AI-Based Real-Time Hyphema Detection Using Deep Learning

Abstract

Eye-related disorders are rapidly increasing due to modern lifestyle factors such as prolonged screen exposure, accidental injuries, and delayed medical consultation. Among these conditions, Hyphema is a serious ocular issue characterized by the presence of blood in the anterior chamber of the eye. If not detected at an early stage, it may lead to complications such as blurred vision, increased intraocular pressure, or even permanent vision loss. Traditional diagnosis of Hyphema depends heavily on clinical examination by ophthalmologists, which may not always be immediately accessible, especially in rural or resource-limited areas. This creates a need for an automated, fast, and reliable detection system that can assist users in early screening. This project presents an AI-based real-time hyphema detection and ocular health assistance system using deep learning techniques. A Convolutional Neural Network (CNN) model is utilized to analyze eye images and classify them into normal and hyphema-affected categories. The system is trained using eye image datasets along with preprocessing and augmentation techniques to enhance performance and generalization. The proposed system is deployed as a web application using Flask, allowing users to capture or upload eye images and receive instant predictions along with confidence scores. Additionally, Grad-CAM visualization is used to highlight affected regions in the eye, improving interpretability. The system also includes an AI assistant and nearby hospital locator for user guidance. Experimental results indicate that the system achieves high accuracy and can effectively support early detection. This solution demonstrates the potential of deep learning in real-time healthcare applications and can be extended further with advanced integrations such as offline detection, multilingual support, and real-time medical consultation systems

Introduction

The text describes the development of an AI-based system for real-time detection of hyphema, a serious eye condition involving bleeding in the anterior chamber that can lead to vision loss if not treated early. Traditional diagnosis depends on ophthalmologists and specialized equipment, which is slow, inaccessible in rural areas, and unsuitable for immediate self-assessment. Existing digital and AI-based eye disease systems also face limitations such as lack of real-time use, poor accessibility, limited interpretation tools, and weak user support features.

To address these issues, the proposed solution is a web-based AI system that allows users to capture or upload eye images for automatic analysis. A trained Convolutional Neural Network (CNN) classifies images as normal or hyphema-affected, enabling fast preliminary screening. The system enhances interpretability using Grad-CAM visualizations, highlights affected regions, and improves user understanding of predictions. It also provides additional support features such as basic medical guidance and location-based suggestions for nearby eye hospitals.

Conclusion

Detecting eye-related conditions such as hyphema using traditional methods often depends on clinical examination by specialists, which can be time-consuming and not always immediately accessible. In many cases, the lack of early detection and timely medical attention can lead to serious complications, including vision loss. To address these challenges, this paper presented an AI-Based Real-Time Hyphema Detection & Ocular Health Assistance System, designed to provide a fast, accessible, and user-friendly solution for early eye disease detection. The proposed system enables users to capture eye images using a camera or upload images from their device for analysis. The system is developed using modern technologies including Python, Flask, TensorFlow, OpenCV, and SQLite, ensuring efficient processing, reliable performance, and structured data management. The integration of a deep learning-based Convolutional Neural Network (CNN) allows automatic feature extraction and classification of eye images into normal and hyphema-affected categories. One of the key contributions of the system is the inclusion of Grad-CAM visualization, which provides heatmaps highlighting the affected regions in the eye. This improves transparency and helps users understand how the model makes predictions. In addition, the system includes supportive features such as a rule-based AI assistant for guidance, nearby hospital recommendations using location-based services, and a vision screening module for basic eye health evaluation. These features enhance the overall usability and extend the system beyond simple detection. The implementation and testing results demonstrate that the system provides accurate predictions, fast response time, and a smooth user experience. It reduces dependency on immediate clinical diagnosis for initial screening and enables users to take timely action. The web-based interface ensures easy accessibility, while the history module allows users to track their previous results effectively.

References

[1] R. Szeliski, Computer Vision: Algorithms and Applications, Springer, 2010. [2] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” Advances in Neural Information Processing Systems, 2012. [3] K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual Learning for Image Recognition,” IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016. [4] R. R. Selvaraju et al., “Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization,” IEEE International Conference on Computer Vision (ICCV), 2017. [5] G. Bradski, “The OpenCV Library,” Dr. Dobb’s Journal of Software Tools, 2000. [6] F. Chollet, Deep Learning with Python, Manning Publications, 2017. [7] M. Abadi et al., “TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems,” 2015. [8] M. Grinberg, Flask Web Development: Developing Web Applications with Python, O’Reilly Media, 2018. [9] R. Walton, M. Von Hagen, A. Grigorian, and D. Zarbin, “Management of Traumatic Hyphema,” Survey of Ophthalmology, vol. 47, no. 4, pp. 297–334, 2002. [10] S. Renusree, K. S. Sreeja, and M. N. Nair, “Deep Learning Based Eye Disease Detection Using Convolutional Neural Networks,” in Proceedings of IEEE International Conference on Intelligent Computing and Control Systems (ICICCS), 2020.

Copyright

Copyright © 2026 Bethapudi Haritha, Gollavelli Manju Bhargavi, Chanda Sunayana , Kunchala Venkata Madhuri . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.