Study finds AI can detect hypoglycaemia via ECG without fingerpick test

Research findings detailed Monday in the journal Scientific Reports demonstrated that new artificial intelligence (AI) technology was able to detect nocturnal hypoglycaemia using raw electrocardiogram (ECG) signals recorded with non-invasive wearable sensors. Study author Leandro Pecchia remarked that "our innovation consisted in using AI for [automatically] detecting hypoglycaemia via few ECG beats. This is relevant because ECG can be detected in any circumstance, including sleeping."

Investigators were looking to spot low glucose levels in healthy individuals based on ECG signals and actigraphy, recorded continuously during a period of 14 nights for each subject. ECG, actigraphy and continuous glucose monitoring (CGM) readings were recorded using commercial wearable sensors. Four healthy volunteers met the inclusion criteria. Data recorded for a participant during the first days were used for training the AI model, which was tested using data from the same subject acquired in the remaining days.

Two pilot studies found that the average sensitivity and specificity was approximately 82% for hypoglycaemia detection, which the authors said is comparable with current CGM performance. They also presented a method that allows clinicians to visualise which part of the ECG signal is significantly associated with the hypoglycaemic event in each subject, "overcoming the intelligibility problem of deep-learning methods."

Pecchia said "our approach enables personalised tuning of detection algorithms and emphasises how hypoglycaemic events affect ECG in individuals." However, he added that further research is required to confirm these results in wider populations.

Did you like this article?