A photograph can reveal much but doesn’t tell the whole story of a person. Similarly, AI analysis of an eye image has its limitations. This invention begins with color fundus photographs and employs image-processing techniques alongside neural networks to emphasize structures linked to glaucoma. A biomedical image-segmentation network detects four key features: retinal blood vessels, the macula, optic disc, and optic cup. In addition to the photo, it accesses the patient’s electronic medical records, which include history, lifestyle, medications, and exam results. Natural-language processing and data extraction then turn this information into a glaucoma knowledge graph, a structured map of significant medical relationships. The system combines this data with features from the retinal image to help predict glaucoma. Think of it as giving AI both the image and the patient’s case file. This approach is valuable because it integrates visual clues with personal medical data, going beyond simple eye image analysis. Since early detection and understanding disease progression are vital to preventing vision loss from glaucoma, this broader context enhances AI’s usefulness for clinicians rather than just serving as an image reader.
Adding more context to the patient's eye photograph
Patent number: CN 118711790 A
Inventor(s): K. Zhang, Y. Gao, and Z. Zou
Citation: K. Zhang, Y. Gao, and Z. Zou, “青光眼辅助预测方法及系统” [Glaucoma-assisted prediction method and system], China Patent Application Publication CN 118711790 A, Sep. 27, 2024.
