FCN achieved accuracies of 94.42% and 94.11% on arteries and veins, respectively. This represents a decrease in error of 25% over the previous state of the art reported by Xu et al. (2017). Additionally, we introduce the HRF A/V ground truth, on which our model achieves 96.98% accuracy on all discovered centerline pixels. HRF A/V ground truth validated by an ophthalmologist, predicted A/V annotations and evaluation code are available at https://github.com/rubenhx/av-segmentation.

Participatory Patterns in an International Air Quality Monitoring Initiative.

The issue of sustainability is at the top of the political and societal agenda, being considered of extreme importance and urgency. Human individual action impacts the environment both locally (e.g., local air/water quality, noise disturbance) and globally (e.g., climate change, resource use). Urban environments represent a crucial example, with an increasing realization that the most effective way of producing a change is involving the citizens themselves in monitoring campaigns (a citizen science bottom-up approach). This is possible by developing novel technologies and IT infrastructures enabling large citizen participation. Here, in the wider framework of one of the first such projects, we show results from an international competition where citizens were involved in mobile air pollution monitoring using low cost sensing devices, combined with a web-based game to monitor perceived levels of pollution. Measures of shift in perceptions over the course of the campaign are provided, together with insights into participatory patterns emerging from this study. Interesting effects related to inertia and to direct involvement in measurement activities rather than indirect information exposure are also highlighted, indicating that direct involvement can enhance learning and environmental awareness. In the future, this could result in better adoption of policies towards decreasing pollution.

The Aeroflex: a bicycle for mobile air quality measurements.

Fixed air quality stations have limitations when used to assess people's real life exposure to air pollutants. Their spatial coverage is too limited to capture the spatial variability in, e.g., an urban or industrial environment. Complementary mobile air quality measurements can be used as an additional tool to fill this void. In this publication we present the Aeroflex, a bicycle for mobile air quality monitoring. The Aeroflex is equipped with compact air quality measurement devices to monitor ultrafine particle number counts, particulate mass and black carbon concentrations at a high resolution (up to 1 second). Each measurement is automatically linked to its geographical location and time of acquisition using GPS and Internet time. Furthermore, the Aeroflex is equipped with automated data transmission, data pre-processing and data visualization. The Aeroflex is designed with adaptability, reliability and user friendliness in mind. Over the past years, the Aeroflex has been successfully used for high resolution air quality mapping, exposure assessment and hot spot identification.