Deep convolutional neural network-based diabetic eye disease detection and classification using thermal images

Abstract

Listen

Translate article

Infrared thermography which is non-contact and non-invasive technique is widely accepted as a medical diagnostic tool. Thermal images are processed for abnormality detection and quantification. It has been used in the diagnosis of dry eye, meibomian gland dysfunction, thyroid eye disease and glaucoma. Diabetic eye disease (DED) detection using thermal images is an absolutely new attempt. The early detection of the occurrence of DED can be very helpful for clinical treatment. In this paper, we are attempting towards finding an automatic way to classify DEDs in thermal images using a deep learning-based convolutional neural network (CNN) methodology. The sensitivity of 92.30%, specificity of 98.46%, and accuracy of 95.38% on testing dataset with reference to expert's ground truth results are obtained. The results attained evidently exhibit that the thermal imaging is promising modality and proposed deep learning method is capable for automatic diagnosis of diabetic eye disease classification.