Home About Login Current Archives Announcements Editorial Board
Submit Now For Authors Call for Submissions Statistics Contact
Home > Archives > Volume 20, No 8 (2022) > Article

DOI: 10.14704/nq.2022.20.8.NQ44409

Glaucoma diagnosis utilising Le-Net and supervised machine learning techniques in retinal fundus images

Dr.R.Muthalagu, Dr.S.Vijayanand, Dr.M.Ganesh Mani

Abstract

Several types of health care systems use content-based image analysis and computer vision to find diseases. Glaucoma is thought to be the second most common eye disease that can lead to a neurodegenerative illness. Glaucoma is an eye disease that starts when the pressure inside the eye is too high. When it gets worse, it can make it impossible to see. Whereas early treatment based on glaucoma screening can keep a person from losing all of their vision. Accurate screening methods depend on the availability of experts who can look at samples of the retina by hand to find the areas where glaucoma is present. But because glaucoma screening is hard and there aren't enough people to do it, we often have to wait, which can cause more people around the world to lose their sight. There is an immediate and pressing need to develop an efficient automated framework that is capable of reliably identifying Optic Disc (OD) and Optic Cup (OC) lesions in their early stages to reportthe issues that are caused by manual methods.Identifying and classifying glaucomatous areas is hard because lesions vary in size, colour, orientation, and shape. Also, there are a lot of similarities between the lesion and the colour of the eye, which makes it harder to classify. To solve these problems, we've come up with a Deep Learning (DL)-a based method called EfficientDet-D0, which uses EfficientNet-B0 as its backbone. Based on the CDR structure, a deep learning method is then used to figure out the CDR value. In this proposed method, the fundus images are cleaned up using wavelet-based denoising. In this proposed method, the optic disc and optic cup should be taken out using the best methods and two similar neural networks with deep convolutions architectures. One method gives perfect results. The OC and OD segmentation design was tested and trained using data from SCES and ACRIMA. With this data set, it was found that 96 per cent of glaucoma diagnoses were correct. Lastly, we talk about the different problems with research and how to fix them, which can help researchers do more work on glaucoma detection.

Keywords

Classification, cup-to-disc ratio, glaucoma, Convolutional Neural Network, Deep Convolution Neural Network, DIARETDB1 data set, MESSIDOR data set.

Full Text

PDF

References

?>