Effect of Noise Reduction on CNN Classification of OCT Images

Decision support systems can be used to diagnose ophthalmic diseases from Optical Coherence Tomography (OCT) images. Almost all of the OCT devices used by experts today have preprocessing steps such as filtering and masking. In this paper, we propose different filtering methods to noise reduction and masking OCT images. Then observes the effect of the preprocessing methods on classification results. Our experimental results on OCT images demonstrate the effectiveness of the filtering on the classification of DME, AMD and NORMAL classes are not so meaningful and time consuming.