Automated Segmentation of Retinal Biomarkers Using U-Net3+ in Optical Coherence Tomography Images for Enhanced Diagnosis of Age-Related Macular Degeneration

An imaging technique used to diagnose eye diseases and analyze changes in the structure of the retina, also known as optical coherence tomography (OCT). The importance of segmenting biomarkers such as drusen, confluent drusen, fluid and scars in OCT images is that it allows the diagnosis and evaluation of various retinal diseases, namely the various stages of age-related macular degeneration (AMD). Over the past few years, a significant amount of research has been conducted on the use of machine learning algorithms, including classical machine learning and deep learning methods, to automate the segmentation of retinal cysts, fluid and other anomalies. Ophthalmologists can use these automated methods to improve the interpretation and quantification of retinal features, which can lead to more accurate diagnoses and better treatment decisions. As part of this study, a neural network model was developed and tested on the productive U-Net3+ architecture, on the OCTDL data set when analyzing optical coherence tomography images, and a comparative analysis of the performance of the resulting model with the results of other authors was presented in order to test its stability to new data in the form of images optical coherence tomography, for further application of the resulting neural network model in a decision support system for ophthalmologists.