SE-MIDNet Based on Deep Learning for Diabetic Retinopathy Classification

Diabetic Retinopathy (DR) is one of the most serious complications of diabetes. At present, DR detection mainly relies on detailed analysis by ophthalmologists. However, manual diagnosis is time-consuming and low efficiency. Aiming at the task of DR automatic classification, this paper proposes a classification method of DR based on deep learning. In view of the different sizes of the lesion area, firstly, an improved Inception module is proposed, which enables the network to efficiently extract multi-scale features of DR images. Then, the dense connection method is used to splice the output feature maps of the improved Inception module and send them to the subsequent layers to realize the multi-scale feature reuse of DR images and enhance the feature representation of small targets. Finally, the Squeeze-and-Excitation (SE) module is used to obtain the global information of the feature map on each channel, and the dynamic nonlinear modeling of each channel is carried out to improve the generalization ability of the network. The experimental results show that the network structure designed in this paper has good generalization ability, and the accuracy of DR automatic classification reaches 88.24%, the sensitivity reaches 99.43%, and the specificity reaches 97.6%, which can meet the needs of hospitals for DR classification.