Akademik Veri Yönetim Sistemi
4th International Conference on Artificial Intelligence and Applied Mathematics in Engineering.ICAIAME 2022, Fatos Xhafa,D. Jude Hemanth,Tuncay Yigit,Utku Kose,Ugur Guvenc, Editör, Springer Nature, Zug, ss.671-679, 2023
Diabetic macular edema (DME) is a condition in which the blood vessels of the retina become disrupted, and fluid accumulates between the retinal layers due to long-term hyperglycemia. It is a complication unnoticeable in the early stages of diabetes but can cause visual impairment and blindness, affecting millions of people with diabetes. Therefore, monitoring of retinal morphology and fluid accumulation is required properly to protect diabetic patients from blindness. In this study, deep learning methods were used to segment the retinal layers and fluid, which is a crucial step in diagnosing eye diseases. The U-Net and DeepLabV3+ model was trained with different backbones on OCT B-scan images obtained from 10 patients. According to the experimental results, the best Dice score for retinal layers was obtained with different ResNet backbones of the U-Net model. For fluid segmentation, the best Dice score (65.94%) was obtained with the ResNet101 backbone architecture of the DeepLabV3+ model.