Detection and categorization of severe cardiac disorders based solely on pulse interval measurements

Cardiac disorders are common conditions associated with a high mortality rate. Due to their potential for causing serious symptoms, it is desirable to constantly monitor cardiac status using an accessible device such as a smartwatch. While electrocardiograms (ECGs) can make the detailed diagnosis of cardiac disorders, the examination is typically performed only once a year for each individual during health checkups, and it requires expert medical practitioners to make comprehensive judgments. Here we describe a newly developed automated system for alerting individuals about cardiac disorders solely based on pulse interval measurements. For this purpose, we examined two metrics of heart rate variability (HRV) and analyzed 1-day ECG recordings of more than 1,000 subjects in total. We found that a newly introduced metric of local variation was more efficient than conventional HRV metrics for detecting premature contraction, and furthermore, that a suitable combination of the old and new metrics resulted in much superior detectability particularly for atrial fibrillation, which requires more attention. Even with a 1-minute recording of pulse intervals, our new detection system had a diagnostic performance even better than that of the conventional analysis method applied to a 1-day recording.