Impact of Novel Pre-Processing Techniques on Retinal Images

Glaucoma, a gradually developing eye condition that can damage the optic nerve, is one of the leading causes of irreversible blindness worldwide. The detection of glaucoma is required in its initial phases for the prevention of the disease. Several machine learning and deep learning models have been developed to detect the disease. These models require a preprocessed structured data set for better efficiency and accuracy of detection. The image datasets used for this purpose have been found with several issues of outliers and noise that limit the accuracy of the results. Hence, proper handling of the dataset in terms of pre-processing is vital. This paper presents a novel pre-processing technique for better glaucoma detection as it is necessary to remove outliers from the retinal images. The proposed “EMF” model for pre-processing comprises Enhancement, Morphology, and Filtering of the images to get better results on the Retina. Drishti dataset has been used to evaluate the efficiency of the proposed technique. This dataset has several outliers that limit the accuracy, making it difficult to diagnose glaucoma. However, with the usage of the proposed technique, the dataset could be efficiently utilized for glaucoma detection. Compared to current state-of-the-art techniques, the findings from the proposed approach were shown to be better in terms of MSE, PSNR, and entropy.