Diabetic Macular Edema Detection Using End-to-End Deep Fusion Model and Anatomical Landmark Visualization on an Edge Computing Device

Purpose

Diabetic macular edema (DME) is a common cause of vision impairment and blindness in patients with diabetes. However, vision loss can be prevented by regular eye examinations during primary care. This study aimed to design an artificial intelligence (AI) system to facilitate ophthalmology referrals by physicians.

Methods

We developed an end-to-end deep fusion model for DME classification and hard exudate (HE) detection. Based on the architecture of fusion model, we also applied a dual model which included an independent classifier and object detector to perform these two tasks separately. We used 35,001 annotated fundus images from three hospitals between 2007 and 2018 in Taiwan to create a private dataset. The Private dataset, Messidor-1 and Messidor-2 were used to assess the performance of the fusion model for DME classification and HE detection. A second object detector was trained to identify anatomical landmarks (optic disc and macula). We integrated the fusion model and the anatomical landmark detector, and evaluated their performance on an edge device, a device with limited compute resources.

Results

For DME classification of our private testing dataset, Messidor-1 and Messidor-2, the area under the receiver operating characteristic curve (AUC) for the fusion model had values of 98.1, 95.2, and 95.8%, the sensitivities were 96.4, 88.7, and 87.4%, the specificities were 90.1, 90.2, and 90.2%, and the accuracies were 90.8, 90.0, and 89.9%, respectively. In addition, the AUC was not significantly different for the fusion and dual models for the three datasets (p = 0.743, 0.942, and 0.114, respectively). For HE detection, the fusion model achieved a sensitivity of 79.5%, a specificity of 87.7%, and an accuracy of 86.3% using our private testing dataset. The sensitivity of the fusion model was higher than that of the dual model (p = 0.048). For optic disc and macula detection, the second object detector achieved accuracies of 98.4% (optic disc) and 99.3% (macula). The fusion model and the anatomical landmark detector can be deployed on a portable edge device.

Conclusion

This portable AI system exhibited excellent performance for the classification of DME, and the visualization of HE and anatomical locations. It facilitates interpretability and can serve as a clinical reference for physicians. Clinically, this system could be applied to diabetic eye screening to improve the interpretation of fundus imaging in patients with DME.

Intravitreal aflibercept partially reverses severe non-proliferative diabetic retinopathy in treatment-naïve patients

Objective

To evaluate whether diabetic retinopathy can be reversed after aflibercept, based on improvements in diabetic macular edema, hard exudates (HEs) of the posterior pole, and retinal microaneurysms (MAs).

Methods

This was a single-center retrospective study of 30 patients (34 eyes) with severe non-proliferative diabetic retinopathy (NPDR) who were treated between August and October 2018. Best-corrected visual acuity (BCVA), central foveal thickness (CFT), area of HEs, and number of MAs were compared before and after treatment.

Results

The mean patient age was 61.4 ± 7.1 years; 14 patients (46.7%) were men. The mean number of injections per patient was 3.5 ± 0.5. The time between the last injection and the last follow-up was 82 days (range, 78–110 days). Six months after the first intravitreal injection, significant improvement was observed in BCVA (from 0.70 ± 0.18 to 0.42 ± 0.19 logMAR), CFT (from 377.17 ± 60.41 to 261.21 ± 31.50 µm), and number of MAs (from 182.2 ± 77.4 to 101.5 ± 59.6). Observations over 6 months after the first intravitreal injection showed a statistically significant reduction in the area of HEs (P = 0.007). No adverse events occurred during the treatment period.

Conclusion

Diabetic retinopathy might be partially reversed by aflibercept treatment, as indicated by BCVA, CFT, number of MAs, and area of HEs.

Efficacy of intravitreal injection of conbercept on non-proliferative diabetic retinopathy: a retrospective study

Objective

This study assessed the efficacy of conbercept for patients with non-proliferative diabetic retinopathy (NPDR).

Methods

In this retrospective clinical study, 54 patients with NPDR (54 eyes) were treated with intravitreal injection of conbercept using a 3+ pro re nata regimen and followed up for 12 months. Best corrected visual acuity (BCVA), central foveal thickness (CFT), area of hard exudate (HE), and number of microaneurysms (MAs) were used as indicators of therapeutic effects. Systemic adverse reactions were recorded to assess safety.

Results

During the 12-month follow-up period, the mean number of injections was 6.12 ± 1.89 on demand. From baseline to the 12-month follow-up, the BCVA of patients with NPDR increased from 0.71 ± 0.20 logMAR to 0.43 ± 0.16 logMAR, CFT decreased from 424.26 ± 64.89 μm to 269.27 ± 44.79 μm, and the number of MAs declined from 79.53 ± 27.18 to 33.34 ± 16.53. Moreover, the area of HE was significantly reduced after 9 months of treatment. There were no serious systemic adverse events during the follow-up.

Conclusions

Intravitreal injection of conbercept has a stable and robust effect on patients with NPDR over a 12-month follow-up period. Thus, conbercept is an effective and feasible treatment for NPDR.

Quantitative measurement of hard exudates in patients with diabetes and their associations with serum lipid levels

PURPOSE

To describe a reproducible method of quantifying macular hard exudates (HEs) in diabetic maculopathy and determine the associations of HEs with systemic risk factors.

METHODS

Patients with diabetes were recruited from a tertiary eye hospital in Melbourne, Australia. Total macular area covered by HEs (total HE area) and the distance from the foveal center to the nearest HE were measured in a semi-automated manner. Associations between HE parameters and diabetic complication risk factors were examined using multiple linear regression models.

RESULTS

Of 593 participants (mean age 60.5-years old), 97 (16.4%) had HEs in at least one eye, due to diabetic maculopathy. The intraclass correlation coefficients (ICC) for inter- and intra-observer agreements were 0.982 and 0.999, respectively. Total HE area was positively associated with qualitative HE severity scale determined by photographic graders. The median of total HE area was 0.089 mm(2) (interquartile range, 0.027-0.246). The median distance between foveal center and the nearest HE was 791.1 μm (431.9-1385.4). After adjusting for age, sex, duration of diabetes, glycated hemoglobin, mean arterial pressure, diabetic retinopathy level, and use of lipid-lowering medication, low density lipoprotein (LDL) cholesterol (P = 0.009), and triglyceride levels (P = 0.036) were positively associated with total HE area. Higher triglyceride levels were associated with central involvement (P = 0.023).

CONCLUSIONS

Quantitative measurement of HEs in patients with diabetes is associated with lipid levels, and higher triglyceride levels are associated with a higher risk of central involvement. Quantitative information may be useful to monitor HE progression or treatment response in persons with diabetic maculopathy.