Related Experiment Video
Updated: Jul 19, 2025
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Fairness and generalisability in deep learning of retinopathy of prematurity screening algorithms: a literature
Luis Filipe Nakayama1,2, William Greig Mitchell3, Lucas Zago Ribeiro2
1Laboratory for Computational Physiology, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA luisnaka@mit.edu.
Insights
Artificial intelligence (AI) shows promise for diagnosing retinopathy of prematurity (ROP), but current studies lack fairness and generalizability. Future AI research in ROP must prioritize diverse data and bias control for reliable clinical application.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Retinopathy of prematurity (ROP) is a leading cause of childhood blindness globally.
- Current ROP diagnosis faces challenges due to specialist shortages and inconsistent standards.
- Artificial intelligence (AI) offers potential for cost-effective ROP management and diagnosis.
Approach:
- A systematic review of 220 articles was conducted, with 18 selected for full-text analysis.
- Studies were categorized by AI application in ROP: severity grading, detection, treatment prediction, and retinal zone identification.
- Analysis focused on study characteristics, fairness, and generalizability efforts in AI for ROP.
Key Points:
- Included studies predominantly featured data from high- and middle-income countries, lacking representation from low-income regions.
- Code availability was limited to two articles, and data sharing was absent in all reviewed studies.
- While reporting strong performance metrics, AI studies in ROP demonstrated limited fairness, generalizability, and bias control, with no studies explicitly detailing bias control methods.
Conclusions:
- Existing AI research in ROP exhibits significant limitations in fairness, generalizability, and reproducibility.
- The lack of diverse datasets and data/code sharing hinders the development of robust and unbiased AI tools for ROP.
- Future AI development for ROP requires collaborative efforts, diverse data inclusion, and rigorous bias control to ensure safe and effective clinical deployment.
Background:
Retinopathy of prematurity (ROP) is a vasoproliferative disease responsible for more than 30 000 blind children worldwide. Its diagnosis and treatment are challenging due to the lack of specialists, divergent diagnostic concordance and variation in classification standards. While artificial intelligence (AI) can address the shortage of professionals and provide more cost-effective management, its development needs fairness, generalisability and bias controls prior to deployment to avoid producing harmful unpredictable results. This review aims to compare AI and ROP study's characteristics, fairness and generalisability efforts.
Methods:
Our review yielded 220 articles, of which 18 were included after full-text assessment. The articles were classified into ROP severity grading, plus detection, detecting treatment requiring, ROP prediction and detection of retinal zones.
Results:
All the article's authors and included patients are from middle-income and high-income countries, with no low-income countries, South America, Australia and Africa Continents representation.Code is available in two articles and in one on request, while data are not available in any article. 88.9% of the studies use the same retinal camera. In two articles, patients' sex was described, but none applied a bias control in their models.
Conclusion:
The reviewed articles included 180 228 images and reported good metrics, but fairness, generalisability and bias control remained limited. Reproducibility is also a critical limitation, with few articles sharing codes and none sharing data. Fair and generalisable ROP and AI studies are needed that include diverse datasets, data and code sharing, collaborative research, and bias control to avoid unpredictable and harmful deployments.
