Asymptomatic atrial fibrillation (AF) is common in the ageing population and increases the risk of suffering an ischemic stroke. Earlier identification and appropriate treatment of AF may decrease stroke morbidity and mortality. Consumer-facing dry-electrode ECGs recorded from personal ECG devices have been identified as a possible solution to assess abnormal heart rhythms, such as AF, at scale. This study assessed the performance of a deep neural network (PulseAI, Belfast, United Kingdom) used in conjunction with a dry electrode ECG sensor device (RhythmPad, D&FT, United Kingdom) to automatically detect AF.
The study is registered at ClinicalTrials.gov, with the unique identifier NCT02401451. Simultaneous pairs of 12-lead ECGs (MAC 550, GE Marquette, Milwaukee, USA) and single-lead dry electrode ECGs were collected from 900 patients. The 12-lead ECGs were manually overread and used as reference diagnoses. Fifteen patients (1.7%) were confirmed with AF and had an interpretable 12-lead and single-lead dry electrode ECG recording. The dry-electrode ECGs were analysed by the deep neural network, and performance was compared to the 12-lead interpretation.
Overall, the deep neural network algorithm yielded a sensitivity of 94% (95% CI, 70%-100%), specificity of 99% (95% CI, 99%-100%) and positive predictive value of 71% (95% CI, 53%-85%) for detection of AF episodes.
The PulseAI neural network, when coupled with dry electrode ECG sensors, allows for large-scale and low-cost screening for AF. Widespread implementation of this technology may allow for earlier detection, treatment and management of patients with AF.