Artificial Intelligence based analysis of fundus images of retina to screen for diabetic retinopathy and cataract: A pilot study in North India

Introduction: Fundus examination is the first and foremost step in the diagnosis of vision conditions like diabetic retinopathy, cataract, corneal opacity and vitreous infection.But the fundus examination requires considerable experience and is subject to human errors.Artificial intelligence (AI) based analysis of fundus images of retina can be an answer to this problem.AI can be used even in the remotest areas where expert ophthalmologists are not available.Sevamob provides artificial intelligence enabled healthcare platform to organizations.It uses deep learning for image recognition, machine learning for triaging and computer vision for object counting.AI models of various medical conditions are first trained in the software from anonymized image data procured from various sources.To determine the accuracy of AI based point-of-care screening solution for fundus images of retina, an android smartphone / tablet with Sevamob app and a low end fundus camera were used.The system was operated by a nurse or a technician with minimal training Methods: Fundus images of retina from clinically suspected diabetic retinopathy ,cataract, corneal opacity, vitreous infection and membrane hemorrhage were included in the study.Results: Out of 151 fundus images, an expert ophthalmologist determined that 46 were negative, 53 were positive for diabetic retinopathy and 52 were positive for blur, which indicated the presence of one of more of the following conditions -cataract, corneal opacity, vitreous infection, membrane hemorrhage or obstruction in the eye.These fundus images were also analyzed by the AI system.The sensitivity and specificity of AI based system was 86.79% and 91.30% for diabetic retinopathy and 57.69% and 91.30% for blur.Conclusion: This shows that Sevamob's AI based system can be very useful to screen for conditions like diabetic retinopathy, cataract, corneal opacity, vitreous infection and membrane hemorrhage and has the potential to replace an expert ophthalmologist in the future.Sensitivity and specificity also depend on the threshold used by our AI system.