Automatic Detection and Monitoring of Diabetic Retinopathy Using Efficient Convolutional Neural Networks and Contrast Limited Adaptive Histogram Equalization

• Published in: IEEE access Vol. 8; pp. 136668 - 136673
• Main Authors: Momeni Pour, Asra, Seyedarabi, Hadi, Abbasi Jahromi, Seyed Hassan, Javadzadeh, Alireza
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Artificial neural networks; Biomedical imaging; Blindness; Computational modeling; contrast limited adaptive histogram equalization; convolutional neural networks; Datasets; Diabetes; Diabetic Retinopathy; EfficientNet; Equalization; Histograms; Image contrast; Image quality; Monitoring; Performance evaluation; Retina; Retinopathy; Telemedicine
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2020.3005044

Diabetic retinopathy is a medical condition of the damaged retina that is caused by diabetes and lack of proper monitoring and treatment, which usually leads to blindness. However, diabetic retinopathy monitoring requires an expert ophthalmologist. Recently, automatic monitoring models with acceptable efficiency are suggested as an alternative for expert ophthalmologists. In this paper, a new diabetic retinopathy monitoring model is proposed by using the Contrast Limited Adaptive Histogram Equalization method to improve the image quality and equalize intensities uniformly as the pre-processing step. Then, EfficientNet-B5 architecture is used for the classification step. The efficiency of this network is in uniformly scaling all dimensions of the network. The final model is trained once on a mixture of two datasets, Messidor-2 and IDRiD, and evaluated on the Messidor dataset. The area under the curve (AUC) is enhanced from 0.936, which is the highest value in all recent works, to 0.945. Also, once again, to further evaluate the performance of the model, it is trained on a mixture of two datasets, Messidor-2 and Messidor, and evaluated on the IDRiD dataset. In this case, the AUC is enhanced from 0.796, which is the highest value in all recent works, to 0.932. In comparison to other studies, our proposed model improves the AUC.