Estimating Uncertainty of Geographic Atrophy Segmentations with Bayesian Deep Learning
Ophthalmology Science · 2025-01 · 5 (1) · 100587
DOI: 10.1016/j.xops.2024.100587 ↗Kurz gefasst
Fachartikel in der Zeitschrift Ophthalmology Science, 2025-01. Verfasst von Theodore Spaide und 18 weiteren Autor:innen. Beteiligte Einrichtung: Department of Ophthalmology, University of Washington, Seattle, Washington. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 20 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: NEI NIH HHS, NIA NIH HHS, NIH HHS.
Abstract (Originalsprache)
Purpose
To apply methods for quantifying uncertainty of deep learning segmentation of geographic atrophy (GA).
Design
Retrospective analysis of OCT images and model comparison.
Participants
One hundred twenty-six eyes from 87 participants with GA in the SWAGGER cohort of the Nonexudative Age-Related Macular Degeneration Imaged with Swept-Source OCT (SS-OCT) study.
Methods
The manual segmentations of GA lesions were conducted on structural subretinal pigment epithelium en face images from the SS-OCT images. Models were developed for 2 approximate Bayesian deep learning techniques, Monte Carlo dropout and ensemble, to assess the uncertainty of GA semantic segmentation and compared to a traditional deep learning model. Main Outcome Measures: Model performance (Dice score) was compared. Uncertainty was calculated using the formula for Shannon Entropy.
Results
The output of both Bayesian technique models showed a greater number of pixels with high entropy than the standard model. Dice scores for the Monte Carlo dropout method (0.90, 95% confidence interval 0.87-0.93) and the ensemble method (0.88, 95% confidence interval 0.85-0.91) were significantly higher (P < 0.001) than for the traditional model (0.82, 95% confidence interval 0.78-0.86).
Conclusions
Quantifying the uncertainty in a prediction of GA may improve trustworthiness of the models and aid clinicians in decision-making. The Bayesian deep learning techniques generated pixel-wise estimates of model uncertainty for segmentation, while also improving model performance compared with traditionally trained deep learning models. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Ophthalmology, University of Washington, Seattle, Washington.
- The Roger and Angie Karalis Retina Center, Seattle, Washington.
- Topcon Healthcare, Oakland, New Jersey.
- Department of Bioengineering, University of Washington, Seattle, Washington.
- Carl Zeiss Meditec, Inc., Dublin, California.
- Twenty Twenty Therapeutics, LLC, San Francisco, California.
Förderung
- NEI NIH HHS
- NIA NIH HHS
- NIH HHS
- Carl Zeiss Meditec AG
- National Institutes of Health
- Research to Prevent Blindness
Zitieren
APA
Spaide, T., Rajesh, A. E., Gim, N., Blazes, M., Lee, C. S., Macivannan, N., Lee, G., Lewis, W., Salehi, A., de Sisternes, L., Herrera, G., Shen, M., Gregori, G., Rosenfeld, P. J., Pramil, V., Waheed, N., Wu, Y., Zhang, Q., & Lee, A. Y. (2025). Estimating Uncertainty of Geographic Atrophy Segmentations with Bayesian Deep Learning. Ophthalmology Science, 5(1), 100587. https://doi.org/10.1016/j.xops.2024.100587
Vancouver
Spaide T, Rajesh AE, Gim N, Blazes M, Lee CS, Macivannan N, et al. Estimating Uncertainty of Geographic Atrophy Segmentations with Bayesian Deep Learning. Ophthalmology Science. 2025;5(1):100587. doi:10.1016/j.xops.2024.100587
BibTeX
@article{Spaide2025,
author = {Spaide, Theodore and Rajesh, Anand E and Gim, Nayoon and Blazes, Marian and Lee, Cecilia S and Macivannan, Niranchana and Lee, Gary and Lewis, Warren and Salehi, Ali and de Sisternes, Luis and Herrera, Gissel and Shen, Mengxi and Gregori, Giovanni and Rosenfeld, Philip J and Pramil, Varsha and Waheed, Nadia and Wu, Yue and Zhang, Qinqin and Lee, Aaron Y},
title = {Estimating Uncertainty of Geographic Atrophy Segmentations with Bayesian Deep Learning},
journal = {Ophthalmology Science},
year = {2025},
volume = {5},
number = {1},
pages = {100587},
doi = {10.1016/j.xops.2024.100587}
}
EUVIMED