A team of scientists in Lata Medical Research Foundation, Nagpur, have developed an artificial intelligence (AI) algorithm that can accurately predict diabetes and pre-diabetes from ECG. The AI algorithm has been derived from the features of individual heartbeats recorded on an electrocardiogram (ECG).
The team included clinical data from 1,262 individuals. A standard 12-lead ECG heart trace lasting 10 seconds for each participant was performed. A predictive algorithm named DiaBeats was generated by combining 100 unique structural and functional features for each of the 10,461 single heartbeats recorded.
The DiaBeats algorithm quickly detected diabetes and pre-diabetes based on the size and shape of individual heartbeats. The algorithm did so with an overall accuracy of 97%, irrespective of factors such as gender, age, and co-existing metabolic disorders.
Read More: Artificial Intelligence In Chronic Disease Management
Vital ECG features consistently matched the known biological triggers that imply cardiac changes that are common for diabetes and pre-diabetes. The method could be used to screen for the disease in settings with low resources if validated in more extensive studies, the team said.
In theory, the study provides a relatively non-invasive, inexpensive, and accurate alternative to the current diagnostic methods. This alternative can effectively detect diabetes and pre-diabetes early in its course. Despite that, the adoption of this algorithm into regular practice will need solid validation on external and independent datasets.
The researchers admitted that the participants of the study were all at high risk of diabetes and other metabolic disorders. Therefore, it is unlikely to represent the general population. Also, DiaBeats was slightly less accurate for patients taking prescription medications for high blood pressure, diabetes, and high cholesterol.