Machine learning analysis of multispectral retinal imaging for the early detection of age-related macular degeneration progression

Dear Editor, Age-related macular degeneration (AMD) is a progressive retinal disorder and a leading cause of irreversible vision loss in individuals over 60 years of age. It is characterized by the degeneration of the macula due to oxidative stress, inflammation, and accumulation of drusen deposits between the retinal pigment epithelium and Bruch’s membrane. The disease progresses from early to intermediate and advanced stages, with symptoms including blurred central vision, difficulty reading, and visual distortion.[1] Early detection is critical, as late-stage AMD, particularly the neovascular form, can lead to rapid and severe vision impairment. The current treatment modalities include anti-vascular endothelial growth factor injections, photodynamic therapy, and lifestyle modifications, but these are primarily effective in advanced stages. Therefore, identifying AMD progression before clinical symptoms emerge remains a key challenge in ophthalmology. Multispectral retinal imaging (MSRI) is a noninvasive diagnostic tool that captures retinal images across multiple wavelengths, enhancing the visualization of subtle structural and biochemical changes. When combined with machine learning (ML), MSRI enables the automated detection of early AMD biomarkers by analyzing spectral signatures and retinal layer morphology. ML algorithms such as convolutional neural networks and generative adversarial networks (GANs) have shown high accuracy in classifying AMD stages and predicting progression. This approach has also demonstrated utility in other retinal and systemic diseases, including diabetic retinopathy, glaucoma, and hypertensive retinopathy, where early microvascular changes are detectable through spectral imaging.[2] Recent literature supports the integration of ML with MSRI for AMD detection. In a study by Wang et al., GAN-based synthetic imaging enhanced AMD classification accuracy, achieving over 90% sensitivity in detecting early stage disease.[3] Another study by Leingang et al., utilized deep learning on the real-world optical coherence tomography datasets to stage AMD progression, demonstrating robust performance in identifying intermediate AMD and predicting conversion to advanced forms.[4] In addition to these research prototypes, Food and Drug Administration-cleared tools, such as iPredict AMD, have demonstrated clinical utility, detecting referable disease with approximately 88% accuracy. This underscores the feasibility of integrating validated AI-based platforms into routine AMD screening and referral pathways.[5] These findings emphasize the potential of ML-enhanced imaging to improve diagnostic precision and enable timely intervention across multiple retinal conditions. Despite promising results, several limitations hinder widespread adoption. First, variability in imaging protocols and device calibration can affect data consistency and model generalizability. Second, ML models require large, annotated datasets for training, which may not be readily available in all the clinical settings. Third, there is a lack of standardized validation frameworks to assess algorithm performance across the diverse populations. In addition, ethical concerns regarding data privacy and algorithm transparency remain unresolved, posing barriers to clinical integration. To optimize the utility of ML-based MSRI in AMD care, future strategies should focus on multicenter data harmonization, cost-effective imaging platforms, and clinician education. Public awareness campaigns and clinical validation drives can facilitate early adoption and help monitor long-term efficacy and safety. Incorporating explainable AI models may also enhance trust and regulatory acceptance. Ultimately, this approach holds promise for transforming AMD management by enabling proactive, personalized, and scalable screening solutions. Author contribution statement E.H: Conceptualization, Methodology, Software, Data curation, Writing Original draft preparation, Reviewing and Editing; M.T: Conceptualization, Writing Original draft preparation, Reviewing and Editing; S.N.U.S: Writing Original draft preparation, Reviewing and Editing. Use of generative AI- and AI-assisted technologies No AI- or AI-assisted technologies were used in the preparation of this manuscript. Provenance and peer review Not commissioned, externally peer-reviewed. Data availability statement The data are available for access. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.