Diagnostic accuracy of multivector smartwatch electrocardiography compared with conventional 12-lead ecg
S Aguiar Munoz, A Tronconi, M Paolini, R RiggeriAbstract
Background
Smartwatch electrocardiography is increasingly used for ambulatory rhythm monitoring but is often interpreted as equivalent to a single planar limb lead, potentially limiting specificity and morphological assessment. A multivector sequential acquisition approach may enhance electrocardiographic interpretation and diagnostic performance.
Purpose
To evaluate diagnostic accuracy, agreement and discriminative performance of multivector smartwatch electrocardiography compared with standard 12-lead electrocardiogram (ECG), and to explore its impact on specificity in asymptomatic individuals.
Methods
Prospective observational study including 139 consecutive patients. Sequential smartwatch ECG recordings were obtained from multiple body positions, including bipolar limb-based vectors and personalised unipolar precordial-equivalent vectors. In phase II, signal acquisition was optimised using an external stabilising adaptor. All recordings were compared with conventional 12-lead ECG interpreted by blinded cardiologists. Sensitivity, specificity and concordance were calculated. Agreement was assessed using Cohen’s kappa coefficient. Discriminative performance was evaluated using receiver operating characteristic curves and area under the curve (AUC).
Results
Mean age was 49±12 years; 57% were male; 89% were asymptomatic at the time of recording. Sinus rhythm detection showed 100% concordance (κ=1.00). For supraventricular arrhythmias, sensitivity was 97% and specificity 93% (κ=0.90; AUC=0.95). Ventricular extrasystoles showed sensitivity 91% and specificity 88% (κ=0.84; AUC=0.92). Conduction abnormalities demonstrated sensitivity 89% and specificity 85% (κ=0.81; AUC=0.89). ST-T alterations showed sensitivity 84% and specificity 82% (κ=0.78; AUC=0.86). Multivector acquisition improved specificity particularly in asymptomatic individuals, reducing false-positive interpretations compared with single-vector analysis. Morphological QRS concordance was higher in personalised unipolar precordial vectors (up to 86%) than in frontal bipolar vectors (65–71%). Signal instability decreased significantly after acquisition optimisation (16 vs 1 case).
Conclusion
Multivector smartwatch electrocardiography demonstrates high diagnostic accuracy and substantial agreement with conventional ECG for rhythm and conduction assessment. Improved specificity in asymptomatic individuals suggests that multivector interpretation may mitigate overdiagnosis associated with single-lead wearable screening. This framework may represent a paradigm shift for portable ECG assessment and provides a scalable foundation for future three-dimensional vector reconstruction supported by artificial intelligence.Concept 3D reconstruction