Optimized convolution neural network based multiple eye disease detection

World health organization (WHO) reports around 2.2 billion people in the world as visually challenged which is mostly due to the age-related eye diseases such as age-related macular degeneration (AMD), cataract, diabetic retinopathy (DR) and glaucoma. These diseases lead to blindness if not diagnosed at an early stage. This paper focuses on the identification of the age-related eye diseases at an early stage using retinal fundus images taken from online dataset and pre-processed using maximum entropy transformation. The pre-processed images were fed to a convolution neural network (CNN), which was optimized using a flower pollination optimization algorithm (FPOA) for feature extraction. Hyperparameters were optimized using FPOA for training the CNN. This increased the speed and the accuracy of the network. The CNN output was fed to a Multiclass Support Vector Machine (MSVM) classifier for the classification of the type of disease. The proposed CNN-based multiple disease detection (CNN-MDD) was tested with the online dataset, namelyOcular Disease Intelligent Recognition (ODIR). The proposed model performance was analysed with the other optimized models which yielded the best performance in terms of precision, accuracy, specificity, recall, and F1 score of 98.30%, 95.27%, 95.21%, and 93.3%, respectively. The proposed method assisted automatic detection of the type of disease. Overall, this approach can be of great assistance to the medical professionals concerned in the treatment of eye diseases. •Automatic detection of multiple age-related eye diseases.•CNN hyper parameters are optimized for training the images.•Flower Pollination Optimization algorithm is used for optimization.•Cataract, Glaucoma, DR and AMD affected images are tested.•7.5% increased validation accuracy compared to the non-optimized CNN.