The World’s First Artificial Intelligence-based System for Detecting Multiple Fundus Lesions Using Ultra-widefield Fundus Images
Source: Department of Artificial Intelligence and Big Data, Zhongshan Ophthalmic Center
Written by: Department of Artificial Intelligence and Big Data, Zhongshan Ophthalmic Center
Proofread by: Jiawei Wang
Edited by: Xianjing Wei
Manual fundus examination after mydriasis is a time-consuming and labor-intensive process. If the fundus diseases are not detected and treated at an early stage, it may lead to irreversible visual function loss. Ultra-widefield fundus (UWF) imaging can provide a 200 to 240-degree panoramic retinal view, which is five times larger than traditional fundus imaging with retinal scope of 45 degrees. Utilization of ultra-widefield fundus imaging will increase detection rates of fundus diseases throughout the entire retina. Based on approximately one hundred thousand ultra-widefield fundus images, the team from the Department of Artificial Intelligence and Big Data at Zhongshan Ophthalmic Center has developed the world’s first artificial intelligence-based system for fundus lesions screening. This system can be used to detect lattice degeneration, retinal breaks, retinal detachment, retinal hemorrhage, retinal exudation, drusen, and glaucomatous optic neuropathy in real time, with an accuracy over 98%. More importantly, based on the features of the detected lesions, the system can provide medical guidance automatically, which can help patients to take preventative action before worse effects develop.
Overall, this system can be used to screen fundus lesions as a part of ophthalmic health evaluations in physical examination centers or community hospitals that lack ophthalmologists, or be deployed in hospitals with a large number of patients to assist ophthalmologists by avoiding the need for further examination of evidently normal eyes. Additionally, this system can be applied to detect peripheral retinal lesions in patients who cannot tolerate a dilated fundus examination, such as those with a shallow peripheral anterior chamber.
Figure. Artificial Intelligence-based System for Fundus Lesions Screening via Ultra-widefield Fundus Images
At the beginning of 2020, their latest research, entitled “Deep Learning for Detecting Retinal Detachment and Discerning Macular Status Using Ultra-widefield Fundus Images” was published online in Communications Biology, the sub journal of the Nature. This work was supported by the National Key R&D Program of China, the Key Research Plan for the National Natural Science Foundation of China in Cultivation Project, National Natural Science Foundation of China, the Science and Technology Planning Projects of Guangdong Province, Guangdong Science and Technology Innovation Leading Talents. PhD student Zhongwen Li and research assistant Chong Guo are co-first authors, Prof. Haotian Lin is the corresponding author.
At present, the team has completed negotiations with several domestic and foreign companies on translational plans of the artificial intelligence system, hoping this system can be applied in real world settings as soon as possible to bring benefits to all people.