Development of a Deep Learning-Based Algorithm for Detection and Analysis of Retinal Fluid Regions in OCT Images

Abstract This paper presents the development of an algorithm for detecting and analyzing retinal fluid regions in optical coherence tomography (OCT) images using deep learning methods. The study addresses the urgent problem of automatic segmentation and quantitative analysis of intraretinal, subretinal, and retinal pigment epithelial fluids. The use of convolutional neural networks, in particular the ReLayNet architecture, ensured high segmentation accuracy, which is confirmed by experimental results. On the test dataset, the Dice coefficient of 0.87 and the focal Tversky loss of 0.12 were achieved, indicating a high degree of correspondence between the predicted and true masks. An algorithm for calculating retinal fluid volumes was also developed, which can be useful for differential diagnosis of retinal diseases such as age-related macular degeneration. The results demonstrate the potential of the proposed approach in clinical practice to improve the diagnosis and monitoring of eye diseases.