Tokyo University of Science develops AI model that detects prediabetes from single‑lead ECG data

Summary

Researchers at Tokyo University of Science built an AI model, DiaCardia, that can identify individuals with prediabetes or diabetes using features extracted from 12‑lead electrocardiograms, achieving an AUROC of 0.851 with 85.7% sensitivity and 70% specificity. The team also showed that similar accuracy is possible using only a single‑lead ECG equivalent to data from smartwatch‑style wearables, suggesting a future where early diabetes risk screening could be done non‑invasively at home without blood tests.