Can artificial intelligence with multimodal imaging outperform traditional methods in predicting age-related macular degeneration progression? A systematic review and exploratory meta-analysis

• Published in: BMC medical informatics and decision making Vol. 25; no. 1; pp. 321 - 16
• Main Authors: Chen, Kai-Yang, Chan, Hoi-Chun, Chan, Chi-Ming
• Format: Journal Article
• Language: English
• Published: London BioMed Central 01.09.2025; BioMed Central Ltd; Springer Nature B.V; BMC
• Subjects: Accuracy; Age related diseases; Age-related macular degeneration; Aging; Algorithms; Artificial intelligence; Artificial Intelligence - standards; Atrophy; Business metrics; Clinical medicine; Datasets; Deep learning; Development and progression; Diagnostic imaging; Disease; Disease Progression; Eye diseases; Health Informatics; Heterogeneity; Humans; Image processing; Information Systems and Communication Service; Machine learning; Macular degeneration; Macular Degeneration - diagnosis; Macular Degeneration - diagnostic imaging; Management of Computing and Information Systems; Medical imaging; Medicine; Medicine & Public Health; Meta-analysis; Methods; Multimodal Imaging - methods; Multimodal Imaging - standards; Performance assessment; Predictive models; Retina; Risk factors; Sensitivity analysis; Sensitivity and Specificity; Systematic Review; Tomography; Taiwan; Deep learning; Predictive models; Age-related macular degeneration; Artificial intelligence
• ISSN: 1472-6947; 1472-6947
• DOI: 10.1186/s12911-025-03119-z

Purpose Age-related macular degeneration (AMD) is a leading cause of irreversible vision loss, and its prevalence is expected to rise with aging populations. Early prediction of AMD progression is critical for effective management. This systematic review and meta-analysis evaluate the accuracy, sensitivity, and specificity of artificial intelligence (AI) algorithms in in detecting and predicting progression of AMD. Methods Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, a systematic review and meta-analysis were conducted from inception to February 7th, 2025. We included five studies that assessed the performance of AI algorithms in predicting AMD progression using multimodal imaging. Data on accuracy, sensitivity, and specificity were extracted, and meta-analysis was performed using Comprehensive Meta-Analysis software version 3.7. Heterogeneity was assessed using the I² statistic. Results Of the five studies, AI models demonstrated superior accuracy (mean difference: 0.07, 95% CI: 0.07, 0.07; p  < 0.00001) and sensitivity (mean difference: 0.08, 95% CI: 0.08, 0.08; p  < 0.00001) compared to retinal specialists. Specificity also showed a minimal but significant advantage for AI (mean difference: 0.01, 95% CI: 0.01, 0.01; p  < 0.00001). Importantly, heterogeneity was minimal to absent across all analyses (I² = 0–0.42%), supporting the reliability and consistency of pooled findings. Conclusion AI algorithms outperform retinal specialists in predicting AMD progression, particularly in accuracy and sensitivity. These findings support the potential of AI in AMD prediction; however, given the limited number of included studies, the results should be interpreted as exploratory and in need of validation through future large-scale, prospective studies. Key messages What is known AI algorithms have demonstrated potential in predicting AMD progression using multimodal imaging. Early prediction of AMD progression is crucial for effective management and treatment. What is new AI models outperform retinal specialists in accuracy and sensitivity in predicting AMD progression. AI also shows a minimal advantage in specificity over retinal specialists.