Retinal Disease Classification in Optical Coherence Tomography Images Using a Convolutional Neural Network Algorithm

Machine learning (ML) algorithms and deep learning (DL) algorithm applications in ophthalmology are rapidly advancing, particularly in the field of retinal optical coherence tomography (OCT) pathology classification. The main goal is to enhance the diagnosis and categorization of retinal diseases, leveraging the capabilities of retinal OCT imaging. This noninvasive technique provides detailed cross-sectional views of the retina using convolutional neural networks (CNN). This imaging method aids in the identification and tracking of various ocular ailments. The integration of ML algorithms in the analysis of retinal OCT images has demonstrated significant potential in early disease detection and classification. Diseases like age-related muscular degeneration, diabetic retinopathy, and glaucoma can be more accurately identified and assessed. These algorithms efficiently process retinal OCT images, extracting quantitative data and discerning subtle structural changes and thickness variations that might elude visual inspection. The ML workflow for classifying retinal OCT pathology entails extracting key features from OCT images. These features encompass parameters such as retinal layer thickness, textural attributes, blood vessel characteristics, and anomalies like fluid or exudate. These extracted features, such as logistic regression (LR) and random forest (RF), serve as input for training the ML model. The model's efficacy is then validated on an independent data set to gauge its accuracy and performance, and results are compared with the CNN algorithm.