As the prevalence of diabetes increases, millions of people need to be screened for diabetic retinopathy (DR). Remarkable advances in technology have made it possible to use artificial intelligence to screen DR from retinal images with high accuracy and reliability, resulting in reducing human labor by processing large amounts of data in a shorter time. We developed a fully automated classification algorithm to diagnose DR and identify referable status using optical coherence tomography angiography (OCTA) images with convolutional neural network (CNN) model and verified its feasibility by comparing its performance with that of conventional machine learning model. Ground truths for classifications were made based on ultra-widefield fluorescein angiography to increase the accuracy of data annotation. The proposed CNN classifier achieved an accuracy of 91–98%, a sensitivity of 86–97%, a specificity of 94–99%, and an area under the curve of 0.919–0.976. In the external validation, overall similar performances were also achieved. The results were similar regardless of the size and depth of the OCTA images, indicating that DR could be satisfactorily classified even with images comprising narrow area of the macular region and a single image slab of retina. The CNN-based classification using OCTA is expected to create a novel diagnostic workflow for DR detection and referral.
Polypoidal choroidal vasculopathy (PCV) is a common choroidal vascular disease particularly in Asians. However, the underlying pathogenesis of PCV is still yet to be fully elucidated, and the correlation between choroidal vasculature and treatment response of PCV are poorly understood. Accordingly, we sought to find clues to understand the pathogenesis and prognosis of PCV by quantitatively evaluating choroidal vasculature from the entire fundus using ultra-widefield (UWF) indocyanine green angiography (ICGA). In this study, 32 eyes from 29 patients with treatment naïve PCV and 30 eyes from 30 healthy control participants were enrolled. Choroidal vascular density (CVD) of PCV eyes was higher than normal eyes in majority regions including the periphery. CVD was positively correlated with choroidal thickness and choroidal hyperpermeability, supporting that the pathogenesis of PCV may include choroidal congestion and dilatation. Thicker choroid and higher CVD were also correlated with poor treatment response after anti-VEGF injections. The CVD, quantified from UWF ICGA can also be used as an effective image biomarker to predict the treatment response in PCV.
We aimed to investigate the relationship between non-perfusion on ultra-widefield angiography (UWF FA) and aqueous cytokine levels and central macular thickness (CMT) in eyes with branch retinal vein occlusion (BRVO). Thirty-five eyes with treatment-naïve BRVO were included. Non-perfusion area (NPA) for partial and complete ischemia was manually segmented and the ischemic index (ISI) for each was calculated using stereographically projected UWF FA for four different retinal zones. Partial and complete ischemia had different regional predominance. Partial ischemia was predominant in the posterior regions, while complete ischemia was predominant in the periphery. And partial ischemic area, located posterior to far periphery, showed significant correlation with central macular thickness and concentrations of angiogenic and inflammatory cytokines, while complete ischemic area showed no correlation with any of the parameters. Taken together, partial but not complete ischemia, particularly in the more posterior retina, was associated with higher cytokine levels and more severe macular edema in eyes with BRVO. These findings would help us to better understand the different clinical significance of ischemia in BRVO depending on the severity and regional distribution.
Purpose
To develop an automated diabetic retinopathy (DR) staging system using optical coherence tomography angiography (OCTA) images with a convolutional neural network (CNN) and to verify the feasibility of the system.
Methods
In this retrospective cross-sectional study, a total of 918 data sets of 3 × 3 mm
2
OCTA images and 917 data sets of 6 × 6 mm
2
OCTA images were obtained from 1118 eyes. A deep CNN and four traditional machine learning models were trained with annotations made by a retinal specialist based on ultra-widefield fluorescein angiography. Separately, the same images of the test data sets were independently graded by two human experts. The results of the CNN algorithm were compared with those of traditional machine learning–based classifiers and human experts.
Results
The proposed CNN achieved an accuracy of 0.728, a sensitivity of 0.675, a specificity of 0.944, an F1 score of 0.683, and a quadratic weighted κ of 0.908 for a six-level staging task, which were far superior to the results of traditional machine learning methods or human experts. The CNN algorithm showed a better performance using 6 × 6 mm
2
rather than 3 × 3 mm
2
sized OCTA images and using combined data rather than a separate OCTA layer alone.
Conclusions
CNN-based classification using OCTA images can provide reliable assistance to clinicians for DR classification.
Translational Relevance
This CNN algorithm can guide the clinical decision for invasive angiography or referrals to ophthalmology specialists, helping to create more efficient diagnostic workflow in primary care settings.
Purpose: To evaluate foveal avascular zone (FAZ) microvascular structural changes in healthy Korean subjects stratified by age using optical coherence tomography angiography (OCTA). Methods: Eighty eyes of 43 healthy volunteer subjects who had 20/20 or better best corrected visual acuity without other ocular or systemic disease except vitreous floaters and dry eye syndrome were enrolled and stratified by age group. To measure FAZ size and vascular density (VD) of the superficial capillary plexus (SCP) and deep capillary plexus (DCP), OCTA (DRI OCT Triton, Swept Source OCT, Topcon, Tokyo, Japan) scans were performed over fovea-centered 3 × 3 mm 2 regions, and then compared with central macular thickness (CMT) and subfoveal choroidal thickness.Results: Mean age of the participants was 46.4 ± 16.1 (20-78). The SCP and DCP FAZ sizes were 0.32 ± 0.11 mm 2 and 0.41 ± 0.14 mm 2 , respectively. There was a significant difference by age group (p < 0.001, p < 0.001), respectively. The FAZ VD for SCP and DCP was 28.96 ± 3.05% and 33.15 ± 3.64%, respectively. There was no difference between age groups (p = 0.118, p = 0.637). Univariate and multiple linear regression analysis showed that the FAZ size of SCP and DCP was significantly correlated with age (p = 0.039, p = 0.002) and CMT (p = 0.007, p = 0.013), respectively. The SCP and DCP FAZ size were positively correlated with age (R 2 = 0.279, p < 0.001, R 2 = 0.344, p < 0.001), and negatively correlated with CMT (R 2 = 0.354, p < 0.001, R 2 = 0.285, p < 0.001), respectively.
Conclusions:The FAZ size of SCP and DCP increased with age and were negatively correlated with CMT. These results suggest that consideration of age and CMT is important when performing the clinical evaluation of FAZ size in healthy subjects.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.