A Novel Adaptive Hybrid Focal-Entropy Loss for Enhancing Diabetic Retinopathy Detection Using Convolutional Neural Networks

Diabetes related complications global The application of Artificial Intelligence (AI) in scale and lead to substantial vision loss and in technology in the field of medical imaging has many cases blindness [2]. Millions of people across the globe are affected by DR and this adds sweeping Despite the advancement in diagnostic methods, challenges to healthcare systems especially in areas most of these still necessitate manual assessment, with a rising diabetic population base. Gary et al. which is inefficient, subjective and inconsistent in (2017) further pointed out the flaws in traditional quality. This is quite worrying in countries where the manual detection and treatment methods for DR, incidence of diabetes is high like in the Khyber arguing that such approaches are not only costly but Pakhtunkhwa province of Pakistan, where around time consuming and usually involve an element of 30% of the population is diabetic and 4% of human mistake. For this reason, Convolutional blindness cases are attributed to DR. As manual Neural Networks (CNNs) have become very helpful approaches fail to meet the growing demand, in automating DR diagnosis because they drastically automated strategies using deep learning techniques, enhance the diagnostic accuracy [3]. The network such as Convolutional Neural Networks (CNNs) scans the retina for the presence of microaneurysms, seem to be an efficient and more scalable solution hemorrhages, exudates, and so forth.