A research team from The Chinese University of Hong Kong (CUHK), in collaboration with Beijing Tongren Hospital, has developed an innovative artificial intelligence model designed to streamline ophthalmic diagnosis. This breakthrough stands out for its ability to predict intracranial tumors directly from retinal images.
Vision impairment affects millions of people worldwide, with projections estimating that nearly 474 million individuals will experience moderate to severe vision loss by 2050. This challenge is particularly critical in resource-limited regions where access to specialized ophthalmologists is scarce. While existing AI solutions have shown promise, they face limitations, such as reliance on large labeled datasets and a narrow focus on a limited range of eye diseases.
VisionFM overcomes these challenges by leveraging the world’s largest ophthalmic dataset, encompassing 3.4 million images across eight different imaging modalities. This robust foundation enables the model to diagnose eye diseases, predict disease progression, identify systemic biomarkers, and detect intracranial tumors, among other applications.
What sets VisionFM apart is its diagnostic accuracy, comparable to that of expert ophthalmologists, and its few-shot learning capability, allowing it to adapt quickly to new tools or clinical scenarios. This adaptability makes it particularly useful in community and primary care settings, paving the way for early interventions and improved patient outcomes.
The model is already in use in Henan Province, China, for diagnosing common eye diseases, showcasing its practical value in underserved regions. Furthermore, as an open-source tool, VisionFM encourages collaboration and innovation in the healthcare sector, driving the development of accessible and advanced diagnostic solutions worldwide.
Article written by Alita Sharon
11/01/2025
Source:
Open Gov