Deep learning algorithms have been used to detect diabetic retinopathy (DR) with specialist-level accuracy. This study aims to validate one such algorithm on a large-scale clinical population, and compare the algorithm performance with that of human graders. A total of 25,326 gradable retinal images of patients with diabetes from the community-based, nationwide screening program of DR in Thailand were analyzed for DR severity and referable diabetic macular edema (DME). Grades adjudicated by a panel of international retinal specialists served as the reference standard. Relative to human graders, for detecting referable DR (moderate NPDR or worse), the deep learning algorithm had significantly higher sensitivity (0.97 vs. 0.74, p < 0.001), and a slightly lower specificity (0.96 vs. 0.98, p < 0.001). Higher sensitivity of the algorithm was also observed for each of the categories of severe or worse NPDR, PDR, and DME ( p < 0.001 for all comparisons). The quadratic-weighted kappa for determination of DR severity levels by the algorithm and human graders was 0.85 and 0.78 respectively ( p < 0.001 for the difference). Across different severity levels of DR for determining referable disease, deep learning significantly reduced the false negative rate (by 23%) at the cost of slightly higher false positive rates (2%). Deep learning algorithms may serve as a valuable tool for DR screening.

No abstract

BackgroundNeovascular age-related macular degeneration often requires chronic therapy with anti-VEGF agents, and patients with recurrent disease are challenging to manage.MethodsThis retrospective case series evaluates patients who were switched from bevacizumab or ranibizumab to aflibercept and then back again because of recurrent fluid on optical coherence tomography (OCT) by reporting changes in OCT measurements over the course of medication changes.ResultsTwenty-one eyes in nineteen patients received an average of 20.7 bevacizumab and/or ranibizumab injections and then an average of 7.2 aflibercept injections before being switched back to bevacizumab or ranibizumab because of recurrent fluid on OCT. Median central macular thickness improved on transition from bevacizumab or ranibizumab (317 μm) to aflibercept (285 μm; p = 0.034), then worsened over the course of aflibercept treatment (296 μm; p = 0.080), but improved again with transition from aflibercept back to bevacizumab or ranibizumab (283 μm; p = 0.016). The total volume of subretinal fluid, intraretinal fluid, and pigment epithelial detachments also decreased on transition from bevacizumab or ranibizumab (2.56 mm3) to aflibercept (2.44 mm3; p = 0.080), then worsened over the course of aflibercept treatment (3.18 mm3; p = 0.019), and improved again on transition back to bevacizumab or ranibizumab (2.11 mm3; p = 0.016).ConclusionsWhile aflibercept appears initially effective, some patients develop recurrent fluid with aflibercept that improves with transition back to bevacizumab or ranibizumab. Rotating anti-VEGF agents may be beneficial with recurrent neovascular activity.

Secondary OHT induced by the dexamethasone implant can usually be controlled by medications, but the incidence of OHT requiring glaucoma surgery was high (4.62%) in our study relative to rates previously reported in the literature. All patients, especially those with preexisting glaucoma, should be advised of the possible need for glaucoma surgery prior to undergoing treatment with the dexamethasone implant. [Ophthalmic Surg Lasers Imaging Retina. 2018;49:680-685.].

This is the first case series demonstrating foveal outer retinal atrophy on spectral domain optical coherence tomography after resolution of glaucoma surgery-associated hypotony maculopathy. Both patients had previous uveitis that was controlled at the time of hypotony. Eyes with previous hypotony and compromised ciliary body function may be at greater risk of photoreceptor damage.