Interval-Level Validation of a Wearable Dry-Electrode ECG Biosensor Platform: A Proof-of-Concept Study

Figure 1. Schematic of an ECG waveform shows the principal deflections (P wave, QRS complex, and T wave) and the intervals derived from them. The RR interval and the P-to-R peak interval, the two parameters analyzed in this study, are highlighted.

Figure 2. System overview. (A) Components of the X-trodes wearable ECG system: the flexible electrode array and the wireless data acquisition unit (DAU). (B) Schematic of the experimental setup illustrating concurrent placement of the wearable patch on the anterior chest and the wired reference system (Biopac) Ag/AgCl electrodes.

Figure 3. Study workflow. Data acquisition (top): simultaneous 3-min recordings from 20 participants with the wearable X-trodes (Fs = 4000 Hz) and wired Biopac (Fs = 1000 Hz) systems, followed by temporal synchronization. ECG analysis pipeline (bottom): preprocessing, wavelet-based R/P detection, manual adjudication, interval computation, and agreement analysis.

Figure 4. Representative simultaneous ECG recordings. The figure shows 10-s traces from the wearable (X-trodes; upper) and wired reference (Biopac; lower) systems in (A) a healthy and (B) a cardiac patient. In these representative recordings, both systems captured visually concordant beat timing and the major waveform components (P wave, QRS complex, and T wave) in the healthy volunteer, and the more complex morphology associated with cardiac disease in the patient. Amplitude in µV, time in seconds.

Figure 5. Bland–Altman agreement and correlation analysis for RR intervals. (A) Correlation plot. Each point represents a paired RR interval measurement for the healthy (teal) and cardiac (grey) cohorts. The Pearson correlation coefficients exceeded 0.99 for both groups (r = 0.99 for healthy patients; r = 1.0 for cardiac patients). (B) Bland–Altman plot showing the difference between systems (y-axis) vs. their mean (x-axis); horizontal lines indicate mean bias and the 95% limits of agreement. The marginal histograms and density curves show the distribution of measurement differences. No significant between-group differences were observed (p > 0.05 for all metrics).

Figure 6. Bland–Altman agreement and correlation analysis for P-to-R peak intervals. (A) Correlation plot. Each point represents a paired P-to-R peak interval measurement for the healthy (teal) and cardiac (grey) cohorts. The Pearson coefficients exceeded 0.92 for both groups, indicating a consistent linear relationship. (B) Bland–Altman plot showing inter-system differences (y-axis) against their mean (x-axis), with horizontal lines for the mean bias and the 95% limits of agreement. The marginal histograms and density curves show the distribution of differences. The between-group bias and correlation did not differ significantly (p > 0.05); the MAE differed significantly (p = 0.040), and the higher MAE in the cardiac cohort was consistent with greater P-wave delineation variability in morphologically complex ECG signals.