(IJACSA) International Journal of Advanced Computer Science and Applications, Vol. 13, No. 9, 2022 401 | Page www.ijacsa.thesai.org Multiple Eye Disease Detection using Hybrid Adaptive Mutation Swarm Optimization and RNN P. Glaret Subin 1 , P. Muthu Kannan 2 Department of ECE, Saveetha School of Engineering, Saveetha Institute of Medical and Technical Sciences Chennai-602105, India. 1, 2 Department of ECE, College of Engineering and Technology, SRM Institute of Science and Technology, Vadapalani Campus Chennai- 600026, India 1 Abstract—The major cause of visual impairment in aged people is due to age related eye diseases such as cataract, diabetic retinopathy, and glaucoma. Early detection of eye diseases is necessary for better diagnosis. This paper concentrates on the early identification of various eye disorders such as cataract, diabetic retinopathy, and glaucoma from retinal fundus images. The proposed method focuses on the automated early detection of multiple diseases using hybrid adaptive mutation swarm optimization and regression neural networks (AED-HSR). In the proposed work, the input images are preprocessed and then multiple features such as entropy, mean, color, intensity, standard deviation, and statistics are extracted from the collected data. The extracted features are segmented by using an adaptive mutation swarm optimization (AMSO) algorithm to segment the disease sector from the fundus image. Finally, the features collected are fed to a regression neural network (RNN) classifier to classify each fundus image as normal or abnormal. If the classifier output is abnormal, then it is classified by the corresponding diseases in terms of cataract, glaucoma, and diabetic retinopathy, which improves the accuracy of detection and classification. Ultimately, the results of the classifiers are evaluated by several performance analyses and the viability of structural and functional features is considered. The proposed system predicts the type of the disease with an accuracy of 0.9808, specificity of 0.9934, sensitivity of 0.9803 and F1 score of 0.9861 respectively. Keywords—Adaptive mutation swarm optimization; fundus image; feature extraction; RNN classifier; standard deviation I. INTRODUCTION Nowadays, aged people are mostly affected by chronic diseases such as cataract, glaucoma and diabetic retinopathy, which lead to visual impairment. The optic nerve is damaged due to glaucoma, which results in loss of vision. Glaucoma occurs due to a slow rise in the normal fluid pressure inside the eyes. Cataract occurs due to the clouding of the eye's lens. The progressive damage in the retina’s blood vessels, which are essential for good vision of the eye, leads to Diabetic Retinopathy [1]. Based on the supervised learning method, blood vessels are segmented from the fundus image that can be done using Zernike moment-based Shape descriptors and training can be performed using an ANN-based binary classifier to predict cardio vascular diseases [2]. Multiple instances A learning technique is used to classify the diseased image and healthy image, in which the classification can be done by binary classification [3]. Microaneurysms can be recognized using principal component analysis, morphological processing, averaging filter, and support vector machine classifier. Diabetic Retinopathy disease can also be identified [4]. The early signs of diabetic retinopathy can be identified by applying nineteen features extracted from the fundus image to an artificial neural network, which is trained by Levenberg- Marquardt, and the disease is classified by using Bayesian Regularization [5]. Red lesions can be detected in the blood vessels by using a Gaussian filter and the disease can be predicted using an SVM classifier [6]. Based on the singular value decomposition algorithm, dictionary learning methods can be used to classify healthy people from diabetic patients based on singular Value Decomposition Algorithm [7]. The fundus image is segmented by using a Deep Convolution Neural Network and it increases the accuracy and efficiency in predicting non-proliferated diabetic retinopathy [8]. The blood vessels can be segmented by using dilated convolution, which leads to more accurate detection of ophthalmologic diseases [9]. The two filtering methods, namely median filtering and Gaussian derivative filtering, are used to define the bifurcation point of a blood vessel image segment [10]. DR (Diabetic Retinopathy) can be recognized using the ANN classifier and region growing segmentation to extract exudates, optic plate and veins from the fundus images [11]. DR can be detected by using a reformed capsule network, which attains an accuracy of 97.98% [12]. A hierarchical severe grading system model was developed to detect and classify the different grades of DR. The classifier accuracy is 94% [13]. The optic disc and optic cup boundary of the fundus images are segmented and by using Weighted Least Square fit, holistic features and disc ratio are extracted, and then they are fed to a Convolutional Multi-Layer Neural Network Classifier to classify the glaucoma [14]. A classification method of multi feature analysis along with a Discrete Wavelet transform is used to detect glaucoma. This model classifies glaucoma with an accuracy of 95% [15]. The input fundus image is validated using Le-Net architecture and the optic disc and optic cup are segmented using U-Net Architecture. Glaucoma can be detected with the use of SVM Classifier, Neural Network Classifier, and Adaboost Classifiers [16]. The eyeball area is extracted from the fundus image using an object detection network and multi task learning is applied to detect the cataract [17]. A Deep Convolution Neural Network with Resnet for classification can be used to identify cataract. The systems show an accuracy of 95.77% [18]. Gray Level Co-occurrence Matrix is utilized for feature extraction, and the classification of different levels of cataract can be done by Back Propagation Neural Network Classifier. This system