Diabetic retinopathy detection through generative AI techniques: A review

Abstract

Diabetes, a burgeoning health issue, especially among the youth, stems from poor dietary habits and unhealthy lifestyles. India stands as the second most afflicted country, witnessing a rapid diabetes epidemic. Diabetic Retinopathy (DR), a significant complication, threatens vision loss and blindness. Early detection, alongside lifestyle adjustments, can manage DR effectively. Traditional methods for DR detection are time consuming, costly and require specialized skills. Computer assisted screening systems, leveraging technologies like Fundus images and Optical Coherence Tomography (OCT), streamline DR detection, with Artificial Intelligence (AI) playing a pivotal role. Technological advancements and abundant data fuel significant progress in AI-based DR screening, promising enhanced accuracy and efficiency, even in remote regions. In healthcare, “normal” and “abnormal” statuses characterize patient health. AI applications in healthcare often focus on anomaly detection, leveraging distinct data distributions. Generative architectures, originally designed for content generation, find application across various domains, including healthcare. By adjusting architecture and data pipelines, controlled and specific samples can be generated, offering solutions for anomaly detection.This paper reviews fundamental aspects of diabetes and DR, exploring the utilization of generative AI in analyzing retinal data for DR detection. It also discusses recent advancements in Generative AI and their potential to enhance AI solutions in healthcare.

Keywords

• Diabetic Retinopathy
• Generative AI
• Anomaly Detection
• Explainable AI
• Computer Vision
• OCT