Retinal Image Laser Marks Detection using a Convolutional Neural Network

This paper presents a retinal image classifier based on Convolutional Neural Networks which are commonly deep learning used tool. This classifier uses some texture based features to decide if an input image shows evidence of previous photocoagulation laser treatments. However, this classifier could help the expert human to examine the case which make the process easier and spend less time. The solution proposed as well as the datasets used in the training and performance evaluation are presented in detail. The results obtained are described and analyzed. The accuracy obtained was above 90%. Finally, some considerations on possible improvements and future developments are made to close the article.