This study proposes a novel diagnostic approach to retinal disease detection by combining deep learning-based segmentation with fractal dimension (FD) analysis on optical coherence tomography (OCT) images. Our primary goal is to enhance early detection of retinal diseases such as age-related macular degeneration (AMD), diabetic retinopathy (DR), central serious retinopathy (CSR), and macular hole (MH). We introduce the Attention U-shaped Network (AUNet), which builds upon the UNet architecture with Attention Gates (AGs) to improve focus on pathological structures in complex cases, achieving a high segmentation accuracy of 98.5% and a mean Intersection over Union (mIoU) of 0.91, outperforming existing models like UNet++ and DeepLabV3+. Coupled with Fourier and Higuchi FD analysis, our method quantitatively assesses the complexity of retinal layers, identifying structural patterns that serve as early indicators of neural degeneration. Statistical tests reveal significant differences in FD values between diseased and healthy groups, underscoring the predictive power of retinal layers like the retinal pigment epithelium (RPE), inner nuclear layer (INL), outer nuclear layer (ONL), and ellipsoid zone (EZ). This combined AUNet-FD approach represents an innovative tool for early diagnosis of retinal diseases, potentially enhancing clinical decision-making through precise, non-invasive analysis.