How DeepMind Is Being Used to End Avoidable Blindness

As many of you may be aware, for our founder, Tej Kohli, robotics and AI have a vital part to play in countless areas of humanitarian and philanthropic endeavour – including in the work of the Tej Kohli Cornea Institute.

Now, DeepMind, Google’s pioneering AI project, is being used to analyse retinal scans for signs of disease – and it may already be better at it than human experts. Here, the Tej Kohli Cornea Institute explores how this application could help doctors, and further Tej Kohli’s personal mission to eradicate curable blindness by the year 2020.

AI: Our Most Powerful Diagnostic Tool?

The potential application of AI as a tool for diagnosing medical conditions has been one of the technology’s most discussed and anticipated possibilities. In particular, machine learning algorithms have already proved to be capable of diagnosing tuberculosis in x-ray images, as well as certain types of heart disease and cancers, with such a degree of accuracy that they are already rivalling their human counterparts.

Now, scientists from DeepMind, in partnership with the National Health Service and London’s Moorfields Eye Hospital, have turned their attention to using similar tools to help diagnose the health of our eyes.

The technology works through the application of machine learning – the branch of AI whereby complex algorithms are used to analyse inputted data, and then calculate new data sets and models based on their analysis. In the case of this research, the data that was inputted was thousands of anonymised retinal scans that had been labelled as demonstrating varying signs of disease by medical professionals.

Armed with this data, the algorithm was then able to teach itself to recognise and diagnose signs of eye disease in human patients. In fact, it was able to do so significantly more quickly, and with a greater degree of accuracy, than the same task performed by human specialists. The researchers have now submitted their findings to a medical journal to be peer reviewed.

What Does This Mean for Avoidable Blindness?

The implications of DeepMind’s research are significant and confirm the belief stated last year by Tej Kohli that machine learning algorithms have a productive part to play in the work of the Tej Kohli Cornea Institute. Mr Kohli was referring specifically to machine learning’s potential for organising and analysing patient records, but this research may prove to have more far-reaching implications, as it will allow medical professionals to identify health issues in eyes, and begin the appropriate treatment, with a much greater degree of accuracy and efficiency.

While other fields of research, including the development of biosynthetic corneas, have proven to be promising, diagnosing curable blindness quickly and efficiently remains one of the key factors in tackling the overall issue.

Following the success of the partnership between DeepMind and Moorfields, there are further plans for partnerships with University College London to analyse radiotherapy scans, and with Imperial College London to analyse mammograms.

While there is still a ways to go to help the millions of people across the globe who suffer from curable corneal blindness, this latest development is a step towards a world in which Mr Kohli’s vision becomes a reality – improving the sight, and quality of life, of countless people.