An Efficient Neovascularization Detection in Fundus Images Using Transfer Learning

This chapter explores the critical task of detecting neovascularization in retinal images, a key indicator of proliferative diabetic retinopathy. The study employs transfer learning with MobileNet, achieving a remarkable accuracy of 97.6%. The methodology involves a comprehensive comparison with other models like GoogleNet, ResNet18, and a combination of CNN with SVM. The results demonstrate the superior performance of MobileNet in identifying neovascularization, making it a valuable tool for early diagnosis. The chapter also discusses the importance of early detection to prevent vision loss and the challenges associated with differentiating new blood vessels from normal ones. The detailed analysis and comparative evaluation provide professionals with a clear understanding of the most effective techniques for neovascularization detection.