AI-Powered DR Screening in Africa: Bridging Deep Learning, DICOM Tools, and Healthcare Gaps

Artificial intelligence (AI) is increasingly recognized as a promising approach for the early diagnosis of diabetic retinopathy (DR), a major cause of preventable blindness world-wide. In Senegal, and particularly in rural areas, DR remains a significant public health issue due to limited access to specialized care and diagnostic infrastructure. This review aims to provide a comprehensive and critical analysis of current deep learning-based approaches for DR screening, with a focus on their applicability in low-resource contexts. We examine the socio-demographic factors influencing disease prevalence, the potential of open-source platforms such as Orthanc for managing medical imaging data, and the performance of state-of-the-art AI models, including Convolutional Neural Networks (CNNs) such EfficientNet, GoogleNet, VGG16, etc. By highlighting existing gaps and challenges, this review outlines key considerations for the development of accessible, portable, and context-adapted AI tools for DR diagnosis in sub-Saharan Africa particularly in Senegal.