Optical Coherence Tomography (OCT) is a key imaging modality for the diagnosis and monitoring of retinal diseases, enabling high-resolution visualization of retinal micro-structures. However, the complexity and growing volume of OCT data place a significant burden on clinicians and increase the risk of differences between observers. This motivates the development of automated and reliable decision-support tools for retinal disease classification.
A deep learning-based framework is presented for retinal disease classification using OCT images. The method focuses on clinically relevant conditions such as choroidal neovascularization (CNV), diabetic macular edema (DME), and drusen, in addition to normal cases.
The study begins with a two-dimensional single-slice analysis using convolutional neural networks (CNNs) and transfer learning, establishing a strong baseline for disease classification. In addition, a hierarchical classification strategy is evaluated to align the classification process with clinically motivated decision processes.
