Automated analysis of digital fundus autofluorescence images of geographic atrophy in advanced age-related macular degeneration using confocal scanning laser ophthalmoscopy (cSLO)

Geographic Atrophy Segmentation Using Multimodal Deep Learning

Purpose

To examine deep learning (DL)-based methods for accurate segmentation of geographic atrophy (GA) lesions using fundus autofluorescence (FAF) and near-infrared (NIR) images.

Methods

This retrospective analysis utilized imaging data from study eyes of patients enrolled in Proxima A and B (NCT02479386; NCT02399072) natural history studies of GA. Two multimodal DL networks (UNet and YNet) were used to automatically segment GA lesions on FAF; segmentation accuracy was compared with annotations by experienced graders. The training data set comprised 940 image pairs (FAF and NIR) from 183 patients in Proxima B; the test data set comprised 497 image pairs from 154 patients in Proxima A. Dice coefficient scores, Bland-Altman plots, and Pearson correlation coefficient (r) were used to assess performance.

Results

On the test set, Dice scores for the DL network to grader comparison ranged from 0.89 to 0.92 for screening visit; Dice score between graders was 0.94. GA lesion area correlations (r) for YNet versus grader, UNet versus grader, and between graders were 0.981, 0.959, and 0.995, respectively. Longitudinal GA lesion area enlargement correlations (r) for screening to 12 months (n = 53) were lower (0.741, 0.622, and 0.890, respectively) compared with the cross-sectional results at screening. Longitudinal correlations (r) from screening to 6 months (n = 77) were even lower (0.294, 0.248, and 0.686, respectively).

Conclusions

Multimodal DL networks to segment GA lesions can produce accurate results comparable with expert graders.

Translational Relevance

DL-based tools may support efficient and individualized assessment of patients with GA in clinical research and practice.

Detection of macular atrophy in age-related macular degeneration aided by artificial intelligence

INTRODUCTION

Age-related macular degeneration (AMD) is a leading cause of irreversible visual impairment worldwide. The endpoint of AMD, both in its dry or wet form, is macular atrophy (MA) which is characterized by the permanent loss of the RPE and overlying photoreceptors either in dry AMD or in wet AMD. A recognized unmet need in AMD is the early detection of MA development.

AREAS COVERED

Artificial Intelligence (AI) has demonstrated great impact in detection of retinal diseases, especially with its robust ability to analyze big data afforded by ophthalmic imaging modalities, such as color fundus photography (CFP), fundus autofluorescence (FAF), near-infrared reflectance (NIR), and optical coherence tomography (OCT). Among these, OCT has been shown to have great promise in identifying early MA using the new criteria in 2018.

EXPERT OPINION

There are few studies in which AI-OCT methods have been used to identify MA; however, results are very promising when compared to other imaging modalities. In this paper, we review the development and advances of ophthalmic imaging modalities and their combination with AI technology to detect MA in AMD. In addition, we emphasize the application of AI-OCT as an objective, cost-effective tool for the early detection and monitoring of the progression of MA in AMD.

Normalization of series of fundus images to monitor the geographic atrophy growth in dry age-related macular degeneration

BACKGROUND AND OBJECTIVE

Age-related macular degeneration (ARMD) is a degenerative disease that affects the retina, and the leading cause of visual loss. In its dry form, the pathology is characterized by the progressive, centrifugal expansion of retinal lesions, called geographic atrophy (GA). In infrared eye fundus images, the GA appears as localized bright areas and its growth can be observed in series of images acquired at regular time intervals. However, illumination distortions between the images make impossible the direct comparison of intensities in order to study the GA progress. Here, we propose a new method to compensate for illumination distortion between images.

METHODS

We process all images of the series so that any two images have comparable gray levels. Our approach relies on an illumination/reflectance model. We first estimate the pixel-wise illumination ratio between any two images of the series, in a recursive way; then we correct each image against all the others, based on those estimates. The algorithm is applied on a sliding temporal window to cope with large changes in reflectance. We also propose morphological processing to suppress illumination artefacts.

RESULTS

The corrected illumination function is homogeneous in the series, enabling the direct comparison of grey-levels intensities in each pixel, and so the detection of the GA growth between any two images. To demonstrate that, we present numerous experiments performed on a dataset of 18 series (328 images), manually segmented by an ophthalmologist. First, we show that the normalization preprocessing dramatically increases the contrast of the GA growth areas. Secondly, we apply segmentation algorithms derived from Otsu's thresholding to detect automatically the GA total growth and the GA progress between consecutive images. We demonstrate qualitatively and quantitatively that these algorithms, although fully automatic, unsupervised and basic, already lead to interesting segmentation results when applied to the normalized images. Colored maps representing the GA evolution can be derived from the segmentations.

CONCLUSION

To our knowledge, the proposed method is the first one which corrects automatically and jointly the illumination inhomogeneity in a series of fundus images, regardless of the number of images, the size, shape and progression of lesion areas. This algorithm greatly facilitates the visual interpretation by the medical expert. It opens up the possibility of treating automatically each series as a whole (not just in pairs of images) to model the GA growth.

Unsupervised Approaches for the Segmentation of Dry ARMD Lesions in Eye Fundus cSLO Images

Age-related macular degeneration (ARMD), a major cause of sight impairment for elderly people, is still not well understood despite intensive research. Measuring the size of the lesions in the fundus is the main biomarker of the severity of the disease and as such is widely used in clinical trials yet only relies on manual segmentation. Artificial intelligence, in particular automatic image analysis based on neural networks, has a major role to play in better understanding the disease, by analyzing the intrinsic optical properties of dry ARMD lesions from patient images. In this paper, we propose a comparison of automatic segmentation methods (classical computer vision method, machine learning method and deep learning method) in an unsupervised context applied on cSLO IR images. Among the methods compared, we propose an adaptation of a fully convolutional network, called W-net, as an efficient method for the segmentation of ARMD lesions. Unlike supervised segmentation methods, our algorithm does not require annotated data which are very difficult to obtain in this application. Our method was tested on a dataset of 328 images and has shown to reach higher quality results than other compared unsupervised methods with a F1 score of 0.87, while having a more stable model, even though in some specific cases, texture/edges-based methods can produce relevant results.

Manual Versus Semi-Automated Measurement of Geographic Atrophy Area in Eyes With Age-Related Macular Degeneration

Purpose

To investigate the agreement between and correlation of manual and semi-automated area measurements of geographic atrophy (GA) in eyes with age-related macular degeneration (AMD) using Heidelberg Eye Explorer and ImageJ software.

Methods

Fundus autofluorescence (FAF) images of eyes with GA secondary to AMD were analyzed. Two graders measured the atrophic area using Heidelberg Eye Explorer manual and semi-automated (RegionFinder) software, as well as ImageJ manual and semi-automated (Color Threshold) software.

Results

Fifty-four FAF images were analyzed. The mean (SD) areas were 10.55 (11.4) mm2 and 9.6 (9.8) mm2 using the Heidelberg manual and semi-automated tools, respectively. The mean (SD) areas were 11.04 (12.25) mm2 and 9.75 (10.3) mm2 using ImageJ manual and semi-automated tools, respectively. Compared with the semi-automated Heidelberg RegionFinder (gold standard) area measurements, Bland-Altman plots showed mean differences of 0.96 mm2, 1.4 mm2, and 0.16 mm2 with manual Heidelberg, manual ImageJ, and semi-automated ImageJ measurements, respectively. Homogeneous GA lesions showed less disparity in area measurements across modalities compared with non-homogeneous lesions.

Conclusions

ImageJ appears to be a reliable tool for GA area measurements when proprietary OCT software is unavailable. Manual measurements with Heidelberg Eye Explorer and ImageJ were comparable, as were semi-automated measurements with Heidelberg RegionFinder and ImageJ Color Threshold.

Translational Relevance

Novel GA measurement techniques using open-source software appear to be comparable to established techniques using proprietary platform-specific software, which may permit more widespread analysis of GA progression from multiple platforms and databases.

Deep Learning Applied to Automated Segmentation of Geographic Atrophy in Fundus Autofluorescence Images

Purpose

This study describes the development of a deep learning algorithm based on the U-Net architecture for automated segmentation of geographic atrophy (GA) lesions in fundus autofluorescence (FAF) images.

Methods

Image preprocessing and normalization by modified adaptive histogram equalization were used for image standardization to improve effectiveness of deep learning. A U-Net-based deep learning algorithm was developed and trained and tested by fivefold cross-validation using FAF images from clinical datasets. The following metrics were used for evaluating the performance for lesion segmentation in GA: dice similarity coefficient (DSC), DSC loss, sensitivity, specificity, mean absolute error (MAE), accuracy, recall, and precision.

Results

In total, 702 FAF images from 51 patients were analyzed. After fivefold cross-validation for lesion segmentation, the average training and validation scores were found for the most important metric, DSC (0.9874 and 0.9779), for accuracy (0.9912 and 0.9815), for sensitivity (0.9955 and 0.9928), and for specificity (0.8686 and 0.7261). Scores for testing were all similar to the validation scores. The algorithm segmented GA lesions six times more quickly than human performance.

Conclusions

The deep learning algorithm can be implemented using clinical data with a very high level of performance for lesion segmentation. Automation of diagnostics for GA assessment has the potential to provide savings with respect to patient visit duration, operational cost and measurement reliability in routine GA assessments.

Translational Relevance

A deep learning algorithm based on the U-Net architecture and image preprocessing appears to be suitable for automated segmentation of GA lesions on clinical data, producing fast and accurate results.

Artificial Intelligence Algorithms for Analysis of Geographic Atrophy: A Review and Evaluation

Purpose

The purpose of this study was to summarize and evaluate artificial intelligence (AI) algorithms used in geographic atrophy (GA) diagnostic processes (e.g. isolating lesions or disease progression).

Methods

The search strategy and selection of publications were both conducted in accordance with the Preferred of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. PubMed and Web of Science were used to extract literary data. The algorithms were summarized by objective, performance, and scope of coverage of GA diagnosis (e.g. lesion automation and GA progression).

Results

Twenty-seven studies were identified for this review. A total of 18 publications focused on lesion segmentation only, 2 were designed to detect and classify GA, 2 were designed to predict future overall GA progression, 3 focused on prediction of future spatial GA progression, and 2 focused on prediction of visual function in GA. GA-related algorithms reported sensitivities from 0.47 to 0.98, specificities from 0.73 to 0.99, accuracies from 0.42 to 0.995, and Dice coefficients from 0.66 to 0.89.

Conclusions

Current GA-AI publications have a predominant focus on lesion segmentation and a minor focus on classification and progression analysis. AI could be applied to other facets of GA diagnoses, such as understanding the role of hyperfluorescent areas in GA. Using AI for GA has several advantages, including improved diagnostic accuracy and faster processing speeds.

Translational Relevance

AI can be used to quantify GA lesions and therefore allows one to impute visual function and quality-of-life. However, there is a need for the development of reliable and objective models and software to predict the rate of GA progression and to quantify improvements due to interventions.

Merging Information From Infrared and Autofluorescence Fundus Images for Monitoring of Chorioretinal Atrophic Lesions

Purpose

To develop a method for automated detection and progression analysis of chorioretinal atrophic lesions using the combined information of standard infrared (IR) and autofluorescence (AF) fundus images.

Methods

Eighteen eyes (from 16 subjects) with punctate inner choroidopathy were analyzed. Macular IR and blue AF images were acquired in all eyes with a Spectralis HRA+OCT device (Heidelberg Engineering, Heidelberg, Germany). Two clinical experts manually segmented chorioretinal lesions on the AF image. AF images were aligned to the corresponding IR. Two random forest models were trained to classify pixels of lesions, one based on the AF image only, the other based on the aligned IR-AF. The models were validated using a leave-one-out cross-validation and were tested against the manual segmentation to compare their performance. A time series from one eye was identified and used to evaluate the method based on the IR-AF in a case study.

Results

The method based on the AF images correctly classified 95% of the pixels (i.e., in vs. out of the lesion) with a Dice's coefficient of 0.80. The method based on the combined IR-AF correctly classified 96% of the pixels with a Dice's coefficient of 0.84.

Conclusions

The automated segmentation of chorioretinal lesions using IR and AF shows closer alignment to manual segmentation than the same method based on AF only. Merging information from multimodal images improves the automatic and objective segmentation of chorioretinal lesions even when based on a small dataset.

Translational Relevance

Merged information from multimodal images improves segmentation performance of chorioretinal lesions.

Analyzing Age-Related Macular Degeneration Progression in Patients with Geographic Atrophy Using Joint Autoencoders for Unsupervised Change Detection

Age-Related Macular Degeneration (ARMD) is a progressive eye disease that slowly causes patients to go blind. For several years now, it has been an important research field to try to understand how the disease progresses and find effective medical treatments. Researchers have been mostly interested in studying the evolution of the lesions using different techniques ranging from manual annotation to mathematical models of the disease. However, artificial intelligence for ARMD image analysis has become one of the main research focuses to study the progression of the disease, as accurate manual annotation of its evolution has proved difficult using traditional methods even for experienced practicians. In this paper, we propose a deep learning architecture that can detect changes in the eye fundus images and assess the progression of the disease. Our method is based on joint autoencoders and is fully unsupervised. Our algorithm has been applied to pairs of images from different eye fundus images time series of 24 ARMD patients. Our method has been shown to be quite effective when compared with other methods from the literature, including non-neural network based algorithms that still are the current standard to follow the disease progression and change detection methods from other fields.

Comparison of Green Versus Blue Fundus Autofluorescence in ABCA4-Related Retinopathy

Purpose

To investigate the interreader and intermodality agreement for grading of retinal pigment epithelium (RPE) atrophy lesion size in ABCA4-related retinopathy using green (GAF) and blue fundus autofluorescence (BAF) imaging.

Methods

In this cross-sectional case series, 97 eyes of 49 patients with RPE atrophy secondary to ABCA4-related retinopathy underwent GAF- (518 nm excitation light) and BAF- (488 nm excitation light) imaging using confocal scanning laser ophthalmoscopy (Spectralis HRA, Heidelberg Engineering, Heidelberg, Germany). Lesions with definitely decreased autofluorescence (DDAF) and questionably decreased autofluorescence (QDAF) in GAF and BAF imaging were analyzed separately by five independent readers using semiautomated software (RegionFinder, Heidelberg Engineering). Intermodality and interreader agreements were assessed for the square-root lesion size, lesion perimeter, and circularity.

Results

GAF- and BAF-based measurements of DDAF and QDAF showed high intermodality and interreader agreement concerning square-root lesion size, as well as shape descriptive parameters (perimeter and circularity). Interreader agreement of square-root lesion size was slightly, hence not significantly higher for GAF-based grading ([95% coefficients of repeatability, intraclass correlation coefficient] DDAF: 0.215 mm, 0.997; QDAF: 0.712 mm, 0.981) compared to BAF-based grading (DDAF: 0.232 mm, 0.997; QDAF: 0.764 mm, 0.978). However, DDAF-measurements revealed distinctly more reproducible results than QDAF-measurements. Foveal sparing did not interfere with intermodality agreement.

Conclusions

Both GAF- and BAF-based quantification of RPE atrophy showed very reliable results with possible superiority of GAF in the context of less energetic excitation light.

Translational Relevance

The high interreader agreement qualifies the use of DDAF progression in GAF and BAF imaging as potential morphologic outcome measure for interventional clinical trials and disease monitoring.

Beyond Retinal Layers: A Deep Voting Model for Automated Geographic Atrophy Segmentation in SD-OCT Images

PURPOSE

To automatically and accurately segment geographic atrophy (GA) in spectral-domain optical coherence tomography (SD-OCT) images by constructing a voting system with deep neural networks without the use of retinal layer segmentation.

METHODS

An automatic GA segmentation method for SD-OCT images based on the deep network was constructed. The structure of the deep network was composed of five layers, including one input layer, three hidden layers, and one output layer. During the training phase, the labeled A-scans with 1024 features were directly fed into the network as the input layer to obtain the deep representations. Then a soft-max classifier was trained to determine the label of each individual pixel. Finally, a voting decision strategy was used to refine the segmentation results among 10 trained models.

RESULTS

Two image data sets with GA were used to evaluate the model. For the first dataset, our algorithm obtained a mean overlap ratio (OR) 86.94% ± 8.75%, absolute area difference (AAD) 11.49% ± 11.50%, and correlation coefficients (CC) 0.9857; for the second dataset, the mean OR, AAD, and CC of the proposed method were 81.66% ± 10.93%, 8.30% ± 9.09%, and 0.9952, respectively. The proposed algorithm was capable of improving over 5% and 10% segmentation accuracy, respectively, when compared with several state-of-the-art algorithms on two data sets.

CONCLUSIONS

Without retinal layer segmentation, the proposed algorithm could produce higher segmentation accuracy and was more stable when compared with state-of-the-art methods that relied on retinal layer segmentation results. Our model may provide reliable GA segmentations from SD-OCT images and be useful in the clinical diagnosis of advanced nonexudative AMD.

TRANSLATIONAL RELEVANCE

Based on the deep neural networks, this study presents an accurate GA segmentation method for SD-OCT images without using any retinal layer segmentation results, and may contribute to improved understanding of advanced nonexudative AMD.

GEOGRAPHIC ATROPHY: Semantic Considerations and Literature Review

PURPOSE

There is a lack of agreement regarding the types of lesions and clinical conditions that should be included in the term "geographic atrophy." Varied and conflicting views prevail throughout the literature and are currently used by retinal experts and other health care professionals.

METHODS

We reviewed the nominal definition of the term "geographic atrophy" and conducted a search of the ophthalmologic literature focusing on preceding terminologies and the first citations of the term "geographic atrophy" secondary to age-related macular degeneration.

RESULTS

According to the nominal definition, the term "geography" stands for a detailed description of the surface features of a specific region, indicating its relative position. However, it does not necessarily imply that the borders of the region must be sharply demarcated or related to any anatomical structures. The term "geographical areas of atrophy" was initially cited in the 1960s in the ophthalmologic literature in the context of uveitic eye disease and shortly thereafter also for the description of variants of "senile macular degeneration." However, no direct explanation could be found in the literature as to why the terms "geographical" and "geographic" were chosen. Presumably the terms were used as the atrophic regions resembled the map of a continent or well-defined country borders on thematic geographical maps. With the evolution of the terminology, the commonly used adjunct "of the retinal pigment epithelium" was frequently omitted and solely the term "geographic atrophy" prevailed for the nonexudative late-stage of age-related macular degeneration itself. Along with the quantification of atrophic areas, based on different imaging modalities and the use of both manual and semiautomated approaches, various and inconsistent definitions for the minimal lesion diameter or size of atrophic lesions have also emerged.

CONCLUSION

Reconsideration of the application of the term "geographic atrophy" in the context of age-related macular degeneration seems to be prudent given ongoing advances in multimodal retinal imaging technology with identification of various phenotypic characteristics, and the observation of atrophy development in eyes under antiangiogenic therapy.

CLINICAL ENDPOINTS FOR THE STUDY OF GEOGRAPHIC ATROPHY SECONDARY TO AGE-RELATED MACULAR DEGENERATION

PURPOSE

To summarize the recent literature describing the application of modern technologies in the study of patients with geographic atrophy (GA) secondary to age-related macular degeneration.

METHODS

Review of the literature describing the terms and definitions used to describe GA, imaging modalities used to capture and measure GA, and the tests of visual function and functional deficits that occur in patients with GA.

RESULTS

In this paper, we describe the evolution of the definitions used to describe GA. We compare imaging modalities used in the characterization of GA, report on the sensitivity and specificity of the techniques where data exist, and describe the correlations between these various modes of capturing the presence of GA. We review the functional tests that have been used in patients with GA, and critically examine their ability to detect and quantify visual deficits.

CONCLUSION

Ophthalmologists and retina specialists now have a wide range of assessments available for the functional and anatomic characterization of GA in patients with age-related macular degeneration. To date, studies have been limited by their unimodal approach, and we recommend that future studies of GA use multimodal imaging. We also suggest strategies for the optimal functional testing of patients with GA.

COMPARISON OF MANUAL AND SEMIAUTOMATED FUNDUS AUTOFLUORESCENCE ANALYSIS OF MACULAR ATROPHY IN STARGARDT DISEASE PHENOTYPE

PURPOSE

To evaluate manual and semiautomated grading techniques for assessing decreased fundus autofluorescence (DAF) in patients with Stargardt disease phenotype.

METHODS

Certified reading center graders performed manual and semiautomated (region finder-based) grading of confocal scanning laser ophthalmoscopy (cSLO) fundus autofluorescence (FAF) images for 41 eyes of 22 patients. Lesion types were defined based on the black level and sharpness of the border: definite decreased autofluorescence (DDAF), well, and poorly demarcated questionably decreased autofluorescence (WDQDAF, PDQDAF). Agreement in grading between the two methods and inter- and intra-grader agreement was assessed by kappa coefficients (κ) and intraclass correlation coefficients (ICC).

RESULTS

The mean ± standard deviation (SD) area was 3.07 ± 3.02 mm for DDAF (n = 31), 1.53 ± 1.52 mm for WDQDAF (n = 9), and 6.94 ± 10.06 mm for PDQDAF (n = 17). The mean ± SD absolute difference in area between manual and semiautomated grading was 0.26 ± 0.28 mm for DDAF, 0.20 ± 0.26 mm for WDQDAF, and 4.05 ± 8.32 mm for PDQDAF. The ICC (95% confidence interval) for method comparison was 0.992 (0.984-0.996) for DDAF, 0.976 (0.922-0.993) for WDQDAF, and 0.648 (0.306-0.842) for PDQDAF. Inter- and intra-grader agreement in manual and semiautomated quantitative grading was better for DDAF (0.981-0.996) and WDQDAF (0.995-0.999) than for PDQDAF (0.715-0.993).

CONCLUSION

Manual and semiautomated grading methods showed similar levels of reproducibility for assessing areas of decreased autofluorescence in patients with Stargardt disease phenotype. Excellent agreement and reproducibility were observed for well demarcated lesions.

Semiautomatic Segmentation of Rim Area Focal Hyperautofluorescence Predicts Progression of Geographic Atrophy Due to Dry Age-Related Macular Degeneration

Purpose

To develop image analysis software usable by nonexpert graders to segment geographic atrophy (GA) from dry AMD and to quantify rim area focal hyperautofluorescence (RAFH) surrounding GA on fundus autofluorescence (FAF) images. To compare the GA progression predictions based on RAFH with those of a validated qualitative classification system.

Methods

Retrospective analysis of serial FAF images from 49 eyes of 30 subjects with GA was performed using MATLAB-based software (MathWorks, Natick, MA, USA). Correlation between RAFH and progression of GA was analyzed using Spearman correlation. Comparisons of lesion growth rate between RAFH tertiles used generalized estimating equations and Kruskal-Wallis testing. Interobserver variability in lesion size, growth rate and RAFH were compared between two expert and one nonexpert grader using Bland-Altman statistics.

Results

Rim area focal hyperautofluorescence was positively correlated with GA progression rate (ρ = 0.49, P < 0.001). Subjects in the middle or highest RAFH tertile were at greater risk of progression (P = 0.005 and P = 0.001, respectively). Mean difference in RAFH was 0.012 between expert and −0.005 to 0.017 between expert and nonexperts. Mean difference in lesion size (mm²) was 0.11 between expert and −0.29 to 0.41 between expert and nonexperts. Mean difference in lesion growth rate (mm²/mo) was 0.0098 between expert and −0.027 to 0.037 between expert and nonexperts. Risk stratification based on RAFH tertile was 96% identical across all graders.

Conclusions

Our semiautomated image analysis software facilitates stratification of progression risk based on RAFH and enabled a nonexpert grader with minimal training to obtain results comparable to expert graders. Predictions based on RAFH were similar to those of a validated qualitative classification system.

Evaluation of Geographic Atrophy from Color Photographs and Fundus Autofluorescence Images: Age-Related Eye Disease Study 2 Report Number 11

PURPOSE

To compare measurements of area of geographic atrophy (GA) and change in GA area from color photographs and fundus autofluorescence (FAF) images.

DESIGN

The Age-Related Eye Disease Study 2 (AREDS2) was a prospective multicenter randomized clinical trial evaluating progression of dry age-related macular degeneration (AMD) using color photographs at annual visits over a 5-year study period. The FAF images were acquired in a subset of participants who joined the FAF ancillary study at any of the annual visits over the study period.

PARTICIPANTS

The AREDS2 FAF ancillary study included 8070 corresponding color and FAF visits of 2202 participants with variable follow-up.

METHODS

Corresponding color and FAF images were independently evaluated at a central reading center for GA area measurement, lesion growth, and involvement of the macula center.

MAIN OUTCOME MEASURES

Presence, area, growth rate of GA, and involvement of center of macula from color and FAF images.

RESULTS

Hypoautofluorescence was visible in 2048 visits (25.4%). Agreement for the presence of GA between the 2 modalities had a kappa of 0.79, with 23% of visits with hypoautofluorescence not presenting with GA on color photographs. Percentage agreement for GA presence ranged from 43% at baseline to 81% at year 5 with improving agreement over time. The mean difference in GA area between the 2 modalities was 0.5 mm², with larger areas on FAF. Growth rate of GA was 1.45 mm² from color photographs and 1.43 mm² from FAF images. The center of the macula was involved in 51% of color photographs and 56% with FAF images.

CONCLUSIONS

Geographic atrophy may be detected earlier by the use of FAF images, but over the course of the study, the 2 modalities become comparable. Progression of GA area is comparable between color photographs and FAF images, but evaluating involvement of the center of the macula may differ, probably because of macular pigmentation blocking autofluorescence.

Modeling Visual Acuity in Geographic Atrophy Secondary to Age-Related Macular Degeneration

Purpose: To analyze and model visual acuity (VA) in geographic atrophy (GA) secondary to age-related macular degeneration (AMD). Methods: The course of VA was analyzed using Turnbull's estimator in 226 eyes with uni- or bilateral GA due to AMD (151 patients; mean age 74.0 ± 7.6 years; mean follow-up time 33.4 ± 23.4 months) from the natural history FAM (Fundus-Autofluorescence Imaging in AMD) study. The variables ‘age at baseline', ‘gender', ‘lesion size', ‘diagnosis of the fellow eye', ‘status of the fovea', ‘focality of the lesion' and ‘pattern' were evaluated for effects on predicting VA using linear mixed-effects models. Results: Mean VA at baseline was 0.6 (Snellen 20/80) ± 0.4 logMAR [range -0.1 to 1.8 (20/17 to hand motions)], showing an estimated mean increase of 0.181 (95% CI 0.152-0.210) and 0.256 (0.214-0.300) after 2 and 4 years of follow-up, respectively. The percentage of eyes with a loss of ≥3 lines was 34% by 2 years and 47% by 4 years. Linear mixed model analysis suggested that 65% of VA variability could be explained by the assessed predictor variables. The strongest effect was found for the ‘status of the fovea' (0.69 logMAR units between ‘definitively spared fovea' and ‘definitive foveal involvement', p < 0.001). The second strongest effect was identified for ‘total lesion size' (effects between 0.02 and 0.09 logMAR units for each mm depending on foveal involvement, p < 0.001, square root transformed values). Conclusions: These findings underscore the importance of GA lesion characteristics as these have the strongest impact on VA. Natural history data and modeling VA to other variables will be helpful for refining outcome parameters and estimating possible benefits of therapeutic interventions.

Natural History of Geographic Atrophy Progression Secondary to Age-Related Macular Degeneration (Geographic Atrophy Progression Study)

PURPOSE

The Geographic Atrophy Progression (GAP) study was designed to assess the rate of geographic atrophy (GA) progression and to identify prognostic factors by measuring the enlargement of the atrophic lesions using fundus autofluorescence (FAF) and color fundus photography (CFP).

DESIGN

Prospective, multicenter, noninterventional natural history study.

PARTICIPANTS

A total of 603 participants were enrolled in the study; 413 of those had gradable lesion data from FAF or CFP, and 321 had gradable lesion data from both FAF and CFP.

METHODS

Atrophic lesion areas were measured by FAF and CFP to assess lesion progression over time. Lesion size assessments and best-corrected visual acuity (BCVA) were conducted at screening/baseline (day 0) and at 3 follow-up visits: month 6, month 12, and month 18 (or early exit).

MAIN OUTCOME MEASURES

The GA lesion progression rate in disease subgroups and mean change from baseline visual acuity.

RESULTS

Mean (standard error) lesion size changes from baseline, determined by FAF and CFP, respectively, were 0.88 (0.1) and 0.78 (0.1) mm(2) at 6 months, 1.85 (0.1) and 1.57 (0.1) mm(2) at 12 months, and 3.14 (0.4) and 3.17 (0.5) mm(2) at 18 months. The mean change in lesion size from baseline to month 12 was significantly greater in participants who had eyes with multifocal atrophic spots compared with those with unifocal spots (P < 0.001) and those with extrafoveal lesions compared with those with foveal lesions (P = 0.001). The mean (standard deviation) decrease in visual acuity was 6.2 ± 15.6 letters for patients with image data available. Atrophic lesions with a diffuse (mean 0.95 mm(2)) or banded (mean 1.01 mm(2)) FAF pattern grew more rapidly by month 6 compared with those with the "none" (mean, 0.13 mm(2)) and focal (mean, 0.36 mm(2)) FAF patterns.

CONCLUSIONS

Although differences were observed in mean lesion size measurements using FAF imaging compared with CFP, the measurements were highly correlated with one another. Significant differences were found in lesion progression rates in participants stratified by hyperfluorescence pattern subtype. This large GA natural history study provides a strong foundation for future clinical trials.

Automated segmentation of geographic atrophy of the retinal epithelium via random forests in AREDS color fundus images

BACKGROUND

Age-related macular degeneration (AMD), left untreated, is the leading cause of vision loss in people older than 55. Severe central vision loss occurs in the advanced stage of the disease, characterized by either the in growth of choroidal neovascularization (CNV), termed the "wet" form, or by geographic atrophy (GA) of the retinal pigment epithelium (RPE) involving the center of the macula, termed the "dry" form. Tracking the change in GA area over time is important since it allows for the characterization of the effectiveness of GA treatments. Tracking GA evolution can be achieved by physicians performing manual delineation of GA area on retinal fundus images. However, manual GA delineation is time-consuming and subject to inter-and intra-observer variability.

METHODS

We have developed a fully automated GA segmentation algorithm in color fundus images that uses a supervised machine learning approach employing a random forest classifier. This algorithm is developed and tested using a dataset of images from the NIH-sponsored Age Related Eye Disease Study (AREDS). GA segmentation output was compared against a manual delineation by a retina specialist.

RESULTS

Using 143 color fundus images from 55 different patient eyes, our algorithm achieved PPV of 0.82±0.19, and NPV of 0:95±0.07.

DISCUSSION

This is the first study, to our knowledge, applying machine learning methods to GA segmentation on color fundus images and using AREDS imagery for testing. These preliminary results show promising evidence that machine learning methods may have utility in automated characterization of GA from color fundus images.

Directional Kinetics of Geographic Atrophy Progression in Age-Related Macular Degeneration with Foveal Sparing

PURPOSE

To describe the directional kinetics of the spread of geographic atrophy (GA) spread in eyes with age-related macular degeneration and foveal sparing.

DESIGN

Prospective, noninterventional natural history study: Fundus Autofluorescence Imaging in Age-Related Macular Degeneration (FAM; clinicaltrials.gov identifier, NCT00393692).

SUBJECTS

Participants of the FAM study exhibiting foveal sparing of GA.

METHODS

Eyes were examined longitudinally with fundus autofluorescence (FAF; excitation wavelength, 488 nm; emission wavelength, >500 nm) and near infrared (NIR) reflectance imaging (Spectralis HRA+OCT or HRA2; Heidelberg Engineering, Heidelberg, Germany). Areas of foveal sparing and GA were measured by 2 independent readers using a semiautomated software tool that allows for combined NIR reflectance and FAF image grading (RegionFinder; Heidelberg Engineering). A linear mixed effect model was used to model GA kinetics over time.

MAIN OUTCOME MEASURE

Change of GA lesion size over time (central vs. peripheral progression).

RESULTS

A total of 47 eyes of 36 patients (mean age, 73.8±7.5 years) met the inclusion criteria. Mean follow-up time was 25.2±16.9 months (range, 5.9-74.6 months). Interreader agreement for measurements of GA and foveal-sparing size were 0.995 and 0.946, respectively. Mean area progression of GA toward the periphery was 2.27±0.22 mm(2)/year and 0.25±0.03 mm(2)/year toward the center. Analysis of square root-transformed data revealed a 2.8-fold faster atrophy progression toward the periphery than toward the fovea. Faster atrophy progression toward the fovea correlated with faster progression toward the periphery in presence of marked interindividual differences.

CONCLUSIONS

The results demonstrate a significantly faster centrifugal than centripetal GA spread in eyes with GA and foveal sparing. Although the underlying pathomechanisms for differential GA progression remain unknown, local factors may be operative that protect the foveal retina-retinal pigment epithelial complex. Quantification of directional spread characteristics and modeling may be useful in the design of interventional clinical trials aiming to prolong foveal survival in eyes with GA.

Ciliary neurotrophic factor for macular telangiectasia type 2: results from a phase 1 safety trial

PURPOSE

To evaluate the safety and tolerability of intraocular delivery of ciliary neurotrophic factor (CNTF) using an encapsulated cell implant for the treatment of macular telangiectasia type 2.

DESIGN

An open-label safety trial conducted in 2 centers enrolling 7 participants with macular telangiectasia type 2.

METHODS

The participant's more severely affected eye (worse baseline visual acuity) received the high-dose implant of CNTF. Patients were followed for a period of 36 months. The primary safety outcome was a change in the parameters of the electroretinogram (ERG). Secondary efficacy outcomes were changes in visual acuity, en face measurements of the optical coherence tomography of the disruption in the ellipsoid zone, and microperimetry when compared with baseline.

RESULTS

The ERG findings demonstrated a reduction in the amplitude of the scotopic b-wave in 4 participants 3 months after implantation (month 3). All parameters returned to baseline values by month 12 and remained so at month 36 with no clinical impact on dark adaptation. There was no change in visual acuity compared with baseline. The area of the defect as measured functionally by microperimetry and structurally by the en face OCT imaging of the ellipsoid zone loss appeared unchanged from baseline.

CONCLUSIONS

The intraocular delivery of CNTF in the encapsulated cell implant appeared to be safe and well tolerated in eyes with macular telangiectasia type 2. Further evaluation in a randomized controlled clinical trial is warranted to test for efficacy.

Semi-automatic geographic atrophy segmentation for SD-OCT images

Geographic atrophy (GA) is a condition that is associated with retinal thinning and loss of the retinal pigment epithelium (RPE) layer. It appears in advanced stages of non-exudative age-related macular degeneration (AMD) and can lead to vision loss. We present a semi-automated GA segmentation algorithm for spectral-domain optical coherence tomography (SD-OCT) images. The method first identifies and segments a surface between the RPE and the choroid to generate retinal projection images in which the projection region is restricted to a sub-volume of the retina where the presence of GA can be identified. Subsequently, a geometric active contour model is employed to automatically detect and segment the extent of GA in the projection images. Two image data sets, consisting on 55 SD-OCT scans from twelve eyes in eight patients with GA and 56 SD-OCT scans from 56 eyes in 56 patients with GA, respectively, were utilized to qualitatively and quantitatively evaluate the proposed GA segmentation method. Experimental results suggest that the proposed algorithm can achieve high segmentation accuracy. The mean GA overlap ratios between our proposed method and outlines drawn in the SD-OCT scans, our method and outlines drawn in the fundus auto-fluorescence (FAF) images, and the commercial software (Carl Zeiss Meditec proprietary software, Cirrus version 6.0) and outlines drawn in FAF images were 72.60%, 65.88% and 59.83%, respectively.

Long-term effectiveness of intravitreal bevacizumab for choroidal neovascularization secondary to angioid streaks in pseudoxanthoma elasticum

PURPOSE

To investigate the long-term effectiveness of intravitreal bevacizumab for treating active choroidal neovascularizations in pseudoxanthoma elasticum (PXE).

METHODS

Fourteen patients (16 eyes) received intravitreal bevacizumab (1.5 mg) and were investigated monthly. Further treatments were administered depending on disease activity. Examinations included best-corrected visual acuity, biomicroscopy, optical coherence tomography, fluorescein angiography and indocyanine green angiography, fundus autofluorescence, and digital fundus photography. Areas of atrophy of the retinal pigment epithelium and retinal fibrosis were quantified using semiautomated detection on fundus autofluorescence images.

RESULTS

Mean age of the cohort was 55 ± 13 years, and mean best-corrected visual acuity at baseline was 20/80 (logarithm of the minimum angle of resolution, 0.56, SD, 0.51). At last follow-up, after an average of 6.5 ± 5.7 injections over 28 months, best-corrected visual acuity was 20/40 (logarithm of the minimum angle of resolution, 0.31, SD, 0.32; P = 0.04). Central retinal thickness was reduced from 254 ± 45 μm to 214 ± 40 μm (P = 0.035). The size of retinal pigment epithelial atrophy and retinal fibrosis measured on fundus autofluorescence images increased in both the treated eye and the fellow eye (P < 0.05). Best-corrected visual acuity of patients with early disease compared with that of those with advanced disease improved significantly more over the treatment course (20/25 vs. 20/63; P = 0.008).

CONCLUSION

Intravitreal bevacizumab therapy demonstrates long-term effectiveness by preserving function in advanced disease and improving function in early disease. Best results of treating active choroidal neovascularizations in PXE are achieved when treatment starts the earliest possible.

Imaging geographic atrophy in age-related macular degeneration

Advances in retinal imaging technology have largely contributed to the understanding of the natural history, prognostic markers and disease mechanisms of geographic atrophy (GA) due to age-related macular degeneration. There is still no therapy available to halt or slow the disease process. In order to evaluate potential therapeutic effects in interventional trials, there is a need for precise quantification of the GA progression rate. Fundus autofluorescence imaging allows for accurate identification and segmentation of atrophic areas and currently represents the gold standard for evaluating progressive GA enlargement. By means of high-resolution spectral-domain optical coherence tomography, distinct microstructural alterations related to GA can be visualized.

Retinal pigment epithelial changes in chronic Vogt-Koyanagi-Harada disease: fundus autofluorescence and spectral domain-optical coherence tomography findings

PURPOSE

The purpose of this study was to determine whether fundus autofluorescence (FAF) and spectral domain-optical coherence tomography (SD-OCT) imaging allow better assessment of retinal pigment epithelium and the outer retina in subjects with chronic Vogt-Koyanagi-Harada disease compared with examination and angiography alone.

METHODS

A cross-sectional analysis of a series of seven consecutive patients with chronic Vogt-Koyanagi-Harada disease undergoing FAF and SD-OCT was conducted. Chronic disease was defined as duration of intraocular inflammation >3 months. Color fundus photographs were correlated to FAF and SD-OCT images. The images were later correlated to fluorescein angiography and indocyanine green angiography.

RESULTS

All patients had sunset glow fundus, which resulted in no apparent corresponding abnormality on FAF or SD-OCT. Lesions with decreased autofluorescence signal were observed in 11 eyes (85%), being associated with loss of the retinal pigment epithelium and involvement of the outer retina on SD-OCT. In 5 eyes (38%), some of these lesions were very subtle on clinical examination but easily detected by FAF. Lesions with increased autofluorescence signal were seen in 8 eyes (61.5%), showing variable involvement of the outer retina on SD-OCT and corresponding clinically to areas of retinal pigment epithelium proliferation and cystoid macular edema.

CONCLUSION

Combined use of FAF and SD-OCT imaging allowed noninvasive delineation of retinal pigment epithelium/outer retina changes in patients with chronic Vogt-Koyanagi-Harada disease, which were consistent with previous histopathologic reports. Some of these changes were not apparent on clinical examination.

CFH, C3 and ARMS2 are significant risk loci for susceptibility but not for disease progression of geographic atrophy due to AMD

Background

Age-related macular degeneration (AMD) is a prevalent cause of blindness in Western societies. Variants in the genes encoding complement factor H (CFH), complement component 3 (C3) and age-related maculopathy susceptibility 2 (ARMS2) have repeatedly been shown to confer significant risks for AMD; however, their role in disease progression and thus their potential relevance for interventional therapeutic approaches remains unknown.

Methodology/Principal Findings

Here, we analyzed association between variants in CFH, C3 and ARMS2 and disease progression of geographic atrophy (GA) due to AMD. A quantitative phenotype of disease progression was computed based on longitudinal observations by fundus autofluorescence imaging. In a subset of 99 cases with pure bilateral GA, variants in CFH (Y402H), C3 (R102G), and ARMS2 (A69S) are associated with disease (P = 1.6×10⁻⁹, 3.2×10⁻³, and P = 2.6×10⁻¹², respectively) when compared to 612 unrelated healthy control individuals. In cases, median progression rate of GA over a mean follow-up period of 3.0 years was 1.61 mm²/year with high concordance between fellow eyes. No association between the progression rate and any of the genetic risk variants at the three loci was observed (P>0.13).

Conclusions/Significance

This study confirms that variants at CFH, C3, and ARMS2 confer significant risks for GA due to AMD. In contrast, our data indicate no association of these variants with disease progression which may have important implications for future treatment strategies. Other, as yet unknown susceptibilities may influence disease progression.

Fundus autofluorescence and progression of age-related macular degeneration

Fundus autofluorescence imaging is an imaging method that provides additional information compared to conventional imaging techniques. It permits to topographically map lipofuscin distribution of the retinal pigment epithelial cell monolayer. Excessive accumulation of lipofuscin granules in the lysosomal compartment of retinal pigment epithelium cells represents a common downstream pathogenetic pathway in various hereditary and complex retinal diseases including age-related macular degeneration (AMD). This comprehensive review contains an introduction in fundus autofluorescence imaging, including basic considerations, the origin of the signal, different imaging methods, and a brief overview of fundus autofluorescence findings in normal subjects. Furthermore, it summarizes cross-sectional and longitudinal fundus autofluorescence findings in patients with AMD, addresses the pathophysiological significance of increased fundus autofluorescence, and characterizes different fundus autofluorescence phenotypes as well as fundus autofluorescence alterations with disease progression.

Interactive segmentation for geographic atrophy in retinal fundus images

Fundus auto-fluorescence (FAF) imaging is a non-invasive technique for in vivo ophthalmoscopic inspection of age-related macular degeneration (AMD), the most common cause of blindness in developed countries. Geographic atrophy (GA) is an advanced form of AMD and accounts for 12-21% of severe visual loss in this disorder [3]. Automatic quantification of GA is important for determining disease progression and facilitating clinical diagnosis of AMD. The problem of automatic segmentation of pathological images still remains an unsolved problem. In this paper we leverage the watershed transform and generalized non-linear gradient operators for interactive segmentation and present an intuitive and simple approach for geographic atrophy segmentation. We compare our approach with the state of the art random walker [5] algorithm for interactive segmentation using ROC statistics. Quantitative evaluation experiments on 100 FAF images show a mean sensitivity/specificity of 98.3/97.7% for our approach and a mean sensitivity/specificity of 88.2/96.6% for the random walker algorithm.

Evaluation of autofluorescence imaging with the scanning laser ophthalmoscope and the fundus camera in age-related geographic atrophy

PURPOSE

To compare fundus autofluorescence images (FAF) between a modified fundus camera (mFC) and a confocal scanning laser ophthalmoscope (cSLO).

DESIGN

Evaluation of diagnostic technology.

METHODS

Thirty-two eyes of 16 patients with age-related geographic atrophy (GA) treated in an institutional setting were included. FAF images were obtained with both the cSLO (excitation, 488 nm; emission, > 500 nm) and the mFC (excitation, approximately 500 to 610 nm; emission, approximately 675 to 715 nm). Using established algorithms, images were graded by two independent observers and agreements were evaluated. The main outcome measures were image quality, quantification of total atrophy, and classification of FAF patterns.

RESULTS

In two eyes with advanced cataract (lens grade 7 according to the Age-Related Eye Disease Study classification), FAF image quality with both systems was not sufficient for any meaningful analysis. In the remaining 30 eyes, the mean differences of the interobserver agreements for atrophy quantification were 0.16 mm2 (95% confidence interval [CI], 0.07 to 0.38) for mFC and 0.15 mm2 (95% CI, -0.04 to 0.33) for cSLO images. Because of inferior signal-to-noise ratios, FAF pattern classification was possible in a lower number of mFC images (69%) compared with cSLO images (88%).

CONCLUSIONS

This study suggests that the agreements for atrophy quantification are similar with both devices. The lesser visualization of FAF patterns with the mFC and thus inferior determination of disease markers may be the result of the nonconfocality and the use of single instead of mean images compared with the cSLO. These findings may be important for the design of interventional trials as well as the routine use of FAF imaging in age-related geographic atrophy.

Progression of geographic atrophy and impact of fundus autofluorescence patterns in age-related macular degeneration

PURPOSE

To test if fundus autofluorescence (FAF) patterns around geographic atrophy (GA) have an impact on GA progression rates over time in atrophic age-related macular degeneration (AMD).

DESIGN

Prospective longitudinal multicenter natural history study.

METHODS

Standardized digital FAF images were obtained from 195 eyes of 129 patients with GA using confocal scanning laser ophthalmoscopy (excitation 488 nm, emission >500 nm). Areas of GA were quantified and patterns of abnormal FAF in the junctional zone were classified. Repeated FAF images were obtained over a median follow-up period of 1.80 years (interquartile range [IQR], 1.28 to 3.34).

RESULTS

Areas of GA (median, 7.04 mm(2) at baseline; IQR, 3.12 to 10.0) showed a median enlargement of 1.52 mm(2)/year (IQR, 0.81 to 2.33). Progression rates in eyes with the banded (median 1.81 mm(2)/year) and the diffuse FAF pattern (1.77 mm(2)/year) were significantly higher compared to eyes without FAF abnormalities (0.38 mm(2)/year) and focal FAF patterns (0.81 mm(2)/year, P < .0001). Within the group of the diffuse pattern, eyes with a diffuse trickling pattern could be identified that exhibited an even higher spread rate (median 3.02 mm(2)/year) compared to the other diffuse types (1.67 mm(2)/year, P = .001).

CONCLUSIONS

The results indicate that distinct phenotypic FAF patterns have an impact on disease progression in eyes with atrophic AMD and may therefore serve as prognostic determinants. The findings underscore the relevance of FAF imaging and the pathogenetic role of excessive retinal pigment epithelium (RPE) lipofuscin (LF) accumulation in GA. Natural history data and identification of high-risk characteristics will be helpful to design interventional studies aiming at slowing the spread of atrophy.

Modelling the natural history of geographic atrophy in patients with age-related macular degeneration

PURPOSE

To model the natural course of geographic atrophy (GA) in patients with age-related macular degeneration (AMD).

METHODS

Data on the natural course of GA were collected in the multi-center, longitudinal, prospective observational FAM study. The size of GA was measured by autofluorescence scanning laser ophthalmoscopy. The natural course of GA is modelled by two different mixed effect models (MEM). Both models are compared with respect to the correctness of the model assumptions, goodness of fit, and predictive behavior.

RESULTS

The linear model results in better prediction, the non-linear model is more in agreement with the model assumptions. The non-linear model fits the data for small and large areas of GA better, while the linear model seems to be more adequate for the medial areas. More data will be needed to study the interplay of both models in more detail.

CONCLUSIONS

The natural course of GA varies extremely between individuals. However, reliable factors for the explanation of this variability have so far not been established. MEM are useful for describing "inter-individual" as well as "intra-individual" influences without the need for precise knowledge of the influencing factors. Using MEM to evaluate data on the natural history of GA allows one to derive parameter estimates, which could be used to design interventional trials for modes of therapy with a potential to reduce or stop the progression of GA in patients with AMD.

Retinal image analysis: concepts, applications and potential

As digital imaging and computing power increasingly develop, so too does the potential to use these technologies in ophthalmology. Image processing, analysis and computer vision techniques are increasing in prominence in all fields of medical science, and are especially pertinent to modern ophthalmology, as it is heavily dependent on visually oriented signs. The retinal microvasculature is unique in that it is the only part of the human circulation that can be directly visualised non-invasively in vivo, readily photographed and subject to digital image analysis. Exciting developments in image processing relevant to ophthalmology over the past 15 years includes the progress being made towards developing automated diagnostic systems for conditions, such as diabetic retinopathy, age-related macular degeneration and retinopathy of prematurity. These diagnostic systems offer the potential to be used in large-scale screening programs, with the potential for significant resource savings, as well as being free from observer bias and fatigue. In addition, quantitative measurements of retinal vascular topography using digital image analysis from retinal photography have been used as research tools to better understand the relationship between the retinal microvasculature and cardiovascular disease. Furthermore, advances in electronic media transmission increase the relevance of using image processing in 'teleophthalmology' as an aid in clinical decision-making, with particular relevance to large rural-based communities. In this review, we outline the principles upon which retinal digital image analysis is based. We discuss current techniques used to automatically detect landmark features of the fundus, such as the optic disc, fovea and blood vessels. We review the use of image analysis in the automated diagnosis of pathology (with particular reference to diabetic retinopathy). We also review its role in defining and performing quantitative measurements of vascular topography, how these entities are based on 'optimisation' principles and how they have helped to describe the relationship between systemic cardiovascular disease and retinal vascular changes. We also review the potential future use of fundal image analysis in telemedicine.