Vision Transformers for Retinal Disease Classification using Optical Coherence Tomography Images

Optical coherence tomography is an imaging method used in medical ophthalmology for retinal disease diagnosis and treatment. With the advancements in deep learning techniques, categorization of Optical coherence tomography pictures can be done with higher accuracies. Recently vision transformers have evolved as more powerful networks than convolutional neural networks in computer vision tasks since transformers exhibit long range relationships using self-attention strategy. This work explores popular vision transformfers namely ViT and BEiT for optical coherence tomography picture categorization of retinal diseases into 7 labels. ViT and BeiT has classification accuracy of 90.32 and 91.94 respectively. Furthermore in this work vision transformers are compared to traditional convolutional neural networks in classifying retinal diseases using optical coherence tomography pictures.