Many eye diseases like cataracts, trachoma, or corneal ulcer can cause vision problems. Progression of these eye diseases can only be prevented if they are recognized accurately at the early stage. Visually observable symptoms differ a lot among these eye diseases. However, a wide variety of symptoms is necessary to be analyzed for the accurate detection of eye diseases. In this paper, we propose a novel approach to provide an automated eye disease recognition system using visually observable symptoms applying digital image processing techniques and machine learning techniques such as deep convolution neural network (DCNN) and support vector machine (SVM). We apply the principal component analysis and t-distributed stochastic neighbor embedding methods for better feature selection. The proposed system automatically divides the facial components from the frontal facial image and extracts the eye part. The proposed method analyzes and classifies seven eye diseases including cataracts, trachoma, conjunctivitis, corneal ulcer, ectropion, periorbital cellulitis, and Bitot's spot of vitamin A deficiency. From the experimental results, we see that the DCNN model outperforms SVM models. We also compare our method with some other existing methods. Our method shows improved accuracy compared to other methods. The average accuracy rate of our DCNN model is 98.79% with sensitivity of 97% and specificity of 99%.
AKRAM, ASHRAFI and DEBNATH, RAMESWAR
"An automated eye disease recognition system from visual content of facial imagesusing machine learning techniques,"
Turkish Journal of Electrical Engineering and Computer Sciences: Vol. 28:
2, Article 23.
Available at: https://journals.tubitak.gov.tr/elektrik/vol28/iss2/23