DeepMind’s AI can detect eye disease as accurately as world-leading doctors
System jointly-developed with Moorfields Eye Hospital can analyse 3D scans and recommend treatment within seconds
DeepMind has announced its artificial intelligence (AI) system can now diagnose sight-threatening eye conditions as accurately as the world's best clinicians.
Working with London-based Moorfields Eye Hospital and UCL's Institute of Ophthalmology, the results of the preliminary research, published in Nature, can pave the way for the rollout of AI systems in hospitals across the country, the Alphabet subsidiary has said.
The AI system aims to drastically reduce the time doctors spend studying thousands of optical coherence tomography (OCT) eye scans and can diagnose patients within seconds. Moreover, Moorfields Eye Hospital says the technology can recommend the best path of treatment for more than 50 diseases with 94% accuracy.
"The number of eye scans we're performing is growing at a pace much faster than human experts are able to interpret them," said Dr Pearse Keane, consultant ophthalmologist at Moorfields Eye Hospital.
"There is a risk that this may cause delays in the diagnosis and treatment of sight-threatening diseases, which can be devastating for patients. The AI technology we're developing is designed to prioritise patients who need to be seen and treated urgently by a doctor or eye care professional.
"If we can diagnose and treat eye conditions early, it gives us the best chance of saving people's sight. With further research, it could lead to greater consistency and quality of care for patients with eye problems in the future."
OCT scans produce a 3D map of the back of the eyes, which trained clinicians must manually examine to identify the key features of eye diseases and make a diagnosis. Conventionally, they have been challenging to interpret, and have proven time-consuming, but DeepMind says its technology can analyse these scans within seconds and make recommendations instantly.
Clinicians studied the same OCT scans that were fed into the DeepMind system and made independent recommendations to establish whether the AI was making correct referrals. The study concluded this technology was able to make the correct referral with a very high level of accuracy.
How a user interacts with DeepMind's OCT viewer
The AI system has also been structured in such a way as to avoid the "black box problem" - given how important it will be for clinical staff and patients to understand how these decisions are reached. The system, which operates using two neural networks, shows visuals of the eye disease it has identified, and the level of confidence it has in its recommendations as a percentage.
"While we're incredibly proud of this progress, this initial research would need to be turned into a product and then undergo rigorous clinical trials and regulatory approval before being used in practice," DeepMind said.
"But we're confident that, in time, this system could transform the diagnosis, treatment and management of eye disease. Our partners at Moorfields want our research to help them improve care, reduce some of the strain on clinicians, and lower costs - all at the same time. So we've also worked hard on what comes next."
This milestone marks the end of the first phase of a partnership struck between Moorfields and DeepMind two years ago. DeepMind is hoping, following further research and approval, that its AI technology can be deployed in Moorfields' more than 30 UK hospitals and community clinics, serving 300,000 patients a year.
The 3D skills report
Add 3D skills to your creative toolkits and play a sizeable role in the digital futureFree Download
The increasing need for environmental intelligence solutions
How sustainability has become a major business priority and is continuing to grow in importanceFree Download
2022 State of the multi-cloud report
What are the biggest multi-cloud motivations for decision-makers, and what are the leading challengesFree Download
Solve global challenges with machine learning
Tackling our word's hardest problems with MLFree Download