Skip Ribbon Commands
Skip to main content

Research Bulletin

BONSAI study uses Artificial Intelligence to detect Papilledema and serious health risks from eye images

When patients complain of headache or nausea and vomiting, it is common practice for a doctor to use an ophthalmoscope to look into the eye. They use it to view any change in the appearance of the optic nerve and the blood vessels that pass through it, which could be the root cause of these symptoms.

Papilledema is the swelling of the optic nerve, which is located at the back of the eye. One of the important causes of this swelling is fluid build-up around the brain, leading to an increase in intracranial pressure. However, there could be underlying serious conditions causing this swelling such as brain abscess, brain tumours and infections such as meningitis or encephalitis. Even though Papilledema is rare, it can lead to blindness or even death, hence it is crucial to detect this early.

An international consortium called BONSAI (BONSAI: Brain and Optic Nerve Study with Artificial Intelligence)  has successfully used an AI-based, deep learning system to look at multiple photographs of the back of the eye (optical disk) to infer whether the eye is normal, has papilledema due to intracranial pressure or papilledema due to other abnormalities.

This is a landmark study, because the AI system has shown 96% sensitivity in detecting eye images that show papilledema and that too in a few seconds, at minimal cost! This could be particularly helpful in emergency departments, in neurology practices and even in general practitioner clinics where there may be a lack of trained ophthalmologists to diagnose.

This AI system was developed by a global collaboration of scientists, including researchers from Singapore Eye Research Institute (SERI), Duke-NUS Medical school and Agency for Science, Technology and Research (A*STAR). The study involved over 7,532 patients from multi-ethnic communities from 25 centres around the world and the AI-system was trained to detect papilledema and normal optical discs from over 15,846 ocular fundus photographs.

“Papilledema is rare. In this study, the machine was trained by being exposed to more pictures of papilledema and other optic disc abnormalities than what one specialist can see in a normal practice over a long career.

Transferring such skills to AI-based medical devices may contribute to evolving new ways of practising medicine at a distance, in order to better protect patients and healthcare providers, especially in our COVID-19 era”, said Professor Dan Milea, a neuro-ophthalmologist and a senior consultant at the Singapore National Eye Centre (SNEC) who was part of the study.

The study was published in the prestigious medical journal, New England Journal of Medicine  titledArtificial Intelligence To Detect Papilledema From Ocular Fundus Photographs” in April 2020.

Currently, this AI-based system is being used in a pilot, prospective, real-life study at Singapore National Eye Centre.

It is noteworthy to mention that SNEC’s deep-learning AI software, Selena+ was approved for use in October 2019, to detect the presence of conditions such as diabetic retinopathy, glaucoma and age-related macular degeneration from the photography of a person’s ocular fundus.


Artificial Intelligence to Detect Papilledema from Ocular Fundus Photographs
N Engl J Med 2020; 382:1687-1695
DOI: 10.1056/NEJMoa1917130



Dan Milea, M.D., Ph.D., Raymond P. Najjar, Ph.D., Zhubo Jiang, M.Sc., Daniel Ting, M.D., Ph.D., Caroline Vasseneix, M.D., Xinxing Xu, Ph.D., Masoud Aghsaei Fard, M.D., Pedro Fonseca, M.D., Kavin Vanikieti, M.D., Wolf A. Lagrèze, M.D., Chiara La Morgia, M.D., Ph.D.,
Carol Y. Cheung, Ph.D., et al., for the BONSAI Group*

Image of the Optic Nerve Head/Disk at the back of the eye in a normal patient (left) compared to the Optic disk of a patient with brain tumor showing subtle changes in them (right)