IoT & AI-ML Based Portable Non-Contact Retinal Imaging for Glaucoma and Eye Disease Monitoring at Home

A major contributor to irreversible blindness, glaucoma necessitates early detection for efficient treatment. Traditional diagnostic methods rely on specialized clinical equipment, limiting accessibility. This paper introduces a Non-Contact Portable Glaucoma Detection Device that incorporates machine learning (ML) and Internet of Things (IoT) for real-time automated diagnosis. It employs a 20D Double Aspheric Lens to enlarge the retina and the Raspberry Pi Camera takes an image that is processed. A YOLO11x-cls model on Raspberry Pi diagnoses the image and puts it into one of four categories: Normal, Diabetic Retinopathy, Cataract, or Glaucoma. The output is presented on a Touchscreen LCD for immediate feedback to the user. Training was conducted on 5,897 retinal images with TensorFlow, PyTorch, and Ultralights Hub, and performance was optimized using transfer learning and hyperparameter optimization. Testing yielded 94.6 accuracy with 95 precision and 100 recall at 50% confidence threshold. The proposed technique is a cost-effective, portable AI driven system for the detection of early glaucoma with home screening and reduced dependence on specialty centers. Upgradation plans incorporate cloud integration, self-enhancement of images, and wider disease classification to make the diagnosis more accessible.