With two million people suffering from the loss of some or all of their sight in the UK, eye care professionals have long being looking for ways to improve the early diagnosis of diseases such as diabetic retinopathy and age-related macular degeneration. This is what led Pearse Keane, Consultant Ophthalmologist at London’s Moorfields Eye Hospital, to approach DeepMind, Google’s artificial intelligence company, to collaborate on the use of technology to aid the analysis of eye scans with a view to spotting and understanding eye disease in new ways.
What are the aims of this exciting new project?
The project, which will last for five years, will see Moorfield pass on anonymised scans of the backs of peoples’ eyes and optical coherence tomography (OCT), which gives much more detail than standard scans are able to provide. As DeepMind’s systems learn automatically, the hope is that large data sets that would overwhelm most optician software systems can be analysed and used to help improve future care.
According to Mustafa Suleyman, who co-founded DeepMind, the project was set up to explore how artificial intelligence can help solve some of the biggest problems faced by society. He went on to say that he believed the project’s work has the potential to benefit patients across the NHS in the future.
How will the DeepMind project impact existing optician management systems?
Current systems, such as the optician software through Blinkoms that is used in many ophthalmologists’ practices, already use OCT scans for diagnostic work. Because of the high level of detail contained in these scans, however, the systems can often become overloaded and doctors sometimes struggle to see the patterns and trends contained in the scans that would allow them to make an accurate diagnosis.
The good news is that if DeepMind’s neural networks do in fact “learn” to spot patterns and trends in these enormous data sets, their output could be utilised by opticians to make early diagnoses of some of the most common eye diseases and herald a turning point in the reduction of sight loss worldwide. It is possible that the future will see machine learning forming an important part of the technology used for everyday checkups, patient monitoring and risk assessments in all areas of medicine.
