A Novel Approach for Enhancement in Detection of Diabetic Retinopathy Using Customized DenseNet121

One of the leading causes of vision loss is Diabetic Retinopathy (DR), highlighting the importance of early diagnosis and treatment to prevent vision loss. Since professional diagnosis requires a careful examination of the retina, early detection of DR is important for effective treatment and prevention of blindness in the future. Traditional diagnostic methods suffer from accuracy and performance problems because they rely on human interpretation of medical images. Compared to traditional methods, this article presents an alternative that uses machine learning algorithms to determine diabetic retinopathy better. To demonstrate the effectiveness of the proposed method, this work uses the APTOS 2019 Blindness Detection dataset, which contains large-scale images of the retina using a variety of images, including eight different types of eye diseases. Circular cropping and two variations of the DenseNet121 model were employed for image Preprocessing. The accuracy of the customized DenseNet121 model surpasses that of the conventional version, highlighting the effectiveness of the proposed enhancements. The results are promising. These technologies will reduce the cost of DR-related blindness by promoting early diagnosis and prompt care and ushering in a new era of patient recovery. Thus, this study not only expands DR diagnosis with new methods and findings but also influences the field’s future direction and encourages early intervention to reduce DR-related blindness and improve behavior.