Multitasking Deep Learning Model for Detection of Five Stages of Diabetic Retinopathy

Abstract

Early diagnosis and treatment of diabetic retinopathy (DR) can reduce the risk of vision loss. There are five stages of DR consisting of no DR, mild DR, moderate DR, severe DR, and proliferate DR. This paper presents a multitask deep learning model to detect all the five stages of DR more accurately than existing methods. The developed multitask model consists of one classification model and one regression model, each with its own loss function. After training the regression model and the classification model separately, the features extracted by these two models are concatenated and inputted to a multilayer perceptron network to classify the five stages of DR. A modified Squeeze Excitation Densely Connected deep neural network is also developed as part of this multitasking approach. The developed multitask model is applied to the two large Kaggle datasets of APTOS and EyePACS. The results obtained indicate that the developed multitask model achieved a weighted Kappa score of 0.90 and 0.88 for the APTOS and EyePACS datasets, respectively. In addition, the micro and macro average area under the receiver operating characteristic (ROC) curve was found to be 0.96, and 0.93, respectively, which are higher than existing methods for detecting the five stages of DR.

Keywords

• Diabetic retinopathy (DR)
• eye fundus images
• five stages of diabetic retinopathy
• multitasking deep neural network
• squeeze excitation densely connected network