Shu Yen Lee

Development and Validation of a Deep Learning System for Diabetic Retinopathy and Related Eye Diseases Using Retinal Images From Multiethnic Populations With Diabetes

Importance A deep learning system (DLS) is a machine learning technology with potential for screening diabetic retinopathy and related eye diseases. Objective To evaluate the performance of a DLS in detecting referable diabetic retinopathy, vision-threatening diabetic retinopathy, possible glaucoma, and age-related macular degeneration (AMD) in community and clinic-based multiethnic populations with diabetes. Design, Setting, and Participants Diagnostic performance of a DLS for diabetic retinopathy and related eye diseases was evaluated using 494 661 retinal images. A DLS was trained for detecting diabetic retinopathy (using 76 370 images), possible glaucoma (125 189 images), and AMD (72 610 images), and performance of DLS was evaluated for detecting diabetic retinopathy (using 112 648 images), possible glaucoma (71 896 images), and AMD (35 948 images). Training of the DLS was completed in May 2016, and validation of the DLS was completed in May 2017 for detection of referable diabetic retinopathy (moderate nonproliferative diabetic retinopathy or worse) and vision-threatening diabetic retinopathy (severe nonproliferative diabetic retinopathy or worse) using a primary validation data set in the Singapore National Diabetic Retinopathy Screening Program and 10 multiethnic cohorts with diabetes. Exposures Use of a deep learning system. Main Outcomes and Measures Area under the receiver operating characteristic curve (AUC) and sensitivity and specificity of the DLS with professional graders (retinal specialists, general ophthalmologists, trained graders, or optometrists) as the reference standard. Results In the primary validation dataset (n = 14 880 patients; 71 896 images; mean [SD] age, 60.2 [2.2] years; 54.6% men), the prevalence of referable diabetic retinopathy was 3.0%; vision-threatening diabetic retinopathy, 0.6%; possible glaucoma, 0.1%; and AMD, 2.5%. The AUC of the DLS for referable diabetic retinopathy was 0.936 (95% CI, 0.925-0.943), sensitivity was 90.5% (95% CI, 87.3%-93.0%), and specificity was 91.6% (95% CI, 91.0%-92.2%). For vision-threatening diabetic retinopathy, AUC was 0.958 (95% CI, 0.956-0.961), sensitivity was 100% (95% CI, 94.1%-100.0%), and specificity was 91.1% (95% CI, 90.7%-91.4%). For possible glaucoma, AUC was 0.942 (95% CI, 0.929-0.954), sensitivity was 96.4% (95% CI, 81.7%-99.9%), and specificity was 87.2% (95% CI, 86.8%-87.5%). For AMD, AUC was 0.931 (95% CI, 0.928-0.935), sensitivity was 93.2% (95% CI, 91.1%-99.8%), and specificity was 88.7% (95% CI, 88.3%-89.0%). For referable diabetic retinopathy in the 10 additional datasets, AUC range was 0.889 to 0.983 (n = 40 752 images). Conclusions and Relevance In this evaluation of retinal images from multiethnic cohorts of patients with diabetes, the DLS had high sensitivity and specificity for identifying diabetic retinopathy and related eye diseases. Further research is necessary to evaluate the applicability of the DLS in health care settings and the utility of the DLS to improve vision outcomes.

Conversion to aflibercept for diabetic macular edema unresponsive to ranibizumab or bevacizumab

Background The purpose of this study was to determine if eyes with diabetic macular edema (DME) unresponsive to ranibizumab or bevacizumab would benefit from conversion to aflibercept. Methods This study was conducted as a retrospective chart review of subjects with DME unresponsive to ranibizumab and/or bevacizumab and subsequently converted to aflibercept. Results In total, 21 eyes from 19 subjects of mean age 62±15 years were included. The majority of subjects were male (63%). The median number of ranibizumab or bevacizumab injections before switching to aflibercept was six, and the median number of aflibercept injections after switching was three. Median follow-up was 5 months after the switch. Mean central foveal thickness (CFT) was 453.52±143.39 mm immediately prior to the switch. Morphologically, intraretinal cysts were present in all cases. Mean CFT after the first injection decreased significantly to 362.57±92.82 mm (Wilcoxon signed-rank test; P<0.001). At the end of follow-up, the mean CFT was 324.17±98.76 mm (P<0.001). Mean visual acuity was 0.42±0.23 logMAR just prior to the switch, 0.39±0.31 logMAR after one aflibercept injection, and 0.37±0.22 log-MAR at the end of follow-up. The final visual acuity was significantly better than visual acuity before the switch (P=0.04). Conclusion Eyes with DME unresponsive to multiple ranibizumab/bevacizumab injections demonstrate anatomical and visual improvement on conversion to aflibercept.

Impact of early and late age-related macular degeneration on vision-specific functioning

Aim To assess the impact of early and late age-related macular degeneration (AMD) on vision-specific functioning in Singapore Malays. Methods AMD was assessed from fundus photographs. The following endpoints were considered for (a) AMD: no AMD, early AMD, and late AMD; (b) drusen: absence and presence; and (c) retinal pigment epithelium (RPE) abnormality: absence and presence. Vision functioning was assessed using the modified VF-11 scale validated using the Rasch analysis. The overall functioning score was used as the main outcome measure. Results Retinal photographs and vision functioning data were available only for 3252 participants. After age standardisation, the prevalence of early AMD was 3.5% and late AMD 0.34%. In multivariate models, after adjusting for age, gender, education, level of income, smoking status, ocular condition and hypertension, only late AMD was independently associated with poorer vision functioning when compared with no AMD or early AMD (β (β regression coefficient)=−6.4 (CI −11.7 to −2.1; p=0.01)). Early AMD or its principal components, drusen or RPE abnormality, were not independently associated with vision functioning (p>0.05). In adjusted multinomial logistic regression models, people with late AMD were twice as likely (OR=2.23; 95% CI 1.16 to 7.11) to have low overall functioning than those without AMD. Conclusions Late AMD has a significant impact on visual functioning, but early AMD, drusen and RPE changes have no impact. These data highlight the importance of preventive public health strategies targeting patients with early AMD signs in order to prevent progression to late AMD when visual function is compromised.

Optical Coherence Tomographic Angiography in Type 2 Diabetes and Diabetic Retinopathy

Importance Optical coherence tomographic angiography (OCT-A) is able to visualize retinal microvasculature without the need for injection of fluorescein contrast dye. Nevertheless, it is only able to capture a limited view of macula and does not show leakage. Objectives To evaluate the retinal microvasculature using OCT-A in patients with type 2 diabetes as well as the association of OCT-A characteristics with diabetic retinopathy (DR) and systemic risk factors. Design, Setting, and Participants A prospective, observational study was conducted from January 1 to June 30, 2016, at medical retina clinics at the Singapore National Eye Center among 50 patients with type 2 diabetes with and without DR (n = 100 eyes). We examined the retinal microvasculature with swept-source OCT-A and a semiautomated software to measure the capillary density index (CDI) and fractal dimension (FD) at the superficial vascular plexus (SVP) and deep retinal vascular plexus (DVP). We collected data on histories of patients’ glycated hemoglobin A1c, hypertension, hyperlipidemia, smoking, and renal impairment. Main Outcomes and Measures The CDI and FD at the SVP and DVP for each severity level of DR and the association of systemic risk factors vs the CDI and FD. Results The mean (SD) glycated hemoglobin A1c of the 50 patients (26 men and 24 women; 35 Chinese; mean [SD] age, 59.5 [8.9] years) was 7.9% (1.7%). The mean (SD) CDI at the SVP decreased from 0.358 (0.017) in patients with no DR to 0.338 (0.012) in patients with proliferative DR (P < .001) and at the DVP decreased in patients with no DR from 0.361 (0.019) to 0.345 (0.020) in patients with proliferative DR (P = .04). The mean (SD) FD at the SVP increased from 1.53 (0.05) in patients with no DR to 1.60 (0.05) in patients with proliferative DR (P < .01) and at the DVP increased from 1.55 (0.06) in patients with no DR to 1.61 (0.05) in patients with proliferative DR (P = .02). For systemic risk factors, hyperlipidemia (odds ratio [OR], 9.82; 95% CI, 6.92-11.23; P < .001), smoking (OR, 10.90; 95% CI, 8.23-12.34; P < .001), and renal impairment (OR, 3.72; 95% CI, 1.80-4.81; P = .05) were associated with reduced CDI, while increased glycated hemoglobin A1c (≥8%) (OR, 8.77; 95% CI, 5.23-10.81; P < .01) and renal impairment (OR, 10.30; 95% CI, 8.21-11.91; P < .001) were associated with increased FD. Conclusions and Relevance Optical coherence tomographic angiography is a novel imaging modality to quantify the retinal capillary microvasculature in patients with diabetes. It can be potentially used in interventional trials to study the effect of systemic risk factors on the microvasculature that was previously not accessible in a noninvasive manner. The relevance of these findings relative to visual acuity, however, remains largely unknown at this time.

A Prospective Study of Treatment Patterns and 1-Year Outcome of Asian Age-Related Macular Degeneration and Polypoidal Choroidal Vasculopathy

Objective To study the treatment patterns and visual outcome over one year in Asian patients with choroidal neovascular membrane secondary to age-related macular degeneration (AMD-CNV) and polypoidal choroidal vasculopathy (PCV). Design Prospective cohort, non-interventional study. Methods 132 treatment-naïve patients who received treatment for AMD-CNV and PCV were included. All patients underwent standardized examination procedures including retinal imaging at baseline and follow-up. AMD-CNV and PCV were defined on fundus fluorescein angiography and indocyanine green angiography at baseline. Patients were treated according to standard of care.We report the visual acuity (VA) and optical coherence tomography (OCT) measurements at baseline, month 3 and month 12 The factors influencing month 12 outcomes were analyzed. Main Outcome Measure Type of treatment, number of Anti-vascular endothelial growth factor (VEGF) treatments, visual outcome over one year. Results Anti-VEGF monotherapy was the initial treatment in 89.1% of AMD-CNV, but only 15.1% of PCV. The mean number of anti-VEGF injections up to month 12 was 3.97 (4.51 AMD-CNV, 3.43 PCV, p = 0.021). Baseline OCT, month 3 OCT and month 3 VA were significant in determining continuation of treatment after month 3. At month12, mean VA improved from 0.82 (∼20/132) at baseline to 0.68 (∼20/96) at month 12 (mean gain 6.5 ETDRS letters, p = 0.002). 34.2% of eyes (38/113 eyes) gained ≥15 ETDRS letters and 14.4% (16/113 eyes) lost ≥15 ETDRS letters. There were no significant differences in visual outcome between AMD-CNV and PCV (p = 0.51). Factors predictive of month 12 visual outcome were baseline VA, baseline OCT central macular thickness, month 3 VA and age. Conclusions There is significant variation in treatment patterns in Asian eyes with exudative maculopathy. There is significant visual improvement in all treatment groups at one year. These data highlight the need for high quality clinical trial data to provide evidence-based management of Asian AMD.

Dynamic responses in retinal vessel caliber with flicker light stimulation in eyes with diabetic retinopathy.

PURPOSE This study investigated the responses of retinal vessels to flickering light in diabetic patients with various stages of diabetic retinopathy (DR). METHODS This cross-sectional observational study evaluated adult subjects with diabetes mellitus. The Dynamic Vessel Analyzer (DVA) was used to measure retinal vascular dilatation in response to diffuse illuminance flicker. Diabetic retinopathy was graded from retinal photography. RESULTS There were 279 subjects in total, with a mean age of 59.9 ± 9.2 years. The majority were male (73%) and the mean HbA1c level and mean duration of diabetes were 7.7% ± 1.4% and 13.9 ± 10.4 years, respectively. After adjustments for age, sex, smoking, duration of diabetes, HbA1c, hypertension, and hyperlipidemia, the responses of both retinal arterioles and venules to flicker stimulation decreased continuously with increasing stages of diabetic retinopathy (P = 0.008 and <0.001, respectively). Subjects with reduced arteriolar dilation responses were more likely to have any DR (odds ratio, OR, 1.20, [95% confidence interval (CI), 1.01-1.45], P = 0.045, per SD decrease). Subjects with reduced venular dilation responses were more likely to have any DR, moderate DR, or vision-threatening DR (OR: 1.27 [1.04-1.53], P = 0.02; OR: 1.27 (1.06-1.49), P = 0.007; and OR: 1.51 (1.14-1.50), P = 0.002; per SD decrease, respectively). CONCLUSIONS The responses of retinal arterioles and venules to flickering light are reduced in subjects with DR, and decrease progressively with more severe stages of DR.

Choroidal thickness and risk characteristics of eyes with myopic choroidal neovascularization

Eye Disease Case-Control Study Group (1993): Risk factors for branch retinal vein occlusion. Am J Ophthalmol 116: 286–296. Ho JD, Tsai CY, Liuo SW et al. (2008): Seasonal variations in the occurrence of retinal vein occlusions: a five year nationwide population-based study from Taiwan. Am J Ophthalmol 145: 722–728. Ho JD, Liou SW & Lin HC (2009): Retinal vein occlusion and the risk of stroke development: a five-year follow-up study. Am J Ophthalmol 147: 283–290. Mitchell P, Smith W & Chang A (1996): Prevalence and associations of retinal vein occlusion in Australia. The Blue Mountains Eye Study. Arch Ophthalmol 114: 1243– 1247. The Central Vein Occlusion Study Group (1997): Natural history and clinical management of central retinal vein occlusion. Arch Ophthalmol 115: 486–491.

Trends and factors related to outcomes for primary rhegmatogenous retinal detachment surgery in a large asian tertiary eye center.

Purpose: To describe trends and outcomes of vitreoretinal surgery for primary rhegmatogenous retinal detachment in a large Asian tertiary eye center. Methods: Retrospective review of 1,530 eyes with primary retinal detachment between 2005 and 2011 managed at the Singapore National Eye Center by one of the following: scleral buckling (SB), pars plana vitrectomy (PPV), and combined SB and PPV (SB + PPV). Anatomical and functional outcomes were assessed. Results: There was a trend toward PPV and PPV + SB as the primary reattachment procedure from 2005 to 2011. The primary anatomical success rate for PPV (78.6%) was worse than that for SB (88.8%) or SB + PPV (89.0%, P = 0.000). Final anatomical success rates were similar for all 3 procedures: SB 97.7%, PPV 95.2%, and SB + PPV 96.4%. Better functional success was achieved in the SB group (86.1%) than both the PPV (72.5%) and SB + PPV groups (77.5%, P = 0.000), partly attributable to the less complex nature of retinal detachments in the SB group. Older age and proliferative vitreoretinopathy were related to the poor functional outcomes in both phakic and pseudophakic eyes. Conclusion: There was an increasing trend toward PPV and PPV + SB as the primary retinal reattachment surgery from 2005 through to 2011. High rates of anatomical and functional outcomes were achieved with SB, PPV, and SB + PPV, proliferative vitreoretinopathy and older age were negatively correlated with the functional success in both phakic and pseudophakic eyes.

THREE-YEAR RESULTS OF POLYPOIDAL CHOROIDAL VASCULOPATHY TREATED WITH PHOTODYNAMIC THERAPY: Retrospective Study and Systematic Review.

Purpose: To evaluate the 3-year outcome in eyes with polypoidal choroidal vasculopathy (PCV) treated with photodynamic therapy with verteporfin. Methods: Retrospective study and review of the literature. We performed a retrospective study of patients with PCV who were treated with photodynamic therapy between January 2007 and December 2008. Patients were excluded if they had received photodynamic therapy before the study period, but those who received previous treatment with other modalities (thermal laser or intravitreal therapies) were allowed. The main outcome measures were best-corrected visual acuity, repeat photodynamic therapy, and recurrence of PCV at the end of Years 1, 2, and 3. We further conducted a systematic review of the literature using the terms “polypoidal choroidal vasculopathy” and “photodynamic therapy” and compared the visual outcome of studies over 3 years using meta-analytical methods. Results: The retrospective study included 68 eyes. The mean best-corrected visual acuity was 0.73 ± 0.56 logMAR (20/107, Snellen equivalent) at baseline, 0.73 ± 0.70 logMAR (20/107, Snellen equivalent) at 1 year, 0.96 ± 0.76 logMAR (20/182, Snellen equivalent) at 2 years, and 1.07 ± 0.81 logMAR (20/235, Snellen equivalent) at 3 years. The cumulative recurrence rates of PCV were 16.1% (1 year), 34.9% (2 years), and 52.7% (3 years) and eyes with recurrence were more likely to suffer ≥3 lines loss compared with eyes without recurrence (63.2 vs. 17.6%, P = 0.006). The systematic review summarized results from 48 published studies and our retrospective study. The pooled analysis from 29 studies (316 eyes reporting the 3-year visual outcome) reported mean best-corrected visual acuity improvement of 0.115 logMAR at 1 year (n = 1,669), 0.066 logMAR at 2 years (n = 701), and 0.027 logMAR at 3 years (n = 316). Reported recurrence rates were 5.9% to 50.0% after 1 year, 9.1% to 83.3% after 2 years, and 40.0% to 78.6% after 3 years or longer of follow-up. Conclusion: The visual outcome in eyes with PCV was stable until 2 years, but the outcome at 3 years worsened, particularly in eyes that experienced recurrence.

Prevalence and clinical correlates of focal choroidal excavation in eyes with age-related macular degeneration, polypoidal choroidal vasculopathy and central serous chorioretinopathy

Purpose To describe the prevalence and clinical characteristics of focal choroidal excavation (FCE) in patients with exudative maculopathy due to age-related macular degeneration with choroidal neovascularisation (AMD-CNV), polypoidal choroidal vasculopathy (PCV) and central serous chorioretinopathy (CSC). Methods Three hundred and forty-three patients (343 presenting eyes and 255 fellow unaffected eyes) from consecutive patients presenting with untreated AMD-CNV, PCV or CSC are prospectively recruited. Two independent retinal specialists masked to the clinical diagnosis graded the presence of FCE by examining the findings from spectral-domain optical coherence tomography (SD-OCT). The frequency and clinical characteristics of FCE in each of the three clinical diagnosis groups were compared. Results The diagnosis in the presenting eye was AMD-CNV in 92 patients, PCV in 149 patients, retinal angiomatous proliferation (RAP) in 3 patients and CSC in 99 patients; 255 fellow eyes free of clinical diseases were also graded. The prevalence of FCE was 2.3% (total 14 eyes; 10 presenting eyes, 4 fellow eyes) out of 598 eyes examined. In presenting eyes, FCE was most prevalent in PCV (6.0%), followed by AMD-CNV (1.0%) and CSC (0%), p=0.02. In fellow eyes, the prevalence of FCE was 2.9%, 0% and 1.2% in patients with PCV, AMD-CNV and CSC, respectively. Eyes with FCE had a significantly longer axial length (24.93±1.65 mm vs 23.49±1.10 mm, p<0.001), but otherwise, all other characteristics were similar. Conclusions FCE is more common in PCV than AMD-CNV and CSC. Disturbance in the choroid/retinal pigment epithelium/Bruch membrane interface affected by FCE may be linked to the pathogenesis of PCV and AMD-CNV.