This study aimed to improve the detection performance of age-related macular degeneration (AMD) in low-quality retinal images. AMD is a leading cause of vision loss among older adults worldwide, and accurate detection is crucial for effective clinical management. However, low-quality optical coherence tomography (OCT) images substantially compromise diagnostic accuracy. To address this issue, we enhanced AMD detection in low-quality images using noise-augmented training data and a redesigned You Only Look Once (YOLO) deep learning model. Public OCT datasets from the University of California, San Diego, and Duke University were utilized. The training dataset contained 24,980 OCT images, comprising both high-quality and noise-augmented low-quality images, and the testing dataset contained 1,000 images (584 AMD and 416 normal). The proposed model was based on the YOLOv8n framework, integrating the squeeze-and-excitation (SE) and adaptive sparse self-attention (ASSA) modules, as well as an additional 160×160 detection layer for detecting small lesions. Evaluation metrics included accuracy, sensitivity, specificity, and F2-score. On the Duke dataset, the proposed model achieved an accuracy of 0.990, a sensitivity of 0.982, a specificity of 1.000, and an F2-score of 0.985; furthermore, the detection rate in low-quality images improved substantially to 0.896 compared with 0.551 for the original YOLOv8n evaluated under the same experimental conditions. Ablation experiments confirmed that each module added to the base framework contributed to performance gain. By adjusting the confidence threshold, the model could be operated at a high-specificity operating point (threshold 0.5; specificity 1.000) suitable for confirmatory assessment, or at a high-detection operating point (threshold 0.1; detection rate 0.998, sensitivity 0.988) suitable for early screening. Overall, the enhanced model using a combination of noise-augmented training data and enhanced attention mechanisms yielded excellent AMD detection performance in low-quality OCT images, demonstrating strong potential for clinical application.