Google's DeepMind AI could soon be diagnosing eye conditions

By Dr. Ananya Mandal, MDAug 14 2018

London-based DeepMind which is owned by Google’s parent company Alphabet, is coming up with Artificial Intelligence based devices that can help diagnose eye conditions accurately.

The recent progress report appears in the latest issue of the journal Nature Medicine. It says that DeepMind, in collaboration with Moorfields Eye Hospital in London has successfully created algorithms that could help diagnose 50 of the more serious eye conditions that could lead to vision loss. The accuracy of the device is equivalent to expert ophthalmologists. The device can not only detect the eye conditions, but it can also sort out the cases that are more serious and need more urgent care. Thereafter the device can outline the therapeutic measures that could be adopted to treat the patient.

This latest project from DeepMind began a couple of years ago when the company started incorporating these algorithms into the machine. The machine was fed with thousands of earlier scans that were in the databases and asked to diagnose them for serious eye conditions. Dr Dominic King, medical director, DeepMind Health explained the process of how they trained the machine to read the scans. He said, “We used two neural networks, which are complex mathematical systems which mimic the way the brain operates, and inputted thousands of eye scans. They divided the eye into anatomical areas and were able to classify whether disease was present.” For initial testing they developed five machine-learning systems that were fed 877 clinical OCT scans. These machines then created the maps of the OCT scans. The five maps were then analysed by a second series of five machine-learning systems. These second learning systems were trained on 14,884 OCT scans from 7,621 patients. Once the machine gave a result it was then trained to provide a referral decision.

The researchers found that the machine was capable of detecting these serious vision-threatening eye conditions with 94 percent accuracy. The team says that this machine could soon change how eye exams take place globally.

The researchers explain that diagnosing eye conditions from scans manually is a time consuming and tedious job and is also quite complex with a high rate of errors. Further, there is a steep rise in the aging population and this means that the prevalence of eye conditions is also on the rise. This can mean a higher burden on the ocular health care system. AI can shorten the process as well as provide the accuracy needed for diagnosis and detection of these conditions. Pearse Keane, consultant ophthalmologist at Moorfields, in a statement said, “The number of eye scans we're performing is growing at a pace much faster than human experts are able to interpret them. There is a risk that this may cause delays in the diagnosis and treatment of sight-threatening diseases, which can be devastating for patients.” He added that using AI could mean earlier and more accurate diagnosis and earlier treatment. This could save sights in many patients.

The AI developed by DeepMind as of now is useful as a type of eye scanner. Researchers add that the machine is compatible with any model. This means that that there are no hardware restrictions for its use and its use could be widespread. The AI can also be used when the machine is being replaced or updated, they add. The AI does not only give out the diagnosis but also details the steps to the diagnosis for the clinician. This can help the doctor go through the findings in details before deciding upon the next course of action.

There are several fields in healthcare where AI is making forays. This year in June, Babylon Health for example, came up with AI incorporated machines for general practitioners that would help them diagnose several diseases. AI was also found to be more effective in classification of the anatomy of the heart by scanning it ultrasonically. This can help emergency heart attack cases.

Mustafa Suleyman head of DeepMind Health, said in a press statement that this was a “ground breaking” research and in time it could, “transform the diagnosis, treatment, and management of patients with sight threatening eye conditions around the world.” For this newly developed AI device to be used routinely, it needs to undergo clinical trials and then be approved by the regulatory bodies.