VISIBILITY OF PREMATURE BEATS IN PHOTOPLETHYSMOGRAPHY SIGNALS

VISIBILITY OF PREMATURE BEATS IN PHOTOPLETHYSMOGRAPHY SIGNALS

Abstrakt

Photoplethysmography (PPG) has recently emerge d as a promising non -invasive tool for cardiac arrhythmia ( CA) detection, particu larly when integrated into wearable or smart devices, offering continuous and accessible cardiac monitoring [1]. However, premature beats (PBs) do not always produce clearly observable peaks in PPG signals [2], which can complicate their detection and, consequently the identification of CAs. To investigate this, we conducted a study with 20 subjects in which PBs were induced by pacing from implantab le cardioverter-def ibrilla tor or pac emaker at coupling in tervals ranging from 300 to 500 ms in 50ms increments. During each m easurement protocol, both PPG signals and reference ECG recordings were collected. The pa cing protocol was struc tured as fol lows: to induce premature atrial contractions (PACs), doub le and triple pacing stimuli were delivered to right atrium across all coupl ing interva ls from 300 to 50 0 ms. In additi on, quadruple and quintuple stimuli were applied at coupling in ter vals of 450 ms and 500 ms, respectively, to induce supraventricular tachycardia . The same sequence was then performed for premature ventricular contract ions (PVCs), with 2× –3× pacing stimuli delivered to right ventricle across 300–50 0 ms coupling intervals, followed by 4× and 5× stimuli at 450 ms and 500 ms to induce ventricular tachycardia . In total, th is protocol yi elded 280 measurements. Each measurement was assigned a binary visibility label based o n the following cr iterion: i f all PPG peaks corresponding to all PBs were clearly ident ifiable, visibility was defined as 1 , o therwise, it was defined as 0. Some measurements were excluded from ana lysis because pacing at very short coupling intervals failed to propagate in the myoc ardium or PPG recordings were corrupted by motion arti facts. Analysis of the resulting PPG da ta showed tha t t he proportion o f visible PBs increased with longer coupling intervals, underscoring the impo rtance of identi fying the threshold at which reliable detection is achieved. Let p denote the p roportion of measurements with all PB peaks visible out of the to tal number o f measurements at a given coupl ing interval . At 300 ms, none of the 22 measurements showed all PBs visible (p = 0, 95% CI [0.0–0.15]); at 350 ms, the proportion increased to p ≈ 0.59 ( 95% CI [0.39– 0.76], 29 measurements); at 4 00 ms, it rose further to p ≈ 0.8 6 (9 5% CI [0.7 1–0.95 ], 3 7 measu rements); and at 450 ms and 5 00 ms, al l PBs were visible in all measurements (77 and 7 8, respectively). These findings indicate tha t PB detection in PPG can be considered reliably achievable from coupling intervals of 400 ms and above, providing quanti tati ve guidance for interpr eting and applying PPG -based CAs monitoring in clinical and research settings.

Photoplethysmography (PPG) has recently emerge d as a promising non -invasive tool for cardiac arrhythmia ( CA) detection, particu larly when integrated into wearable or smart devices, offering continuous and accessible cardiac monitoring [1]. However, premature beats (PBs) do not always produce clearly observable peaks in PPG signals [2], which can complicate their detection and, consequently the identification of CAs. To investigate this, we conducted a study with 20 subjects in which PBs were induced by pacing from implantab le cardioverter-def ibrilla tor or pac emaker at coupling in tervals ranging from 300 to 500 ms in 50ms increments. During each m easurement protocol, both PPG signals and reference ECG recordings were collected. The pa cing protocol was struc tured as fol lows: to induce premature atrial contractions (PACs), doub le and triple pacing stimuli were delivered to right atrium across all coupl ing interva ls from 300 to 50 0 ms. In additi on, quadruple and quintuple stimuli were applied at coupling in ter vals of 450 ms and 500 ms, respectively, to induce supraventricular tachycardia . The same sequence was then performed for premature ventricular contract ions (PVCs), with 2× –3× pacing stimuli delivered to right ventricle across 300–50 0 ms coupling intervals, followed by 4× and 5× stimuli at 450 ms and 500 ms to induce ventricular tachycardia . In total, th is protocol yi elded 280 measurements. Each measurement was assigned a binary visibility label based o n the following cr iterion: i f all PPG peaks corresponding to all PBs were clearly ident ifiable, visibility was defined as 1 , o therwise, it was defined as 0. Some measurements were excluded from ana lysis because pacing at very short coupling intervals failed to propagate in the myoc ardium or PPG recordings were corrupted by motion arti facts. Analysis of the resulting PPG da ta showed tha t t he proportion o f visible PBs increased with longer coupling intervals, underscoring the impo rtance of identi fying the threshold at which reliable detection is achieved. Let p denote the p roportion of measurements with all PB peaks visible out of the to tal number o f measurements at a given coupl ing interval . At 300 ms, none of the 22 measurements showed all PBs visible (p = 0, 95% CI [0.0–0.15]); at 350 ms, the proportion increased to p ≈ 0.59 ( 95% CI [0.39– 0.76], 29 measurements); at 4 00 ms, it rose further to p ≈ 0.8 6 (9 5% CI [0.7 1–0.95 ], 3 7 measu rements); and at 450 ms and 5 00 ms, al l PBs were visible in all measurements (77 and 7 8, respectively). These findings indicate tha t PB detection in PPG can be considered reliably achievable from coupling intervals of 400 ms and above, providing quanti tati ve guidance for interpr eting and applying PPG -based CAs monitoring in clinical and research settings.

Photoplethysmography, Premature beats, Coupling intervals

Photoplethysmography, Premature beats, Coupling intervals

ISBN 978-80-248-4828

Setkání vysokoškolských pracovišť v rámci konference Trendy v biomedicínském inženýrství 2025 a řešitelů projektu LERCO