A novel diabetic retinopathy detection from fundus images using hybrid quantum convolutional neural network models

Diabetic retinopathy (DR) diagnosis from digital fundus images is a long-standing topic of research in medical image processing. The determination of optic disk boundaries in two-dimensional retinal images is difficult due to blurred edges, which makes this field in need of improvement. All these problems cannot be solved by a single technique. An efficient algorithm for identifying DR-related retinal changes and structure is still needed. If DR is recognized and treated in a timely manner, visual deterioration can be managed or avoided. It is based on telemedicine analysis of color fundus pictures or clinical evaluations by medical doctors. However, due to intrinsic human subjectivity, both systems are time-consuming, labor-intensive, and prone to inaccuracy. Due to their great specificity and sensitivity, automated methods capable of analyzing color fundus pictures have become important for the general deployment of DR screening. To study the existence of DR-related characteristics and to cope with the various diabetes severity diagnosis phases, a hybrid quantum convolutional neural network (HQCNN) is presented. Kaggle fundus images database is utilized to test and train the network. Finally, the presented work is compared for analyzing efficiency using the system of measurement like precision, specificity, accuracy, sensitivity, and f1 score. The proposed work obtains accuracy of 98.89%, sensitivity of 99.37%, specificity of 99.57%, precision of 98.89%, and F1 score of 97.58%.