- Researchers from DeepMind, a UK-based artificial intelligence company owned by Google, are progressing towards a clinical decision support product that can accurately identify more than fifty eye diseases and provide treatment recommendations for patients.
Initial results, published this month in Nature Medicine, show that the deep learning tool developed by DeepMind and Moorfields Eye Hospital can identify, diagnose, and recommend care for common clinical eye conditions as accurately as human clinicians.
The algorithm, which is yet to be commercialized, aims to reduce the time to diagnosis for conditions that threaten permanent loss of sight, allowing providers to treat patients or refer them to the correct specialist more quickly.
“Currently, eye care professionals use optical coherence tomography (OCT) scans to help diagnose eye conditions. These 3D images provide a detailed map of the back of the eye, but they are hard to read and need expert analysis to interpret,” explained DeepMind.
“The time it takes to analyze these scans, combined with the sheer number of scans that healthcare professionals have to go through (over 1,000 a day at Moorfields alone), can lead to lengthy delays between scan and treatment – even when someone needs urgent care. If they develop a sudden problem, such as a bleed at the back of the eye, these delays could even cost patients their sight.”
Using deep learning, the research team is able to automatically detect relevant eye features in seconds, and can flag patients at high risk of serious complications to bring their cases to the front of the diagnostic queue.
“This instant triaging process should drastically cut down the time elapsed between the scan and treatment, helping sufferers of diabetic eye disease and age-related macular degeneration avoid sight loss,” the researchers said.
Source: DeepMind / Moorfields
While speed is important, delivering trustworthy and transparent results is an equally critical goal, the blog post says.
The team has focused specifically on avoiding the creation of a “black box” system that does not allow users and patients to understand the decision-making pathways behind a recommendation.
By using a dual layered neural network, DeepMind and Moorfields researchers hope to provide adequate explanation – and trustworthy confidence intervals – to supplement the clinical decision support delivery process.
“The first neural network, known as the segmentation network, analyses the OCT scan to provide a map of the different types of eye tissue and the features of disease it sees, such as hemorrhages, lesions, irregular fluid or other symptoms of eye disease,” the team said.
“This map allows eye care professionals to gain insight into the system’s ‘thinking.’ The second network, known as the classification network, analyses this map to present clinicians with diagnoses and a referral recommendation. Crucially, the network expresses this recommendation as a percentage, allowing clinicians to assess the system’s confidence in its analysis.”
The tool has not yet been put through its paces with clinical trials, and has not received regulatory approval that would allow it to be used in the clinical setting just yet.
However, the researchers are confident that the preliminary results are a positive step towards developing a product that could alter the way eye care providers diagnose and treat their patients.
“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,” DeepMind said.
If approved for general use, the deep learning tool could be used in 30 UK hospitals and community clinics for an initial five-year run.
“These clinics serve 300,000 patients a year and receive over 1,000 OCT scan referrals every day – each of which could benefit from improved accuracy and speed of diagnosis,” said the team.
“For all of us who have worked on this since we signed our agreement with Moorfields in 2016, this is a hugely exciting milestone, and another indication of what is possible when clinicians and technologists work together. We’ll continue to keep you updated as we make progress.”