29 Jan New system uses AI to test for blood sugar
A team of scientists recently developed an AI system that can detect low blood sugar or hypoglycemia.
The researchers hope that this system will enable patients to measure their blood glucose levels without the need for an invasive method called a finger prick test.
The team recently published the results of a pilot study in the journal Scientific Reports.
For many people, measuring blood glucose currently involves pricking a finger with a needle attached to a device. The blood sample is then analyzed by a continuous glucose monitor (CGM), which often needs to be calibrated at least twice a day.
The process can be difficult, uncomfortable, and inconvenient, especially for children and people who need to test their blood in the middle of the night. As a result, some people are unable to measure their levels as often or as accurately as necessary.
The researchers behind the current study hope that a noninvasive method will help improve compliance rates, particularly among those who need to monitor their glucose levels closely, such as people with diabetes.
The new AI technology was developed at the University of Warwick, in the United Kingdom, and it can detect hypoglycemia using electrocardiogram (ECG) signals from the heart.
In their study, the scientists demonstrated that this new technology is accurate 82% of the time, a rate similar to that of current CGM systems. Senior study author Leandro Pecchia, Ph.D., an associate professor of biomedical engineering at the university, commented:
“Our innovation consisted in using [AI] for automatic detecting [of] hypoglycemia via few ECG beats. This is relevant because ECG can be detected in any circumstance, including sleeping.”
Hypoglycemia affects the electrophysiology of the heart, and because it has slightly different effects on each individual’s heart, an AI system makes it possible to monitor glucose levels in a highly personalized way.
In the recent pilot study, the team used AI to automatically detect nocturnal hypoglycemia from just a few heartbeat signals recorded by a wearable device. The study included healthy individuals, whom the scientists monitored for 24 hours a day for 14 consecutive days.
This study was unique because the scientists monitored the participants’ glucose levels individually, whereas previous trials had analyzed results from the participants as a group.
The authors believe that their new approach captures the considerable diversity in ECG signals among individuals, which previous trials could not accurately incorporate.
The wave-shaped readouts from an ECG machine give a detailed picture of how the heart is behaving; each section of the wave provides information about specific cardiac events, such as heartbeats.
The authors behind the current study developed a way to visualize precisely which part of the ECG wave is associated with a hypoglycemic event.
This could result in a real-time alarm system that alerts individuals if their blood sugar levels change dramatically. Having such an early warning could drastically shorten the amount of time that a person experiences hypoglycemia, which can be very dangerous, especially for people with diabetes.
The team’s new method is one example of precision medicine that could vastly improve the way that people manage diabetes. While there is still some way to go before this technology becomes available, the initial results are promising.
If successful, the technology tested in the present study could pave the way for many more uses of AI and electrophysiology of the heart. It could also possibly be used to manage a variety of disorders that result from changes in the blood, with highly personalized precision.