AI Powers Personalized Medicine Approach to Detecting Hypoglycemia

Most patients with diabetes still rely on finger-prick blood tests to measure their glucose levels. Some use wearable continuous glucose monitors, but even those typically require twice daily calibrations against the finger-prick method. It has been known for a while that an electrocardiogram (ECG) can show signs of abnormal glucose levels, but the reliability of this method has been hampered by the individual responses that different patients exhibit in their ECG waveforms.

Now, researchers at the University of Warwick in the UK have shown that they can detect dangerous hypoglycemic events with an accuracy of 82% by simply analyzing ECG graphs generated by a commercially available wearable device (Medtronic Zephyr BioPatch HP). This is comparable to the current capabilities of invasive continuous glucometers. If these results are confirmed in more extensive studies, the technology may help with pediatric patients, aid in preventing hypoglycemia during sleep, and lead to a major improvement in the quality of life for many patients with diabetes.

The new system relies on a deep learning approach that can identify how the heartbeats of different people are affected by hypoglycemia. The ECG signals of each patient are recorded and hypoglycemic events noted using traditional means. Given enough data, the Warwick system notices the ECG signal biomarkers that come up in cases of hypoglycemia, which it can spot during subsequent analysis. Making things easier for clinicians is an annotative approach that makes sure that the aspects of the ECG that point to hypoglycemia are highlighted.


The solid lines represent the average heartbeats for two different subjects when the glucose level is normal (green line) or low (red line). The red and green shadows represent the standard deviation of the heartbeats around the mean. A comparison highlights that these two subjects have different ECG waveform changes during hypo events. In particular, Subject 1 presents a visibly longer QT interval during hypo, while Subject 2 does not. Vertical bars represent the relative importance of each ECG wave in determining if a heartbeat is classified as hypo or normal. From these bars, a trained clinician sees that for Subject 1, the T-wave displacement influences classification, reflecting that when the subject is in hypo, the repolarisation of the ventricles is slower. In Subject 2, the most important components of the ECG are the P-wave and the rising of the T-wave, suggesting that when this subject is in hypo, the depolarisation of the atria and the threshold for ventricular activation are particularly affected.

Study in Scientific Reports: Precision Medicine and Artificial Intelligence: A Pilot Study on Deep Learning for Hypoglycemic Events Detection based on ECG

Via: University of Warwick

Source