RetoNet: a deep learning architecture for automated retinal ailment detection

Researchers are trying to tap the immense potential of big data to revolutionize all aspects of societal activity and to assist in having well informed decisions. Healthcare being one such field where proper analytics of available big medical data can lead to early detection and treatment of many ailments. Machine learning played a significant role in the design of automated diagnostic systems and today we have deep learning models in this arena which are outperforming human expertise in terms of predictive accuracy. This paper proposes RetoNet, a convolutional neural network architecture, which is trained and optimized to detect retinal ailment from fundus images with pronounced accuracy and its performance is also proven to be superior to a transfer learning based model developed for the same. Deep learning based e-diagnostic system can be an accurate, cost effective and convenient solution for the shortage of expertise on demand in the healthcare field.