An Enhanced RNN-LSTM Model for Fundus Image Classification to Diagnose Glaucoma

In the current medical implications, one of the leading ocular diseases is Glaucoma which majorly damage the Optic Nerve Head (ONH) of the eye retina. The intraocular pressure of the eye leads to glaucoma, which may lead to complete or partial vision loss. Regular screening and early detection is the only solution to avoid further vision loss. Due to the laborious and manual procedure in diagnosis, an automatic system is needed to diagnose glaucoma. This paper presents a novel method using deep learning-based RNNLSTM classification model to develop an automatic approach to predict and classify the images to be as healthy or glaucomatous. The RNN and LSTM with dense, dropout, and batch normalization layers are used for training and testing the proposed prediction model. The RNN model is used for training the model and to overcome the problems that occur during training, the LSTM model is applied to increase the performance of the model. The proposed model achieved an accuracy of 97.4%, specificity of 97.9% and sensitivity of 97%in classifying the images. We have made use of the DRISHTI-GS database for training and testing the proposed model.