Artificial Intelligence Can Detect Glucose Levels via ECG

A visualization method then enables the clinicians to establish which part of the ECG signal is significantly associated with a hypoglycemic event in each individual subject.

The AI model is trained with each subject's own dataset, which is comprised of both ECG and glucose recordings as measured by two sensors worn for a period of 8-14 days. The researchers conducted two pilot studies involving eight healthy volunteers, which found that the average sensitivity and specificity of the AI approach for hypoglycemia detection was about 82%, comparable to current continuous glucose monitoring (CGM) device performance. The study was published on January 13, 2020, in Nature Scientific Reports.

“Fingerpicks are never pleasant, and in some circumstances particularly cumbersome. Our innovation consisted of using AI for automatically detecting hypoglycemia via few ECG beats. This is relevant because ECG can be detected in any circumstance, including sleeping,” said senior author Leandro Pecchia, PhD, of the University of Warwick School of Engineering. “Our approach enables personalized tuning of detection algorithms and emphasizes how hypoglycemic events affect ECG. Based on this information, clinicians can adapt the therapy to each individual.”

Hypoglycemia can cause pronounced physiological responses as a consequence of autonomic activation, principally of the sympatho-adrenal system, which results in the release of epinephrine (adrenaline). The autonomic stimulus provokes hemodynamic changes in order maintain a supply of glucose to the brain and promote the hepatic production of glucose. Hemodynamic changes associated with hypoglycemia include an increase in heart rate and peripheral systolic blood pressure, a fall in central blood pressure, reduced peripheral arterial resistance, and an increase in myocardial contractility, stroke volume, and cardiac output.

Related Links:
University of Warwick
University of Napoli Federico II
Western University