A Deep Learning-Based Framework for Early Diabetes Prediction Using Retinal Fundus Images

Diabetes mellitus is a worldwide health issue, with the rising prevalence and serious prolonged issues, such as cardiovascular disease, peripheral neuropathy, and sight loss. Early diagnosis is essential to avoid irredeemable harm, but traditional diagnostic tools, e.g., blood glucose and HbA1c testing, are invasive, periodic and unavailable in low-resource environments. The retinal fundus imaging is a non-invasive alternative that captures microvascular alterations related to early diabetes and therefore has the potential of being an effective modality in predictive screening. The study proposed a deep learning-based predicting framework of early diabetic diseases through retinal fundus images. The framework operates on a convolutional neural network (CNN) that operates on a pre-trained ResNet-50 backbone on a training based on transfer learning and fine-tuning. Preprocessing of the data was done by determining the quality of images, resizing, data normalization, and contrast enhancement with CLAHE, noise removal, and data augmentation to address the class imbalance. The dataset used in experiments contained 12,000 retinal images, subdivided into the training, validation, and testing groups, and such evaluation measures as accuracy, precision, recall, specificity, F1-score, and ROC-AUC were used to evaluate the methods. Grad-CAM visualizations were used so that the interpretations are interpretable and relevant to clinical application. The results demonstrates high predictive performance, with an accuracy of 94.2% precision of 92.8, recall of 93.1 and F1-score of 93.1 and ROC-AUC of 0.96. The model was effective in addressing interpretable clinically meaningful retinal areas. The contributions made are a strong, non-invasive predictive model, elaborate preprocessing plans and thorough assessment. This paper indicates that AI retinal analysis is potentially valuable in prompt detection of diabetes in the initial phases of the illness to provide early interventions and enhance patient health outcomes.

Early diabetes prediction, Retinal fundus imaging, Deep learning

1. Abràmoff, M. D., Lavin, P. T., Birch, M., Shah, N., & Folk, J. C. (2018) 'Autonomous AI for diabetic retinopathy detection', npj Digital Medicine, 1. 10.1038/s41746-018-0040-6. [Google Scholar] [Crossref]

2. Cheung, N.; Mitchell, P.; Wong, T.Y. Diabetic retinopathy. Lancet 2010, 376, 124–136. [CrossRef] [Google Scholar] [Crossref]

3. Colagiuri, S., & Ceriello, A. (2025) 'Diabetes detection, prevention, and hyperglycaemia management', Diabetes Research and Clinical Practice, 112145. doi: 10.1016/j.diabres.2025.112145 [Google Scholar] [Crossref]

4. Elsayed, Nuha & Aleppo, Grazia & Bannuru, Raveendhara & Bruemmer, Dennis & Ekhlaspour, Laya & Gaglia, Jason & Hilliard, Marisa & Johnson, Eric & Khunti, Kamlesh & Lingvay, Ildiko & Matfin, Glenn & McCoy, Rozalina & Perry, Mary & Pilla, Scott & Polsky, Sarit & Prahalad, Priya & Pratley, Richard & Segal, Alissa & Gabbay, Robert. (2023). 2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes—2024. Diabetes Care. 47. S20-S42. 10.2337/dc24-S002. [Google Scholar] [Crossref]

5. Gulshan, V., et al. (2016) 'Deep learning for diabetic retinopathy detection', JAMA, 316, 2402–2410. 10.1001/jama.2016.17216. [Google Scholar] [Crossref]

6. Nathan, D. M., et al. (2009) 'A1C assay role in diabetes diagnosis', Diabetes Care, 32, 1327–1334. 10.2337/dc09-9033. [Google Scholar] [Crossref]

7. Nelson M, Dungan KM. (2025) ‘Diagnostic Tests for Diabetes Mellitus. In: Feingold KR, Adler RA, Ahmed SF, et al., editors. Endotext [Internet]. South Dartmouth (MA): MDText.com, Inc.; 2000-. Available from: https://www.ncbi.nlm.nih.gov/books/NBK278985/ [Google Scholar] [Crossref]

8. Ogurtsova, K.; da Rocha Fernandes, J.D.; Huang, Y.; Linnenkamp, U.; Guariguata, L.; Cho, N.H.; Cavan, D.; Shaw, J.E.; Makaroff, L.E. IDF Diabetes Atlas: Global estimates for the prevalence of diabetes for 2015 and 2040. Diabetes Res. Clin. Pract. 2017, 128, 40–50. [CrossRef] [Google Scholar] [Crossref]

9. Poplin, R., et al. (2018) 'Retinal images predict cardiovascular risk factors', Nature Biomedical Engineering, 2(3), 158–164. doi: 10.1038/s41551-018-0195-0. [Google Scholar] [Crossref]

10. Sona, P & N D, Bisna & James, Ajay. (2025). Retinal Image Analysis for Heart Disease Risk Prediction: A Deep Learning Approach. IEEE Access. PP. 1-1. 10.1109/ACCESS.2025.3562433. [Google Scholar] [Crossref]

11. Ting, D.S.W.; Cheung, C.Y.; Lim, G.; Tan, G.S.W.; Quang, N.D.; Gan, A.; Hamzah, H.; Garcia-Franco, R.; San Yeo, I.Y.; Lee, S.Y.; et al. Development and Validation of a Deep Learning System for Diabetic Retinopathy and Related Eye Diseases Using Retinal Images from Multiethnic Populations with Diabetes. JAMA 2017, 318, 2211–2223. [CrossRef] [Google Scholar] [Crossref]

12. Virgili, G.; Menchini, F.; Murro, V.; Peluso, E.; Rosa, F.; Casazza, G. Optical coherence tomography (OCT) for detection of macularoedema in patients with diabetic retinopathy. Cochrane Database Syst. Rev. 2011, 2011, CD008081. [CrossRef] [Google Scholar] [Crossref]

• Travaalay: An AI-Powered Mobile Platform for Tourism with Student Translator Guides, Agro-Tourism, and Astro-Tourism Experiences
• Integrating QVoC (QR Code with Voice Content) to Enhance Medication Adherence for Geriatric Diabetic Patients
• "Navigating Global Volatility: Assessing the Resilience and Innovation of Bahrain’s Financial Sector Through the 2025 Financial Stability Report"
• Comparison of Similarity Distance-Based Metrics for HODA and BANGLA Dataset for Enhanced Precision
• Stereo Matching Frameworks for Depth-Aware Object Detection: A Comprehensive Review