OCT DEEPNET 1-A Deep Learning Approach for Retinal OCT Image Classification

Machine learning algorithms play a vital role in disease diagnosis and prediction. This research proposes a deep learning architecture framework for the classification of retinal optical coherence tomography images into four classes. The deep learning architecture was termed OCT DEEPNET 1. The tailor-made deep learning frameworkwas evaluated using the public database images for various batch sizes. The input images were classified into 4 classes; normal, choroidal neovascularization, diabetic macular edema, and Drusen. Satisfactory results were obtained, and the performance metrics were also evaluated. The hyper-parameters are tuned using random search optimization. The results reveal the proficiency of the deep learning approach in the classification. An accuracy of 88% was obtained with a batch size of 32 when OCT DEEPNET 1 was employed for the classification problem. The outcome of this research work paves a way toward the automated classification of OCT retinal images for disease diagnosis.