Indian scientists from Innotomy Consulting Data Science Center and Lata Research Center together with colleagues from M&H Research Center in the USA developed a highly accurate and specific DiaBeats algorithm for diagnosing prediabetes from ECG results. The study was published in the journal BMJ Innovations.
The researchers used a set of clinical data from 1,262 people, including their digital electrocardiograms. They divided the dataset into three groups: for algorithm training, for validation, and for independent testing. The prevalence of type 2 diabetes and prediabetes among the participants was 30% and 14%, respectively. DiaBeats algorithm predicted the presence of diabetes or prediabetes with 97.1% accuracy in an independent data set.
Early diagnosis is very important for the prevention of type 2 diabetes and prediabetes. But today its main tools are the glucose tolerance test and the assessment of glycosylated hemoglobin. These are invasive interventions that are difficult for large-scale screening. Diabetes and prediabetes cause structural and functional changes in the heart even before symptoms appear. They can be detected using an EKG, and the scientists’ algorithm could help expand early diagnosis with this simple and inexpensive method.