About

Amitha Domalpally, MD, PhD, is an Assistant Professor and the Scientific Director at the A-EYE Research Unit at the University of Wisconsin–Madison. Her role involves leading research efforts in the field of artificial intelligence for eye-related applications. She is part of a team that focuses on advancing AI technologies to improve eye health and ophthalmology.

Research topics

• Medicine
• Ophthalmology
• Internal medicine
• Optometry
• Surgery
• Pediatrics
• Endocrinology
• Psychology

Selected publications

• Modest and Variable Correlations Between Geographic Atrophy Enlargement Rates in Fellow Eyes in the AREDS2 Study

  JAMA Ophthalmology · 2026-04-02 · 1 citations

  articleOpen access

  Importance: In bilateral geographic atrophy (GA), enlargement rates in fellow eyes are often assumed to be highly correlated. On this basis, researchers have inferred treatment effects and proposed GA trials using the untreated fellow eye as an internal control. Objective: To quantify the correlation in GA enlargement rates between fellow eyes. Design, Setting, and Participants: This was a post hoc analysis of Age-Related Eye Disease Study 2 (AREDS2), a multicenter study of patients in retinal specialty clinics throughout the US. Included in the analysis were participants of the AREDS2 trial with bilateral GA. Study data were analyzed from May 2025 to January 2026. Exposures: GA in the fellow eye. Main Outcomes and Measures: Pearson correlation coefficient for 2-year GA enlargement rates in eye pairs. Rates were derived from planimetry of annual fundus photographs and expressed as untransformed, square root-transformed, and perimeter-adjusted rates. Correlation was analyzed overall and within clinically relevant strata. Results: A total of 386 eyes of 193 AREDS2 participants (mean [SD] age, 75.5 [7.3] years; 118 female [61.1%]) with bilateral GA were included in this analysis. Correlations in enlargement rates varied substantially by transformation used as follows: moderate for untransformed (r = 0.51; 95% CI, 0.41-0.61), weak to moderate for square root-transformed (r = 0.38; 95% CI, 0.25-0.49), and very weak for perimeter-adjusted (r = 0.11; 95% CI, -0.03 to 0.25) rates. Correlation was consistently weaker in large vs small GA: 0.30 (95% CI, 0.07-0.50) vs 0.63 (95% CI, 0.41-0.78; untransformed), 0.38 (95% CI, 0.16-0.56) vs 0.46 (95% CI, 0.20-0.66; square root), and 0.03 (95% CI, -0.21 to 0.26) vs 0.22 (-0.08 to 0.49; perimeter), respectively. For GA location using untransformed rates, correlation was strongest for extrafoveal GA, intermediate for subfoveal GA, and weakest for discordant pairs. Square root rates showed a similar pattern. For focality using untransformed rates, correlation was similar for unifocal, multifocal, and discordant pairs. Using square root rates, it was strongest for unifocal GA, weakest for multifocal GA, and intermediate for discordant pairs. For reticular pseudodrusen (RPD) status, correlation was strongest for RPD absence, weakest for presence, and intermediate for discordant pairs. Differences were smaller using square root rates. Conclusions and Relevance: Results of this post hoc analysis of the AREDS2 trial reveal that correlation in GA enlargement rates between fellow eyes was modest. Key GA characteristics of area, location, focality, and RPD status, as well as the transformation used, were strongly associated with the correlation. Untransformed rates were inflated by a tendency for symmetry in baseline GA characteristics, whereas genuine biological correlation (best reflected by linear enlargement) was only weak to moderate. Researchers should be cautious of trials and analyses relying on assumptions of highly correlated enlargement rates.

  PublisherOA PDFDOI

• Comparison of Clarus, Optos and Heidelberg Systems for Geographic Atrophy Area Measurements (COCO GA)

  Ophthalmology Science · 2026-04-01

  articleOpen accessSenior author
  PublisherDOI

• Reference Standard for Validation of Age-Related Macular Degeneration Screening Algorithms

  Ophthalmology · 2026-04-01

  articleOpen access1st authorCorresponding

  PURPOSE: Artificial intelligence (AI)-based screening models hold promise for identifying individuals with undiagnosed age-related macular degeneration (AMD) in nonspecialist settings. A standardized reference framework for image labeling is needed to enable consistent training, validation, and deployment of AI-based screening algorithms. The goal of the present study was to establish expert consensus on an image-based reference standard for labeling AMD. DESIGN: Modified Delphi consensus study. PARTICIPANTS: Fellowship-trained retina specialists, ophthalmologists, AI specialists, and imaging specialists. METHODS: A prespecified Delphi process was conducted using structured surveys. Over 2 rounds, panelists assessed opinions on existing reference standards, including the Age-Related Eye Disease Study scale and Beckman scale, as well as imaging methods such as color, OCT, and autofluorescence. The surveys also evaluated imaging features of AMD, including drusen, pseudodrusen, and pigment changes, as well as referral criteria. Consensus was defined using a 9-point Likert scale, with predefined statistical thresholds for agreement. MAIN OUTCOME MEASURES: Agreement on key elements of a reference standard. RESULTS: Consensus was reached on adopting the Beckman classification as the level 1 reference standard (median score, 8; agreement). OCT use for identifying key AMD features, including drusen, geographic atrophy (GA), and choroidal neovascularization, also reached consensus (median scores, 8.5-9; agreement). Pigment change detection did not reach consensus (median, 7.5; uncertain), and screening age thresholds showed nonconsensus (median, 8; uncertain). Referral thresholds reached consensus, including urgent referral for neovascular AMD and nonurgent referral for GA and intermediate AMD (median, 9; agreement). CONCLUSIONS: This study defined a consensus-based reference standard for labeling AMD from images for AI-based screening. These recommendations are intended to support consistent AI model development and evaluation, while remaining distinct from clinical practice guidelines. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

  PublisherDOI

• Geographic Atrophy Multifocality in the Age-Related Eye Disease Study 2

  JAMA Ophthalmology · 2025-11-06 · 2 citations

  articleOpen access

  Importance: Geographic atrophy (GA) progression represents the combination of contiguous expansion and new lesion addition. Multifocal GA is associated with faster area-based progression and has been linked to faster decline in visual acuity, but the natural history and risk factors for new lesion addition are unknown. Objective: To examine, in eyes with preexisting GA, the natural history of new GA lesion incidence and identify risk factors for faster incidence. Design, Setting, and Participants: This was a post hoc analysis of the Age-Related Eye Disease Study 2 (AREDS2), which was a multicenter study based in retinal specialty clinics in the US. Included in the analysis were eyes with incident GA considered in 2 cohorts: (1) any-focality GA (eyes with unifocal or multifocal GA at incidence) and (2) unifocal GA (eyes with unifocal GA at incidence). Study data were analyzed from January to May 2025. Exposures: Demographic and dietary characteristics, genotype, GA characteristics, and macular features. Main Outcomes and Measures: The primary outcome was new GA lesion incidence rate, after GA incidence. Annual fundus photographs were graded for GA presence and characteristics and macular features. After GA incidence, new GA lesion incidence events were recorded. Univariable and multivariable analyses were performed by generalized estimating equations regression, with least absolute shrinkage and selection operator regression for variable screening. Results: A total of 689 eyes with incident GA of 570 participants (mean [SD] age, 74.5 [6.9] years; 332 female [58.2%]) were included in this study. Eyes constituted 2 main cohorts: (1) any-focality GA (ie, 689 eyes with unifocal or multifocal GA at incidence) and (2) unifocal GA (ie, 386 eyes with unifocal GA at incidence). Over a mean (SD) follow-up of 2.1 (0.9) years, the mean (SD) GA lesion incidence rate was 0.25 (0.50) lesions per year (range, 0-3 lesions per year) and 0.40 (0.70) lesions per year (range, 0-5 lesions per year) in the unifocal and any-focality cohorts, respectively. For the unifocal cohort, the multivariable model comprised 4 risk factors: reticular pseudodrusen (RPD), noncentral GA, lower nut intake, and higher calorie intake. For the any-focality cohort, the model comprised 7 risk factors: GA lesion number, noncentral GA, greater GA proximity to central macula, RPD, age, lower fish intake, and higher calorie intake. Conclusions and Relevance: In this post hoc analysis of the AREDS2 trial data, the natural history data and multivariable models may be helpful in clinical practice and trials. The models included a combination of GA characteristics and RPD status; the risk factors for first GA occurrence, subsequent lesion incidence rate, and area-based progression overlapped only partially. Their mechanisms may, therefore, be partially distinct, which has implications for therapeutic approaches.

  PublisherOA PDFDOI

• Dietary intake and an atherogenic dietary pattern in relation to retinal vessel caliber 15-years later in the carotenoids in age-related eye disease study

  Clinical Nutrition · 2025-08-06

  article
  PublisherDOI

• Risk Factors for 15-Letter Visual Acuity Loss from Geographic Atrophy Progression over 1 Year in the Age-Related Eye Disease Study 2

  Ophthalmology Retina · 2025-09-11

  articleOpen access
  PublisherOA PDFDOI

• Long-Term Effects of Anti-VEGF Therapy versus Panretinal Photocoagulation on Retinal Vessel Caliber in Eyes with Proliferative Diabetic Retinopathy

  Ophthalmology Retina · 2025-04-10

  articleOpen access
  PublisherOA PDFDOI

• Relationships Between Diet and Geographic Atrophy Progression in the Age-Related Eye Diseases Studies 1 and 2

  Nutrients · 2025-02-22 · 7 citations

  articleOpen access

  Background/Objectives: The objective of this study was to analyze the relationships between diet and geographic atrophy (GA) progression, both area-based and proximity-based, for dietary pattern, components, and micronutrients. Methods: In the Age-Related Eye Diseases Study (AREDS) and AREDS2, an Alternative Mediterranean Diet Index (aMedi), its nine components, and individual micronutrient intakes were calculated. Mixed-model regression was performed for square root GA area, GA foveal proximity, and acuity. Results: The study populations comprised 657 (AREDS) and 1179 eyes (AREDS2). For area-based progression, a higher aMedi was associated with slower progression in AREDS2 and (in analyses excluding MUFA:SFA) AREDS. A higher intake was associated with slower progression for seven components (including vegetables and fruit at Bonferroni) and four components (including fruit and less red meat at Bonferroni), and seven and 15 nutrients, in AREDS1/2, respectively. For proximity-based progression, a higher aMedi was associated with slower progression in AREDS. A higher intake was associated with slower progression for three components (including vegetables at Bonferroni) and two components, and 10 and 8 nutrients, in AREDS1/2, respectively. With increasing oral supplementation, associations between proximity-based progression and aMedi/components/nutrients were weaker. In AREDS2 eyes with non-central GA, higher aMedi was associated with a slower acuity decline. Conclusions: A Mediterranean-type diet is associated with slower GA area-based progression and slower progression to the fovea, accompanied by a slower decline in acuity. The most important components and micronutrients for incidence, area-based progression, and foveal progression overlap only partially. For the latter two, they include vegetables, fruit, and less red meat. These findings suggest the benefits of targeted nutritional and supplementation strategies.

  PublisherOA PDFDOI

• Retinal imaging in an era of open science and privacy protection

  Experimental Eye Research · 2025-03-14

  reviewOpen access

  Artificial intelligence (AI) holds great promise for analyzing complex data to advance patient care and disease research. For example, AI interpretation of retinal imaging may enable the development of noninvasive retinal biomarkers of systemic disease. One potential limitation, however, is government regulation regarding retinal imaging as biometric data, which has been recently under debate in the United States. Although careful regard for patient privacy is key to maintaining trust in the widespread use of AI in healthcare, the designation of retinal imaging as biometric data would greatly restrict retinal biomarker research. There are several reasons why retinal imaging should not be considered biometric data. Unlike images of the iris, high quality images of the retina are more difficult to obtain, requiring specialized training and equipment, and often requiring pupil dilation for optimal quality. In addition, retinal imaging features can vary over time with changes in health status, and retinal images are not currently linked to any large identification databases. While the protection of patient privacy is imperative, there is also a need for large retinal imaging datasets to advance AI research. Given the limitations of retinal imaging as a source of biometric data, the research community should work to advocate for the continued use of retinal imaging in AI research. • Advances in artificial intelligence (AI)-based health research depend on large data. • AI analysis of retinal imaging may enable retinal biomarkers of systemic disease. • Designating retinal imaging as biometric data may limit available data for AI models. • Retinal imaging is less well-suited for biometric identification than the iris. • Researchers should advocate for the use of retinal imaging data for AI research.

  PublisherDOI

• Longitudinal Comparison of Geographic Atrophy Enlargement Using Manual, Semiautomated, and Deep Learning Approaches

  Ophthalmology Science · 2025-04-08 · 3 citations

  articleOpen accessSenior author

  Objective: To compare a fully automated artificial intelligence (AI) model, a semiautomated method, and manual planimetry in the longitudinal assessment of geographic atrophy (GA) using fundus autofluorescence images.Design: A retrospective analysis of 3 GA assessment methods: AI, Heidelberg Eye Explorer semiautomated software (RegionFinder), and manual planimetry.Subjects and Controls: One hundred eight patients (185 eyes) with GA from a phase IIb clinical trial by GlaxoSmithKline, which evaluated an experimental drug that did not reduce GA enlargement compared with the placebo.Methods: Fundus autofluorescence images of 185 eyes were annotated using manual planimetry, semiautomated RegionFinder, and a fully automated AI model trained and validated on manual planimetry annotations at screening, year 1, and year 2. Artificial intelligence masks were compared with human-guided methods, and regression errors were assessed by stacking masks from consecutive visits.Agreement between methods was assessed using BlandAltman plots, Dice similarity coefficient (DSC), and comparisons of GA growth rates.Artificial intelligence performance was evaluated based on its need for human edits and frequency of regression errors.Main Outcome Measures: Agreement between methods was evaluated using BlandAltman plots, DSC, and intraclass correlation coefficients (ICCs).The mean GA growth rate (mm 2 /year) and square root transformation of GA size were compared across methods.Artificial intelligence performance was assessed by the percentage of acceptable masks and the frequency of longitudinal regression errors.Results: At screening, the mean GA area was 7.22 mm 2 with RegionFinder, 8.37 mm 2 with AI, and 8.66 mm 2 with manual planimetry.RegionFinder measured smaller GA areas than both AI and manual, with a mean difference of 1.45 mm 2 (95% confidence interval [CI]: 1.56, 1.35) versus AI (ICC 0.945) and 1.87 mm 2 (95% CI: 1.99, 1.75) versus manual (ICC 0.920).Growth rates were comparable between RegionFinder (1.54 mm 2 /year), AI (1.68 mm 2 / year), and manual (1.80 mm 2 /year) (P 0.25).Artificial intelligence masks were deemed acceptable in 84.8% of visits, and 81.4% of cases showed no regression over time.Conclusions: Artificial intelligence accurately measures GA in approximately 85% of cases, requiring human intervention in only 15%, indicating potential to streamline GA measurement in clinical trials while maintaining human oversight.

  PublisherOA PDFDOI

Frequent coauthors

• Emily Y. Chew

  National Eye Institute

  162 shared

• Ronald P. Danis

  111 shared

• Elvira Agrón

  National Eye Institute

  91 shared

• Barbara A. Blodi

  University of Wisconsin–Madison

  73 shared

• Tiarnán D L Keenan

  National Eye Institute

  71 shared

• Traci E. Clemons

  Eunice Kennedy Shriver National Institute of Child Health and Human Development

  68 shared

• Barbara Blodi

  60 shared

• Jeong W. Pak

  University of Wisconsin–Madison

  57 shared

Labs

• A-EYE Research UnitPI

Education

• M.D., Ophthalmology

  University of Wisconsin-Madison

  2003

• B.S., Biology

  University of Wisconsin-Madison

  1999