A comprehensive review on early detection of drusen patterns in age-related macular degeneration using deep learning models

• Age-related macular degeneration (AMD), a leading cause of vision loss, can be detected early through drusen formation. • Advanced image processing techniques, followed by deep neural networks, enable accurate detection of the tiny and ambiguous drusen structures. • Image analysis plays a significant role in localizing the area in the retina where the drusen is present. • This review examines the impact of deep learning on early AMD detection in ophthalmology. • Neural networks and machine learning significantly improve AMD diagnosis, enhancing accuracy, sensitivity, and specificity. Age-related Macular Degeneration (AMD) is a leading cause of visual impairment and blindness that affects the eye from the age of fifty-five and older. It impacts on the retina, the light-sensitive layer of the eye. In early AMD, yellowish deposits called drusen, form under the retina, which could result in distortion and gradual blurring of vision. The presence of drusen is the first sign of early dry AMD. As the disease progresses, more and larger deposits develop, and blood vessels grow up from beneath the retina leading to leakage of blood, that damages the retina. In advanced AMD, peripheral vision may remain, but the straight vision is lost. Detecting AMD early is crucial, but treatments are limited, and nutritional supplements like AREDS2 formula may slow disease progression. AMD diagnosis is primarily achieved through drusen identification, a process involving fundus photography by ophthalmologists, but the early stages of AMD make this task challenging due to ambiguous drusen regions. Furthermore, the existing models have difficulty in correctly predicting the drusen regions because of the resolution of fundus images, for which a solution is proposed as a model based on deep learning. Performance can be optimized by employing both local and global knowledge when AMD issues are still in the early phases. The area of the retina where drusen forms were identified by image segmentation, and then these deposits were automatically recognized through pattern recognition techniques.