Revolutionizing Age-Related Macular Degeneration Detection Using Artificial Intelligence
NEGAR KHALAF1 , Hossein Parsaei 2 *, Tahereh Mahmoudi2 , Sina Shahparast3 , Zahra Fathollahi4 , Vahid Sadeghi5 , Mohammad Hossein Nowrozzadeh6 , Elias Khalilipour7 , siamak yousefi8
- Department of Medical Physics and Engineering, School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran. Novin Pars emerging intelligent health technologies
- Department of Medical Physics and Engineering, School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran.
- Department of Computer Science, Shiraz University, Shiraz, Fars, Iran. Novin Pars emerging intelligent health technologies
- Department of Computer Science, Shiraz University, Shiraz, Fars, Iran . Novin Pars emerging intelligent health technologies
- Department of Bioelectrics and biomedical Engineering, School of Advanced Technologies in Medicine , Isfahan University of Medical Science, Isfahan, Iran. Novin Pars emerging intelligent health technologies
- Poostchi Ophthalmology Research Center, Department of Ophthalmology, School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran
- Translational ophthalmology research center, Farabi Eye Hospital , Tehran University of Medical Science ,Tehran,Iran
- Department of Genetics, Genomics, and Informatics, University of Tennessee Health Science Center, Memphis, TN, USA
Abstract: Age-related macular degeneration (AMD) is a leading cause of vision loss among individuals over the age of 50, with a projected global prevalence of 288 million cases by 2040. The condition manifests in both dry and wet forms, significantly impairing central vision and adversely affecting daily activities and quality of life. Early detection is vital to mitigate disease progression; however, traditional diagnostic methods can be resource-intensive and depend heavily on the availability of specialists. This study introduces an artificial intelligence (AI)-driven system for the automated detection of AMD in fundus images, aimed at enhancing diagnostic accuracy, improving efficiency, and increasing accessibility for patients.
Methods: The AI system employs convolutional neural networks (CNNs), a state-of-the-art deep learning architecture, to detect hallmark features of AMD, including drusen, retinal pigment epithelium changes, and neovascular signs. The model was trained using a dataset of over 19,000 fundus images representing various AMD stages and imaging conditions. Preprocessing steps like resizing, augmentation, and normalization ensured consistency across diverse inputs. Performance was evaluated using accuracy, sensitivity, specificity, and AUROC, on both internal and external validation datasets: RFMiD and ODIR.
Results: On the internal dataset, the model achieved an accuracy of 88.8%, sensitivity of 85.3%, specificity of 91.6%, and an AUROC of 0.94. For external validation, the model performed as follows: on the RFMiD dataset, sensitivity was 83.5%, specificity was 70%, accuracy was 78.2%, and an AUROC of 0.89; on the ODIR dataset, sensitivity was 95.8%, specificity was 97.4%, accuracy was 96.6%, and an AUROC of 0.98.
Conclusion: The proposed AI-driven system provides an accurate, efficient, and automated solution for AMD detection using images from various fundus devices, with external validation ensuring the robustness of its performance. By reducing the workload on ophthalmologists and enabling early diagnosis, this system has the potential to enhance clinical decision-making and patient outcomes. Future work will focus on integrating multimodal imaging and expanding detection capabilities for other retinal pathologies, further advancing AI-assisted eye care.
