We helped DeepMind create an AI system to help clinicians fast-track patients with serious eye diseases.
Collect and prepare the data required to train a pioneering healthcare artificial intelligence algorithm
Moorfields Eye Hospital, the leading provider of eye health services in the UK and a world-class centre of excellence for ophthalmic research, was working with DeepMind on a pioneering artificial intelligence (AI) research project. The aim was to develop an algorithm capable of accurately spotting signs of conditions such as age-related macular degeneration (AMD) in de-personalised eye scans. AMD is the most common cause of blindness in the UK.
Reviewing eye scans to assess whether a patient has a particular condition currently requires time and input from highly skilled clinicians. With large numbers of referrals, this can take time. If a computer was able to accurately spot those in need of quick treatment, it would free up clinicians to see more patients, and enable them to prioritise patients effectively. Being treated more quickly could save someone’s sight.
To train the AI algorithm, Moorfields Eye Hospital asked Softwire to help it bring together and prepare the necessary data from a number of sources. It also asked us to create a new in-house research support tool. DeepMind provided Moorfields with funding for the research, part of which covered Softwire’s work.
We aggregated and transformed the required data to train the AI algorithm, and built Moorfields Eye Hospital a transformational research support system
Firstly, we worked with Moorfields’ information governance teams to ensure all necessary data was de-identified in line with local and national guidelines. We then brought this together and transformed it into the correct format, ready for the development of AI algorithms.
Second, we created the in-house system to search the newly aggregated data, to support Moorfields Eye Hospital’s wider eye condition research work. The new tool enables researchers to search the enormous volume of data and quickly find relevant information, a process that was previously manual and highly labour-intensive.
DeepMind created an AI system that could interpret eye scans as accurately as world-leading expert doctors. This has the potential to transform eye care for the millions of people affected by preventable or curable sight loss worldwide.
We quickly provided the data required to train the AI algorithm, enabling Moorfields Eye Hospital and DeepMind to produce their research paper in the planned timescales. Pearse Keane, NIHR Clinician Scientist and Honorary Consultant Ophthalmologist at Moorfields Eye Hospital, underlines the importance of Softwire to this process: “ Having Softwire involved, working with our in-house information governance department, gave us a lot of security and trust in how robust the system was in this regard.”
The results, published online in Nature Medicine (and covered on the BBC and in most national newspapers including the front page of the London Evening Standard), show that the DeepMind AI system can quickly interpret eye scans from routine clinical practice with unprecedented accuracy. It can correctly recommend how patients should be referred for treatment for over 50 sight-threatening eye diseases – as accurately as world-leading expert doctors.
Meanwhile, the research support system we built means Moorfields’ researchers are now able to base their work on much larger sample sizes than was previously feasible.
Keane explains the impact this has already had: “We can now really take advantage of the fact that Moorfields is the largest eye hospital in Europe by using all our data. Previously, we would do studies of around 100 patients, and gathering the data could take up to six months. Now, we can gather data on 10,000 or even 100,000 patients in minutes, meaning the insights we get are much more powerful. We’ve had it for a few months and it’s already making a difference: we hope to have at least 10 abstracts to submit to a high-profile conference within the next two months, based on the work from Softwire.”
He concludes: “I would love to work with Softwire again in the future. I feel that if the NHS had access to this type of expertise, at scale, on a regular basis, it would have huge benefits for the UK and the world.”
For more details about this project, you can read DeepMind’s blog post.