Artificial intelligence taught to determine the level of blood glucose ECG

Monitoring of blood sugar levels is of great importance primarily for patients with diabetes. But people who put such a diagnosis, it is important to monitor this indicator.

In recent years, scientists offer more and more alternative non-invasive technologies. But while medicine can’t meet patients ‘ need for inexpensive and simple device that allows you to continuously monitor glucose levels without damaging the skin.

a promising New technology presented to the international scientific group. Scientists propose to use artificial intelligence to detect hypoglycemic events in the electrocardiogram signals (ECG).

Explain what the term “hypoglycemic event” is indicated by a drop in blood sugar levels. The jump may be due to receiving too many medications from diabetes, skipping meals or dehydration, alcohol abuse, excessive exercise and so on.

If the system of regulation of glucose level continuously gives the same crashes, a person may develop hypoglycemic syndrome. But a single event of this kind may have serious consequences, up to hypoglycemic coma.

currently, monitoring of hypoglycemic events, using invasive techniques and the appropriate sensors you need “pay in blood”.

“Prick fingers are never pleasant, and in some circumstances they are extremely burdensome, especially for children, – explains the head of the research group Pecchia Leandro (Leandro Pecchia) from the University of Warwick. Our innovation is to use artificial intelligence for automatic detection of hypoglycemia with the help of ECG. This is important because heart rate can be tracked under any circumstances, including during sleep”.

a Previous study that explored the possibility of tracking the level of glucose in blood according to the ECG, failed because the AI is faced with a huge variety of indicators that explain learnedE. none of the systems of machine learning are unable to find universal patterns that correlate with the glucose level in the blood of people.

the Key breakthrough team Peccia is to develop a more advanced system of deep study that analyzes cardiac rhythms of each individual patient.

the article presented in the journal Scientific Reports, authors reported two pilot studies for training the AI, we used data of healthy volunteers.

the trial was attended by eight people. The monitoring was conducted around the clock for 14 days in a row.

four of the volunteers were recorded for at least two hypoglycemic events during a minimum of two nights.

Tests have shown that the new AI system at this stage identifies hypoglycemic events with an accuracy of 82%. Approximately the same percentage have invasive wearable sensors for continuous glucose monitoring (continuous glucose monitor).

the figure below shows the result of the algorithm: the green line shows normal glucose levels, red – low. Horizontal line represents the threshold value of glucose is four moles per liter. The gray area surrounding the continuous line, represents the range of measurement error.

the Curve constructed by the algorithm of deep learning.Illustration of University of Warwick/translation “Conduct.Science”.

Scientists have provided another vivid example. The image below shows how the ECG changes of two people during a hypoglycemic event. The solid lines show the average value of the rate of heartbeat, when the glucose level is normal (green) or low (red). Red and green shade indicates the standard deviation of heart rate from the mean value.

An example of how much may vary individual ECG data, signaling the drop in the level of glucose in the blood.Illustration the University of Warwick.

it is Easy to see that the changes of the forms of “waves” of the ECG during a hypoglycemic event two people are completely different. This factor was an obstacle in previous works, considers the team Pecka.

we would Add that such models, we build an artificial intelligence useful for clinicians. Analyzing the position of teeth (deviation up or down from the isoelectric line), the doctor will be able to conclude the work of the Atria and ventricles during a hypoglycemic event. This is important data that may affect subsequent therapy and treatment is personalized.

In the near future the authors intend to conduct more extensive clinical studies that should confirm the effectiveness of new technologies and possibly help in its refinement.

by the Way, before “Conduct.Science” ( told me about the paper sensor, the patch and a “smart” contact lenses that can replace a glucometer. The scientists also found that to reduce the risk of Hypo – and hyperglycemia help dogs.

Text: To.Science