Development and validation of a deep-learning algorithm for the detection of neovascular age-related macular degeneration from colour fundus photographs.

IMPORTANCE: Detection of early onset neovascular age-related macular degeneration (AMD) is critical to protecting vision. BACKGROUND: To describe the development and validation of a deep-learning algorithm (DLA) for the detection of neovascular age-related macular degeneration. DESIGN: Development and validation of a DLA using retrospective datasets. PARTICIPANTS: We developed and trained the DLA using 56 113 retinal images and an additional 86 162 images from an independent dataset to externally validate the DLA. All images were non-stereoscopic and retrospectively collected. METHODS: The internal validation dataset was derived from real-world clinical settings in China. Gold standard grading was assigned when consensus was reached by three individual ophthalmologists. The DLA classified 31 247 images as gradable and 24 866 as ungradable (poor quality or poor field definition). These ungradable images were used to create a classification model for image quality. Efficiency and diagnostic accuracy were tested using 86 162 images derived from the Melbourne Collaborative Cohort Study. Neovascular AMD and/or ungradable outcome in one or both eyes was considered referable. MAIN OUTCOME MEASURES: Area under the receiver operating characteristic curve (AUC), sensitivity and specificity. RESULTS: In the internal validation dataset, the AUC, sensitivity and specificity of the DLA for neovascular AMD was 0.995, 96.7%, 96.4%, respectively. Testing against the independent external dataset achieved an AUC, sensitivity and specificity of 0.967, 100% and 93.4%, respectively. More than 60% of false positive cases displayed other macular pathologies. Amongst the false negative cases (internal validation dataset only), over half (57.2%) proved to be undetected detachment of the neurosensory retina or RPE layer. CONCLUSIONS AND RELEVANCE: This DLA shows robust performance for the detection of neovascular AMD amongst retinal images from a multi-ethnic sample and under different imaging protocols. Further research is warranted to investigate where this technology could be best utilized within screening and research settings.

Longitudinal Assessment of Rod Function in Intermediate Age-Related Macular Degeneration With and Without Reticular Pseudodrusen.

Purpose: To evaluate rod function longitudinally in intermediate age-related macular degeneration subjects with reticular pseudodrusen (RPD) and without RPD (AMD). Methods: Retinal sensitivities (505 and 625 nm) during dark adaptation, at 14 locations within the central 12° macula were obtained after photobleaching at baseline and 12-month visits. Pointwise sensitivity differences between both stimuli were used to assess static rod function, while rod intercept time (RIT) and rod recovery rate (RRR) were used to evaluate dynamic function. Changes in function over time were compared between groups. Results: A total of 23 controls, 12 AMD, and 13 RPD cases were followed-up. At baseline, the RPD group had significantly worst static and dynamic rod function compared to AMD and control groups. Static function in AMD was similar to controls. Static and dynamic function across the central 12° was consistent in controls; however, it was most impaired at 4° compared to 12° eccentricity in disease groups. Over 12 months, no AMD cases progressed clinically and static function in AMD improved (P ≤ 0.04), but remained unchanged in control and RPD groups (P ≥ 0.17). The RRR for control and RPD groups remained stable, while the AMD group deteriorated, but only at 12° (P = 0.02). The RIT was stable in AMD (P = 0.75) and RPD (P = 0.71) groups but improved in the control group (P = 0.002). Conclusions: A decrease in RRR was detected over 12 months at 12° eccentricity in the AMD group. Evaluating changes in rod function requires testing at multiple locations including the peripheral macula.

Static and flicker perimetry in age-related macular degeneration.

PURPOSE: The relationship between clinical severity of age-related macular degeneration (AMD) and macular function has not been well established. In this study, we investigated the correlation between clinical severity and functional deficits as detected by static and flicker perimetry. METHODS: This cross-sectional study consisted of 279 AMD subjects and 24 control participants. AMD subjects were allocated into 1 of 10 AMD severity groups depending on the status of the designated study eye and the fellow eye, as assessed by color fundus photographs. Visual acuity, and static and flicker perimetry were tested on one eye during the same session. The geometric means, SDs, and percentage of abnormal eyes of static and flicker sensitivity of each AMD severity group were determined and compared. RESULTS: The pattern of change in sensitivity and percentage of abnormal eyes for static perimetry across all AMD severity groups were similar to flicker perimetry. Eyes with drusen > 125 µm (P[static] = 0.018, P[flicker] = 0.024), drusenoid epithelial detachment (P[static and flicker] < 0.001) and noncentral geographic atrophy (GA; P[static and flicker] < 0.001) had significant reductions in static and flicker sensitivities compared to normal eyes. Static (ß-coefficient -1.59, 95% confidence interval [CI] -4.78-1.60) and flicker (ß-coefficient -1.29, 95% CI -4.66-2.08) sensitivities declined at a similar rate in eyes that showed clinical signs of progression. CONCLUSIONS: Static and flicker perimetry were affected similarly across the spectrum of AMD severity, and methods appeared to be valid techniques for assessing retinal sensitivity in AMD once drusen > 125 µm are present, but before the development of late AMD.

Role of flicker perimetry in predicting onset of late-stage age-related macular degeneration.

OBJECTIVE: To investigate the longitudinal changes in flicker perimetry in patients with age-related macular degeneration (AMD) as the condition progresses from early AMD to geographic atrophy (GA) or choroidal neovascularization (CNV). METHODS: Patients with AMD and control subjects were recruited from a longitudinal study of retinal function in early AMD consisting of 187 participants. Only those who completed at least 4 consecutive, 6-monthly flicker perimetry tests were selected for this study. Study groups consisted of everyone who went on to develop GA (n = 16) or CNV (n = 5), controls (n = 24), and the high-risk, early- AMD participants whose eyes did not progress to GA or CNV (drusen >125 µm; n = 18). The flicker sensitivity was determined, and its rate of change during the 18 months before the clinical detection of late AMD was calculated. RESULTS: Eyes that went on to develop GA or CNV had a significantly reduced mean (SD) flicker sensitivity in the months before clinical detection of GA (15.8 [5.6] dB) or CNV (19.1 [3.8] dB) compared with control eyes (22.9 [3.0] dB) (P < .001) and with eyes that did not progress to GA or CNV (21.4 [3.4] dB) (P < .001). The rate of change in flicker sensitivity was significantly increased in GA eyes (-0.07 dB/mo) (P < .001) but not in CNV eyes (0.006 dB/mo) (P = .56) compared with the control eyes (-0.003 dB/mo). CONCLUSIONS: Flicker sensitivity is reduced in eyes that go on to develop late AMD. The rate of change in flicker sensitivities over time was particularly useful in predicting eyes and areas within the eye that subsequently develop GA.

Identification of urinary biomarkers for age-related macular degeneration.

PURPOSE: Age-related macular degeneration (AMD) can be considered as a chronic low-grade systemic inflammatory disease. This study was undertaken to test the associations of AMD with the urinary proinflammatory cytokines transforming growth factor (TGF)-ß1, macrophage chemoattractant protein (MCP)-1 and C3a-desArg, as potential noninvasive biomarkers for monitoring AMD. METHODS: A cross-sectional study of 103 AMD cases, comprising early AMD (n = 51), geographic atrophy (GA; n = 19), or choroidal neovascularization (CNV; 33), and 54 unrelated controls, aged 73 ± 9 years, who attended the Royal Victorian Eye and Ear Hospital and private practice in Victoria, Australia. AMD status was determined from the bilateral retinal digital photographs and through angiography and optical coherence tomography images when confirmation of CNV was needed. Serum and urine cytokine levels were measured by immunoassay and the rs1061170 (Y402H) single-nucleotide polymorphism of the complement factor H (CFH) gene was determined. RESULTS: Multivariate logistic regression analyses demonstrated significant associations of urinary TGF-ß1 levels (odds ratio [95% confidence interval]: OR = 1.24 [1.02-1.50]; P < 0.031) and MCP-1 levels (OR = 1.07 [1.02-1.12]; P < 0.008), in early AMD, and also MCP-1 levels with GA (OR = 1.10 [1.03-1.17]; P < 0.003). There was no correlation between urinary and serum cytokine levels. Individuals with one or more copies of the C allele (Y402H) were 2.5 times more likely to have urinary MCP-1 above median levels (P < 0.040). CONCLUSIONS: This study demonstrates a novel finding of an association between elevated urinary cytokines TGF-ß1 and MCP-1 and AMD. Further development of a urinary biomarker profile could provide a practical tool for detection of early AMD, progression monitoring, and assessment of treatment efficacy.