Pith. sign in

REVIEW 4 major objections 7 minor 45 references

A deep learning model for segmentation of geographic atrophy to study its long-term natural history

T0 review · 4 major / 7 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A deep learning model can segment geographic atrophy in color fundus images accurately enough to trace its long-term natural history, revealing that GA area grows quadratically up to about 12 mm² and then stabilizes or decreases.

desk verdict Solid segmentation model, but the natural-history claim depends on unvalidated transfer to AREDS—treat the 12 mm² plateau as suggestive until manual validation appears. read the letter →

arxiv 1908.05621 v1 pith:VXRFV5ZV submitted 2019-08-15 eess.IV cs.CV

Bart Liefers , Johanna M. Colijn , Cristina González-Gonzalo , Timo Verzijden , Paul Mitchell , Carel B. Hoyng , Bram van Ginneken , Caroline C.W. Klaver
show 1 more author
Clara I. Sánchez
This is my paper · ORCID
classification eess.IVcs.CV
keywords geographicatrophydeeplearningsegmentationcolorfundusimagingage-relatedmaculardegenerationdiseaseprogressionnaturalhistorystructuralbiomarkersensemblemodel
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper develops a fully automatic deep learning segmentation model for geographic atrophy (GA), the progressive death of retinal pigment epithelium in advanced age-related macular degeneration, and validates it against four expert graders on 409 color fundus images. The model reaches a Dice coefficient of 0.72 and an intraclass correlation of 0.83 for GA area, close to inter-grader agreement. Applied to 5,379 images from a large longitudinal dataset, the segmentations yield square-root annual growth rates across 584 eyes. The central natural-history claim is that GA area grows quadratically with time up to a lesion area of about 12 mm², after which growth rate stabilizes or decreases. The paper also identifies eight automatically computed structural features—area, filled area, convex area, convex solidity, eccentricity, roundness, foveal involvement, and perimeter—that are significantly associated with future growth rate.

What carries the argument

The central object is an ensemble of 20 encoder-decoder deep convolutional networks with residual blocks and shortcut connections, trained on consensus manual delineations from four graders in two population-based cohorts. Each network maps a color fundus image, together with a contrast-enhanced version, to a per-pixel likelihood of GA; ensemble predictions are combined after per-model threshold correction. The other load-bearing device is the square-root transformation of GA area: growth rate is measured as the slope of a linear regression through the square root of area over time, which removes the baseline-size dependence and lets the authors pool growth rates across eyes into a single curve.

What would settle it

Take a stratified sample of images from the target dataset spanning small and large lesions, have expert graders delineate GA, and compare automatic and manual areas per size bin; if the automatic-versus-manual bias changes systematically as lesions approach 12 mm², the plateau may be a measurement artifact rather than true natural history.

Watch

Extended reading notes

Core claim

The paper's central claim is that an ensemble of encoder-decoder deep networks can segment GA in color fundus images with accuracy approaching inter-grader agreement, and that the resulting segmentations, applied at scale to a longitudinal dataset, reveal a consistent natural-history pattern: square-root-transformed GA area grows roughly linearly (that is, area grows quadratically) while lesions are small, but the growth rate in mm² per year stops increasing and stabilizes or declines once the atrophic area reaches approximately 12 mm². The authors argue this pattern explains the dependence of growth rate on baseline area and appears both in cross-sectional pooling of growth rates and in individual eyes followed over many years.

Load-bearing premise

The model is trained on one set of color fundus images and applied to a second large dataset without any manual delineations on that second dataset, so segmentation errors that vary with lesion size or image quality could, in principle, create or mask the observed growth plateau.

Editorial extensions

If this is right

  • Fully automatic segmentation makes it practical to measure GA area and growth in thousands of eyes from standard color fundus images without manual grading.
  • If the plateau near 12 mm² is real, square-root annual growth rate cannot be treated as constant across the disease; clinical trials should stratify or adjust for baseline lesion size.
  • Eight automatically computed structural features predict future growth, offering a route to enrich clinical trials with fast-progressing eyes.
  • The quadratic-to-saturating growth curve provides a quantitative natural-history benchmark that future interventions can be compared against.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The pooled growth curve mixes many eyes at different disease stages, so the plateau near 12 mm² could reflect a cohort effect rather than a universal within-eye phase; tracking individual lesions through that size would directly test the claim.
  • If the plateau is confirmed in within-eye data, the standard square-root transformation, which assumes constant radial expansion, would need revision for large lesions because radial speed would appear to slow.
  • Applying the same segmentation pipeline to fundus autofluorescence or OCT images could test whether the plateau is specific to color fundus imaging or a true biological feature of atrophy growth.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. The manuscript presents an ensemble deep-learning model for automatic segmentation of geographic atrophy (GA) in color fundus images (CFIs), trained and validated on 409 images from the Rotterdam Study and Blue Mountains Eye Study, achieving a Dice coefficient of 0.72 ± 0.26 against consensus manual delineations, comparable to inter-grader Dice values of 0.72–0.82. The model is then applied to 5,379 CFIs from 625 eyes in the Age-Related Eye Disease Study (AREDS) to measure GA growth. The authors report that eight automatically extracted structural biomarkers are significantly associated with the square-root annual growth rate, and that combining all per-eye growth estimates suggests GA area grows quadratically up to about 12 mm², after which the growth rate stabilizes or decreases. The paper concludes that the model enables fully automatic GA segmentation and can support large-scale natural-history analyses.

Significance. If the findings hold, the paper makes two useful contributions. First, it provides a fully automatic deep-learning pipeline for GA segmentation in CFIs, with validation against multiple expert graders on two population-based cohorts; the model's performance is close to the inter-grader variability, which is a meaningful benchmark. Second, it demonstrates the feasibility of applying such a model to a large longitudinal dataset (AREDS) and reproduces known associations (e.g., multifocal and extrafoveal lesions growing faster), while proposing novel morphometric biomarkers. The central natural-history claim—a quadratic growth phase followed by a plateau—would be of substantial clinical and trial-design interest if it survives the additional validation described below. The paper also includes a fairly detailed model description in the appendix, which aids reproducibility, though code and trained models are not provided.

major comments (4)
  1. [Methods, 'GA growth rate' (first paragraph)] The model's segmentation accuracy on AREDS is never assessed; all reported performance metrics (Dice 0.72, ICC 0.83) come from RS/BMES. Because the growth-rate-versus-area curve in Figure 4 is derived entirely from automatic segmentations on AREDS, a size-dependent segmentation bias could create or exaggerate the apparent deceleration after ~12 mm². For example, under-segmentation of large, low-contrast, or confluent lesions would lower estimated growth rates for large lesions, while over-segmentation of small lesions would inflate early growth. The authors should validate the model on a manually graded AREDS subset (the AREDS reading-center gradings referenced in ref. 11 are an appropriate source) and report Dice and ICC stratified by lesion size and image quality, or at least compare automatic areas with available manual AREDS area measurements.
  2. [Methods, 'GA growth rate' (stereoscopic exclusion)] The exclusion of visits with more than 50% relative difference in automatically segmented GA area between stereoscopic images is applied without a reported rationale or a characterization of the 41 excluded eyes (625 enrolled, 584 analyzed). If stereoscopic discordance is more common for large or fast-growing lesions, the exclusion could selectively remove the large-area tail of the growth curve and contribute to the observed plateau. The authors should report the excluded eyes' baseline characteristics and test the robustness of Figure 4 to alternative thresholds (e.g., 30% and 70%).
  3. [Methods, 'GA growth rate' (last paragraph) and Figure 4] The claim that GA area grows quadratically up to ~12 mm² and then stabilizes is based on pooling per-eye 2-year growth slopes across eyes and fitting a quadratic curve with a data-dependent cutoff at 12 mm². This procedure does not demonstrate that individual eyes follow this trajectory; eyes reaching large areas are a selected subset, and cross-sectional pooling can mask heterogeneous individual paths. The authors should analyze individual longitudinal data with a prespecified mixed-effects model (e.g., including linear and quadratic time terms with a changepoint) and report how many eyes cross the 12 mm² threshold and their growth patterns.
  4. [Methods, 'Data' (pixel-to-mm conversion) and Discussion] The fixed pixel-to-millimeter conversion based on an assumed 4.5 mm fovea-to-disc distance affects all area and growth estimates, including the 12 mm² threshold. The authors acknowledge this limitation but do not quantify its impact. A sensitivity analysis with alternative conversion factors (e.g., ±10%) should be reported, and if possible the authors should calibrate magnification per image using the measured fovea-to-disc distance.
minor comments (7)
  1. [Abstract, 'Design' line] The study is described as 'Prospective, multicenter, natural history study,' but the model development and validation are retrospective analyses of existing cohort data; the design should be described as a retrospective analysis of prospective cohorts.
  2. [Abstract and Methods] The participant counts (409 CFIs for development, 5,379 CFIs for analysis) are reported inconsistently with the evaluation N=315 in Results; the distinction between images and unique visits should be stated consistently.
  3. [Introduction and reference 21] The claim that this is 'the first deep learning model for segmentation of GA in CFI' should be reconciled with reference 21 (Keenan et al.), which describes a deep-learning approach for automated detection of GA from color fundus photographs; if that work also performs segmentation, the novelty claim should be narrowed.
  4. [Table 3 and appendix] The feature 'Fovea region' in Table 3 corresponds to 'foveal involvement' in the appendix; the naming should be unified.
  5. [Figure 4] The right panel's 'evolution of GA area over time' is obtained by numerical integration starting from an assumed area of 0.5 mm² at t=0; the dependence of the resulting curve on this initial condition and on the integration method should be stated.
  6. [Methods, 'GA growth rate'] The window selection criterion ('for which the number of available CFIs was highest for the respective eye') should specify how ties are resolved and how many timepoints were typically available in the chosen window.
  7. [General] The paper should state whether the trained model or code is publicly available; this would enhance reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity identified: the segmentation model is trained on independent manual delineations and the growth analysis is empirical curve fitting, not a prediction derived from the fitted parameters.

full rationale

The paper's derivation chain is self-contained rather than circular. The deep learning model is trained and cross-validated on manual consensus delineations from RS and BMES, which are external to the AREDS growth analysis. The central natural-history result is obtained by applying the trained ensemble to AREDS images, measuring GA area, computing per-eye growth rates via linear regression over two-year windows, and then combining these growth estimates as a function of area (Methods: 'GA growth rate'). The quadratic curve in Figure 4 is explicitly described as a best fit to the observed data ('The red dashed line in these graphs represent a quadratic model that best fitted the data for GA area < 12 mm²'), not as an independent prediction derived from the model inputs. Fitting a curve to data and reporting that it fits is empirical summarization, not circularity. No load-bearing argument relies on self-citation: the cited EyeNED workstation is only a tool for manual annotation, and the square-root transformation is justified by external references (Feuer et al., Grunwald et al.). The potential concern that the model was not manually validated on AREDS is a question of external validity or measurement bias, not a circularity of the kind where an output reduces by construction to an input. Therefore the paper receives a circularity score of 0.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central natural history claim rests on several domain assumptions about GA identification, the consensus reference standard, square-root linear growth, the fixed pixel-to-mm conversion, and generalization of the model to AREDS. The first three are standard in the field and supported by references; the last two are load-bearing and could bias the reported growth rates if inaccurate. No new physical entities are introduced. The free parameters are the thresholds and conversions chosen by the authors for the growth analysis and segmentation post-processing.

free parameters (5)
  • Pixel-to-millimeter conversion distance = 4.5 mm
    Used to convert pixel areas to mm². Assumed constant across all images based on average fovea-optic disc distance; directly scales all area and growth-rate values. Acknowledged as a limitation in the Discussion.
  • Stereoscopic difference threshold = 50%
    Visits with >50% relative difference in automatically segmented GA area between left and right stereoscopic images were discarded from the growth analysis. This post hoc exclusion threshold is not justified and could bias the sample.
  • Growth regression window = 2 years
    Square-root and linear growth rates were estimated from CFIs within a 2-year window; this assumes constant growth within that interval and affects all growth-rate estimates.
  • Quadratic fit cutoff for GA area = ~12 mm²
    The claim that GA area grows quadratically up to ~12 mm² is based on fitting a quadratic to the model-derived growth curve; the cutoff is inferred from the same data, not from an independent prediction.
  • Optimal segmentation threshold per model = Selected per model on validation set
    The binary segmentation mask is obtained by thresholding the model's likelihood output; the optimal threshold for each model is chosen to maximize Dice on the validation set and is used in the ensemble, so final predictions depend on these fitted thresholds.
assumptions (5)
  • domain assumption GA is identified on CFI as absence of retinal pigment epithelium and increased visibility of choriocapillaris; manual delineations from four graders are treated as ground truth.
    The entire training and validation depends on this operational definition of GA.
  • domain assumption Consensus grading by four graders is an appropriate reference standard for model evaluation, despite not being independent of individual graders.
    The paper notes consensus is not independent but uses it as the reference for Dice and ICC.
  • domain assumption Square root of GA area grows approximately linearly over time within short intervals, for growth rate measurement.
    Used to compute annual growth; based on prior literature (refs 16, 26).
  • domain assumption The average distance between the fovea and the optic disc center is 4.5 mm across all images.
    Underlies the pixel-to-mm conversion for all area and growth values.
  • domain assumption The model trained on RS/BMES CFIs produces valid segmentations on AREDS CFIs acquired with different cameras and populations.
    The natural history analysis uses AREDS segmentations without manual validation on that dataset; systematic domain shift could bias results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A deep learning model for segmentation of geographic atrophy to study its long-term natural history." pith.science (2026). https://pith.science/paper/VXRFV5ZV

@misc{pith2026190805621,
  author       = {Pith},
  title        = {Pith review of: A deep learning model for segmentation of geographic atrophy to study its long-term natural history},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VXRFV5ZV}},
  note         = {Machine review of arXiv:1908.05621}
}
abstract

Purpose: To develop and validate a deep learning model for automatic segmentation of geographic atrophy (GA) in color fundus images (CFIs) and its application to study growth rate of GA. Participants: 409 CFIs of 238 eyes with GA from the Rotterdam Study (RS) and the Blue Mountain Eye Study (BMES) for model development, and 5,379 CFIs of 625 eyes from the Age-Related Eye Disease Study (AREDS) for analysis of GA growth rate. Methods: A deep learning model based on an ensemble of encoder-decoder architectures was implemented and optimized for the segmentation of GA in CFIs. Four experienced graders delineated GA in CFIs from RS and BMES. These manual delineations were used to evaluate the segmentation model using 5-fold cross-validation. The model was further applied to CFIs from the AREDS to study the growth rate of GA. Linear regression analysis was used to study associations between structural biomarkers at baseline and GA growth rate. A general estimate of the progression of GA area over time was made by combining growth rates of all eyes with GA from the AREDS set. Results: The model obtained an average Dice coefficient of 0.72 $\pm$ 0.26 on the BMES and RS. An intraclass correlation coefficient of 0.83 was reached between the automatically estimated GA area and the graders' consensus measures. Eight automatically calculated structural biomarkers (area, filled area, convex area, convex solidity, eccentricity, roundness, foveal involvement and perimeter) were significantly associated with growth rate. Combining all growth rates indicated that GA area grows quadratically up to an area of around 12 mm$^{2}$, after which growth rate stabilizes or decreases. Conclusion: The presented deep learning model allowed for fully automatic and robust segmentation of GA in CFIs. These segmentations can be used to extract structural characteristics of GA that predict its growth rate.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 44 canonical work pages

  1. [1]

    Age-related macular degeneration

    Lim LS, Mitchell P, Seddon JM, et al . Age-related macular degeneration. Lancet. 2012;379:1728-1738

    work page 2012

  2. [2]

    Visual function abnormalities and prognosis in eyes with age -related geographic atrophy of the macula and good visual acuity

    Sunness JS, Rubin GS, Applegate CA, et al. Visual function abnormalities and prognosis in eyes with age -related geographic atrophy of the macula and good visual acuity. Ophthalmology. 1997;104:1677-1691

    work page 1997

  3. [3]

    The estimated prevalence and incidence of late stage age related macular degeneration in the UK

    Owen CG, Jarrar Z, Wormald R, et al. The estimated prevalence and incidence of late stage age related macular degeneration in the UK. Br J Ophthalmol . 2012;96:752-756

    work page 2012

  4. [4]

    Global prevalence of age-related macular degeneration and disease burden projection for 2020 and 2040: a systematic review and meta- analysis

    Wong WL, Su X, Li X, et al. Global prevalence of age-related macular degeneration and disease burden projection for 2020 and 2040: a systematic review and meta- analysis. Lancet Glob Health. 2014;2:e106-116

    work page 2020

  5. [5]

    Prevalence of age -related macular degeneration in Europe: the past and the fut ure

    Colijn JM, Buitendijk GH, Prokofyeva E, et al. Prevalence of age -related macular degeneration in Europe: the past and the fut ure. Ophthalmology. 2017;124:1753- 1763

    work page 2017

  6. [6]

    Age-related macular degeneration— emerging pathogenetic and therapeutic concepts

    Gehrs KM, Anderson DH, Johnson LV, et al. Age-related macular degeneration— emerging pathogenetic and therapeutic concepts. Ann Med. 2006;38:450-471

    work page 2006

  7. [7]

    The pathophysiology of geographic atrophy secondary to age -related macular degeneration and the complement pathway as a therapeutic target

    Boyer DS, Schmidt -Erfurth U, van Lookeren Campagne M, et al. The pathophysiology of geographic atrophy secondary to age -related macular degeneration and the complement pathway as a therapeutic target. Retina. 2017;37:819

    work page 2017

  8. [8]

    Current therapeutic developments in atrophic age - related macular degeneration

    Hanus J, Zhao F, Wang S. Current therapeutic developments in atrophic age - related macular degeneration. Br J Ophthalmol. 2016;100:122-127

    work page 2016

Show all 45 references
  1. [9]

    Geographic atrophy: clinical features and potential therapeutic approaches

    Holz FG, Strauss EC, Schmitz -Valckenberg S, et al. Geographic atrophy: clinical features and potential therapeutic approaches. Ophthalmology. 2014;121:1079 - 1091

    2014

  2. [10]

    Designin g clinical trials for age - related geographic atrophy of the macula: enrollment data from the geographic atrophy natural history study

    Sunness JS, Applegate CA, Bressler NM, et al. Designin g clinical trials for age - related geographic atrophy of the macula: enrollment data from the geographic atrophy natural history study. Retina. 2007;27:204-210

    2007

  3. [11]

    Change in area of geographic atrophy in the Age -Related Eye Disease Study: AREDS report number 26

    Lindblad AS, Lloyd PC, Clemons TE, et al. Change in area of geographic atrophy in the Age -Related Eye Disease Study: AREDS report number 26. Arch Ophthalmol. 2009;127:1168-1174

    2009

  4. [12]

    The progression of geographic atrophy secondary to age -related macular degeneration

    Fleckenstein M, Mitchell P, Freund KB, et al. The progression of geographic atrophy secondary to age -related macular degeneration. Ophthalmology. 2018;125:369-390

    2018

  5. [13]

    Natural history of geographic atrophy progression secondary to age-related macular degeneration (Geographic Atrophy Progression Study)

    Schmitz-Valckenberg S, Sahel J, Danis R, et al. Natural history of geographic atrophy progression secondary to age-related macular degeneration (Geographic Atrophy Progression Study). Ophthalmology. 2016;123:361-368

    2016

  6. [14]

    Geographic atrophy in patients with advanced dry age -related macular degeneration: current challenges and future prospects

    Danis RP, Lavine JA, Domalpally A. Geographic atrophy in patients with advanced dry age -related macular degeneration: current challenges and future prospects. Clin Ophthalmol. 2015;9:2159

    2015

  7. [15]

    Measuring geographic atrophy in advanced age -related macular degen eration

    Sunness JS, Bressler NM, Tian Y, et al. Measuring geographic atrophy in advanced age -related macular degen eration. Invest Ophthalmol Vis Sci . 1999;40:1761-1769

    1999

  8. [16]

    Progression of geographic atrophy in age-related macular degeneration imaged with spectral domain optical coherence tomography

    Yehoshua Z, Rosenfeld PJ, Gregori G, et al. Progression of geographic atrophy in age-related macular degeneration imaged with spectral domain optical coherence tomography. Ophthalmology. 2011;118:679-686

    2011

  9. [17]

    A survey on deep learning in medical image analysis

    Litjens G, Kooi T, Bejnordi BE, et al. A survey on deep learning in medical image analysis. Med Image Anal. 2017;42:60-88

    2017

  10. [18]

    Automated grading of age -related macular degeneration from color fundus images using deep c onvolutional neural networks

    Burlina PM, Joshi N, Pekala M, et al. Automated grading of age -related macular degeneration from color fundus images using deep c onvolutional neural networks. JAMA Ophthalmol. 2017;135:1170-1176

    2017

  11. [19]

    DeepSeeNet: a deep learning model for automated classification of patient -based age -related macular degeneration severity from color fundus photographs

    Peng Y, Dharssi S, Chen Q, et al. DeepSeeNet: a deep learning model for automated classification of patient -based age -related macular degeneration severity from color fundus photographs. Ophthalmology. 2019;126:565-575

    2019

  12. [20]

    Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs

    Gulshan V, Peng L, Coram M, et al. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA. 2016;316:2402-2410

    2016

  13. [21]

    A d eep learning approach for automated detection of geographic atrophy from color fundus photographs

    Keenan T, Dharssi S, Peng Y, et al. A d eep learning approach for automated detection of geographic atrophy from color fundus photographs. Ophthalmology. 2019;In press

    2019

  14. [22]

    The epidemiology of progression of pure geographic atrophy: the Beaver Dam Eye Study

    Klein R, Meuer SM, Knudtson MD, et al. The epidemiology of progression of pure geographic atrophy: the Beaver Dam Eye Study. Am J Ophthalmol. 2008;146:692- 699

    2008

  15. [23]

    Circularity index as a risk factor for progression of geographic atrophy

    Domalpally A, Danis RP, White J, et al. Circularity index as a risk factor for progression of geographic atrophy. Ophthalmology. 2013;120:2666-2671

    2013

  16. [24]

    The long -term natural history of geographic atrophy from age -related macular degeneration: enlargement of atrophy and implications for interventional clinical trials

    Sunness JS, Margalit E, Srikumaran D, et al. The long -term natural history of geographic atrophy from age -related macular degeneration: enlargement of atrophy and implications for interventional clinical trials. Ophthalmology. 2007;114:271-277

    2007

  17. [25]

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

    Holz FG, Bindewald -Wittich A, Fleckenstein M, et al. Progression of geographic atrophy and impact of fundus autofluorescence patterns in age -related macular degeneration. Am J Ophthalmol. 2007;143:463-472

    2007

  18. [26]

    Feuer WJ, Yehoshua Z, Gregori G, et al. Square root transformation of geographic atrophy area measurements to eliminate dependence of growth rates on baseline lesion measurements: a reanalysis of age -related eye disease study report no

  19. [27]

    2013;131:110-111

    JAMA Ophthalmol. 2013;131:110-111

    2013

  20. [28]

    Natural history of geographic atrophy in untreated eyes with nonexudative age -related macular degeneration: a systematic review and meta-analysis

    Shen L, Liu F, Nardini HG, et al. Natural history of geographic atrophy in untreated eyes with nonexudative age -related macular degeneration: a systematic review and meta-analysis. Ophthalmol Retina. 2018;2:914-921

    2018

  21. [29]

    Fundus autofluorescence and spectral -domain optical coherence tomography characteristics in a rapi dly progressing form of geographic atrophy

    Fleckenstein M, Schmitz -Valckenberg S, Martens C, et al. Fundus autofluorescence and spectral -domain optical coherence tomography characteristics in a rapi dly progressing form of geographic atrophy. Invest Ophthalmol Vis Sci. 2011;52:3761-3766

    2011

  22. [30]

    Evaluation of geographic atrophy from color photographs and fundus autofluorescence images: Age -Related Eye Disease Study 2 Report Number 11

    Domalpally A, Danis R, Agrón E, et al. Evaluation of geographic atrophy from color photographs and fundus autofluorescence images: Age -Related Eye Disease Study 2 Report Number 11. Ophthalmology. 2016;123:2401-2407

    2016

  23. [31]

    Validated automatic segmentation of AMD pathology including drusen and geographic atrophy in SD -OCT images

    Chiu SJ, Izatt JA, O'Connell RV, et al. Validated automatic segmentation of AMD pathology including drusen and geographic atrophy in SD -OCT images. Invest Ophthalmol Vis Sci. 2012;53:53-61

    2012

  24. [32]

    Segmentation of the geographic atrophy in spectral -domain optical coherence tomography and fundus autofluorescence images

    Hu Z, Medioni GG , Hernandez M, et al. Segmentation of the geographic atrophy in spectral -domain optical coherence tomography and fundus autofluorescence images. Invest Ophthalmol Vis Sci. 2013;54:8375-8383

    2013

  25. [33]

    Automated geographic atrophy segmentation for SD -OCT images using region -based CV model via local similarity factor

    Niu S, de Sisternes L, Chen Q, et al. Automated geographic atrophy segmentation for SD -OCT images using region -based CV model via local similarity factor. Biomed Opt Express. 2016;7:581-600

    2016

  26. [34]

    Automated segmentation of geographic atrophy in fundus autofluorescence images using supervised pixel classification

    Hu Z, Medioni GG, Hernandez M, et al. Automated segmentation of geographic atrophy in fundus autofluorescence images using supervised pixel classification. J Med Imaging. 2015;2:014501

    2015

  27. [35]

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

    Feeny AK, Tadarati M, Freund DE, et al. Automated segmentation of geographic atrophy of the retinal epithelium via random forests in AREDS color fundus images. Comput Biol Med. 2015;65:124-136

    2015

  28. [36]

    Prevalence of age -related maculopathy in Australia

    Mitchell P, Smith W, Attebo K, et al. Prevalence of age -related maculopathy in Australia. Ophthalmology. 1995;102:1450-1460

    1995

  29. [37]

    The Rotterdam Study: 2018 update on objectives, design and main results

    Ikram MA, Brusselle GG, Murad SD, et al. The Rotterdam Study: 2018 update on objectives, design and main results. Eur J Epidemiol. 2017;32:807-850

    2018

  30. [38]

    Growth of geographic atrophy in the comparison of age -related macular degeneration treatments trials

    Grunwald JE, Pistilli M, Ying G, et al. Growth of geographic atrophy in the comparison of age -related macular degeneration treatments trials. Ophthalmology. 2015;122:809-816

    2015

  31. [39]

    EyeNED workstation: development of a multi-modal vendor-independent application for annotation, spatial alignment and analysis of retinal images

    van Zeeland H, Meakin J, Liefers B, et al. "EyeNED workstation: development of a multi-modal vendor-independent application for annotation, spatial alignment and analysis of retinal images", in: Association for Research in Vision and Ophthalmology, 2019

    2019

  32. [40]

    Clinical ly applicable deep learning for diagnosis and referral in retinal disease

    De Fauw J, Ledsam JR, Romera -Paredes B, et al. Clinical ly applicable deep learning for diagnosis and referral in retinal disease. Nat Med. 2018;24:1342

    2018

  33. [41]

    Kaggle diabetic retinopathy detection competition report

    Graham B. Kaggle diabetic retinopathy detection competition report. University of Warwick. 2015

    2015

  34. [42]

    Progression of geogra phic atrophy in age-related macular degeneration: AREDS2 report number 16

    Keenan TD, Agron E, Domalpally A, et al. Progression of geogra phic atrophy in age-related macular degeneration: AREDS2 report number 16. Ophthalmology. 2018;125:1913-1928

    2018

  35. [43]

    Circularity, solidity, axes of a best fit ellipse, aspect ratio, and roundness of the foramen ovale: a morphometric analysis with neurosurgical considerations

    Zdilla MJ, Hatfield SA, McLean KA, et al. Circularity, solidity, axes of a best fit ellipse, aspect ratio, and roundness of the foramen ovale: a morphometric analysis with neurosurgical considerations. J Craniofac Surg. 2016;27:222. Table 1: Dice coefficients between model and...

    2016

  36. [44]

    et al., Keras, https://keras.io, 2015

    Chollet, F. et al., Keras, https://keras.io, 2015

    2015

  37. [45]

    Abadi, M. et al. TensorFlow: Large-scale machine learning on heterogeneous systems, https://www.tensorflow.org, 2015 Description of lesion characteristics Area: The total area of all segmented GA. Perimeter: The total perimeter along the border of the segmented area. If lesion...

    2015

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.