All articles Health Tech Business

How smart watches and AI can help people with diabetes



Artificial Intelligence may prove to be a game-changer for people with diabetes by doing away with the need for blood-tests or devices inserted under the skin.

Medicine is advancing at an unprecedented rate due to leaps we are making in technology - and it will make a real difference for patients. This week, scientists have reported using artificial intelligence to spot issues for people with diabetes, simply by using data from wearable devices.

People with diabetes can control their high blood-sugar with insulin and other drugs. One of the risks of using insulin is getting dangerously low blood-sugar. It’s known as hypoglycaemia. Although hypoglycaemia is very easy to treat, it can have severe effects if it’s missed, with symptoms ranging from drowsiness to loss of consciousness. Hypoglycaemia is the reason why patients with diabetes must frequently monitor their blood-sugar levels.

At the moment, patients have to measure their blood-sugar levels using pinprick blood tests, or by having continuous glucose monitors inserted under their skin. Whilst good solutions, these are of course painful, disruptive and laborious. However, new research published in Nature suggests patients may now be able to simply wear the right smartwatch.

The research reports that AI was able to learn the patterns in the electrical signals given off by patients’ hearts. This data, collected from an electrocardiogram (ECG) can be recorded in an easy, painless way by some of the latest smart watches. The AI used to recognise the patterns is known as deep-learning AI - essentially it assessed so many sets of ECG data, comparing them with blood levels in patients with diabetes, that it was able to recognise which ECG patterns meant that a patient had a problem.

When a patient has low blood-sugar it changes levels of potassium and hormones in the blood, and the way the brain controls the heart rate. It’s the result of these changes, being picked up by the ECG, that the AI is spotting in the data. But it’s potentially even more impressive than that. Everyone is different and everyone reacts slightly differently to low blood-sugar. The AI could actually recognise the individual variations for each specific patient. It wasn’t just saying pattern X means hypoglycaemia, it was saying patient 1 should be concerned by pattern Y, patient 2 by pattern Z and so on.

Wearables get smarter all the time and more and more people are using them to understand their bodies and their health in ever greater detail. Imagine if they could also be used to help people with long-term conditions. As sensors get smarter, the ability of AI to analyse large amounts of data from multiple sources and apply what it learns specifically for us - our own personalised health information. It’s one of the most promising developments in digital health.

Here at Babylon, we are focused on behaviour change, giving patients the power to manage their conditions and be the healthiest they can. We give food-swaps and personalised activity plans to our patients with diabetes so they can choose the healthiest options and stay active. By pulling data from smart watches that include calories burned, distance covered and steps, we can give more informed advice than before. We want to help patients work with their health professionals to create their own, personalised care plans.

Advances like this AI could be huge for people with diabetes. It’s exciting to think what else could be around the corner for people with conditions that affect them every single day of their lives.

The full study is called Precision Medicine and Artificial Intelligence: A Pilot Study on Deep Learning for Hypoglycemic Events Detection based on ECG