Abstract
This study uses deep learning methods to detect and predict glaucoma before symptoms appear. A dataset for glaucoma analysis was collected from open data sources provided by the Institute of Eye Diseases in Almaty, Kazakhstan. This study proposes an Efficient Convolutional-Dual-Resolution (ECDR) model that employs a Convolutional Neural Network (CNN) and a Transformer-based hybrid method to segment medical images of glaucoma. The global features obtained from the Swin Transformer module are combined with local representations obtained from a CNN-based encoder. The resulting data are compared with CNN classification methods, showing that the proposed ECDR model outperforms previous solutions. This study highlights the effectiveness of the proposed ECDR model in improving the detection and prediction of glaucoma, suggesting potential implications for early diagnosis and treatment strategies.